This profile draws on an exclusive interview conducted at Yonghe’s Beijing headquarters in March 2025, alongside the company’s newly released FY2025 annual results.

The Founder: Twenty Years, One Industry

Zhang Yu is an unusual figure in China’s consumer healthcare sector. Most founders of his generation built their companies by riding a single market wave — and cashed out when the wave crested. Zhang has stayed, and more importantly, he has kept changing.

“Yonghe has survived twenty years by constantly reforming itself,” he told Pandaily. “Any business model is a product of its era. The real challenge isn’t building a successful company. It’s staying successful as the market evolves.”

He traces the company’s early survival to a bet he made against the industry’s prevailing logic. Where competitors poured money into marketing and packaged proprietary surgical “technologies” — a practice he bluntly calls 包装化 (bāozhuāng huà, or excessive packaging) — Zhang pushed Yonghe toward clinical standardization. “Medical packaging creates a house of cards,” he said. “You claim generation one, then you have to claim generation two, then generation eight, and eventually you have nothing left to claim.”

That instinct sharpened into a formal strategy in 2022, when Zhang initiated what he describes as the company’s deepest self-reinvention: stripping out sales-led culture, de-emphasizing surgical technology marketing, and rebuilding around a tiered physician system — where doctors are ranked by experience and clinical record, and patients choose and pay accordingly, much like choosing between a resident and a senior specialist at a public hospital.

“When we started physician tiering, it was painful,” he admitted. One clinic that previously ran on a single do-everything doctor now needs two or three, each at different price points. Short-term cost pressure was significant. But the system, now maturing, is becoming Yonghe’s structural moat: experienced doctors are tied to the platform because the tiering system rewards their accumulated clinical data. Young surgeons choose Yonghe because the platform offers a career path that independent clinics cannot.

The Financials: A Turnaround Validated

For investors, the FY2025 annual results released on March 31, 2026 tell a clean turnaround story.

Revenue was essentially flat year-on-year at RMB 1.808 billion (+0.2%), but the composition shifted meaningfully. Hair transplant volumes grew 19.7% to 71,380 procedures, while average revenue per procedure declined from RMB 22,306 to RMB 19,265 — reflecting a deliberate price-band widening strategy Zhang describes as necessary to capture a broader patient base without sacrificing premium positioning at the top end.

The more striking numbers are on the cost and margin side. Gross profit grew 10.1% to RMB 1.194 billion, and the gross margin expanded from 60.1% to 66.0% — a nearly six-percentage-point improvement driven by clinic footprint optimization, digital and intelligent upgrades, management model restructuring, and marketing efficiency improvements. Sales and marketing expenditure fell to RMB 806 million from RMB 902 million, as the company’s shift away from paid acquisition toward content marketing and retention began to flow through.

The net result: a swing from a RMB 226.6 million net loss in FY2024 to a RMB 73.6 million net profit in FY2025. EBITDA rose 158% to RMB 363.7 million. The company ended the year with RMB 749 million in cash and zero bank debt, and has proposed a maiden dividend of RMB 0.076 per share.

“A lot of people looked at our 2024 results and saw a company cutting costs to survive,” Zhang said. “What they missed was the structural improvement underneath — the margin expansion, the conversion rate improvement, the cash generation. The surface numbers looked bad. The architecture was getting better.”

The Global Observation: A Founder Who Did His Homework

The most revealing part of our conversation was Zhang’s account of his international research over the past eighteen months. He has visited Turkey, South Korea, Japan, Thailand, and the United States — not on industry junkets, but as a deliberate competitive intelligence exercise, in some cases posing as patients to observe clinic operations firsthand.

His observations are sharp and, for a Western investor trying to understand the global hair restoration market, worth dwelling on.

Turkey surprised him. Istanbul’s hair transplant cluster — he says the city has 600–700 clinics — is a genuine global hub, drawing patients from the US, UK, France, and across Europe. The economics are built on three pillars: government support for medical tourism, Istanbul’s geographic position as an 8-hour flight from most of the world, and labor costs that keep a full hair transplant procedure priced around USD 3,000 — comparable to China and a fraction of US pricing.

But the Turkish model has a structural ceiling. Zhang commissioned a detailed legal and market study from a Big Four firm after his visit. The report confirmed something he had suspected: roughly 50% of Istanbul’s hair clinics operate without proper medical licenses, and the majority are single-location owner-operated shops with no scalability. “Turkey built the world’s biggest hair transplant destination, but almost nobody there has built a chain,” he said. “That’s actually an opportunity.”

South Korea impressed him with its depth of specialization. He visited clinics that refused to serve male patients entirely, focusing only on female hairline aesthetics. Others had narrowed their entire practice to eyebrow transplants. “Korea is intensely competitive, so practitioners have been forced to specialize in ways China hasn’t yet reached,” Zhang observed. He sees this as a preview of where Yonghe’s own clinical segmentation strategy is heading — Yonghe has already begun building a dedicated female aesthetics division, with 30 cities now hosting women-only consultation zones.

Japan is essentially a non-market for surgical hair transplants: cultural preference for wigs, particularly among men, and a high-quality domestic wig industry (he cited Aderans as a dominant player) means demand for surgical intervention is minimal.

The United States has the opposite problem: premium pricing. An average US hair transplant costs USD 10,000–20,000, driven by labor costs that make the surgical economics almost unworkable at scale. “In the US, hair transplantation is still a luxury,” Zhang said. “In China it’s becoming a mainstream elective procedure.”

His summary: “If you’re measuring by total procedure volume, China leads the world. If you’re measuring by international patient draw, Turkey leads. If you’re measuring by clinical refinement, Korea leads. Nobody has put all three together.”

The International Strategy: Two Steps, Timed Differently

Zhang’s global ambitions are real but strategically sequenced. He is explicit about this: Yonghe is currently in Step One, and Step Two involves a different capital model and timeline.

Step One: Draw international patients into China. This is already underway. Yonghe currently derives approximately 3–4% of revenue from patients from outside the Chinese mainland (including Hong Kong, Taiwan, and Macau) and international markets like Singapore. Zhang expects this to exceed 5–6% in 2026. The company is advertising on YouTube, Google, and Facebook, targeting overseas Chinese communities as the primary near-term segment — people who travel back to China regularly and can combine a procedure with a family visit. He is also evaluating Hainan’s Boao special zone, where post-free-trade-zone opening, over 80 countries now have visa-free access, as a potential hub for international medical tourism.

Read more

At EAIDC 2026 in Shenzhen, X Square Robot CEO Wang Qian outlined the company’s push toward scalable embodied AI, predicting that general-purpose robots could eventually operate even in extreme environments like Mars.

On March 30, 2026, in Shenzhen, a quiet but telling milestone took place in the embodied AI world. X Square Robot (自变量机器人) hosted the inaugural Embodied AI Developers Conference — EAIDC 2026 — billed as the world’s first large-scale gathering dedicated specifically to developers building embodied AI systems. The event brought together researchers, engineers, and technology companies from across the industry for live robotic demonstrations, a national-level hackathon, and focused discussions on a topic that most robotics conferences still treat as futuristic: not whether robots can be intelligent, but how to deploy them at scale in the real world.

That framing — from can it work to how do we ship it — is precisely what distinguishes X Square from most companies in this space. While competitors have spent the past two years perfecting demo videos of robots doing backflips, X Square has been obsessively focused on a harder, less photogenic problem: building a foundation model general enough to actually run in a real home environment, a messy and variable space without failing the moment something unexpected happens.

EAIDC was, in some ways, a coming-out party. X Square is no longer a well-funded startup operating in the background. It is increasingly the company setting the terms of the conversation. Following the conclusion of the conference, Wang Qian(王潜), X Square’s founder and CEO, sat down with Pandaily for an exclusive interview to discuss the road ahead.

What X Square Actually Builds

To understand X Square, it helps to understand what problem it is trying to solve.

Traditional industrial robots are essentially very precise, very fast machines that follow pre-programmed instructions. They excel in structured environments — the same task, the same object, the same position, repeated thousands of times. Change any variable and they fail. They have no ability to observe the world, reason about it, and adapt.

The company’s answer is what it calls an Embodied Intelligence Foundation Model — specifically, a Vision-Language-Action (VLA) model that processes sensory input (video, language, haptic signals) and outputs physical action (joint torques, velocities, poses) in a single end-to-end architecture. The model, branded the Great Wall series with its flagship called WALL-A, is designed to enable robots to perceive, decide, and act autonomously across diverse environments and tasks without being explicitly reprogrammed for each one.

X Square claims WALL-A is currently one of the world’s largest-scale unified embodied foundation models — achieved via a novel three-stage training paradigm, a proprietary architecture combining shared attention with expert-routed feed-forward modules, and training data drawn from a mix of self-collected real-world demonstrations, open-source robot datasets, and automatically generated multimodal data.

In September 2025, X Square open-sourced a developer-facing version of its model called WALL-OSS, which has since been integrated into Hugging Face’s LeRobot framework — a move that simultaneously builds community goodwill and feeds the data flywheel. EAIDC itself can be read as an extension of this strategy: by convening the ecosystem, X Square is positioning its open-source model as the platform developers build on top of.

On the hardware side, X Square produces two robot platforms: Quanta X1, a wheeled dual-arm robot designed for precision manipulation in service environments, and Quanta X2, a wheeled humanoid launched in August 2025. It also makes the Artixon Hand, a 20 DoF dexterous robotic hand — which the company claims is the world’s first such hand to be controlled directly by a foundation model rather than traditional motion-planning algorithms.

The company is headquartered in Shenzhen with offices in Beijing and Shanghai, and has deployed its robots in hotel groups, logistics companies, elderly care facilities, supermarkets, and research institutions. Early commercial revenue is coming in from education, hospitality, and elder care, with household services — including a collaboration with classifieds platform 58.com — currently active.

The Founder: A Transformer Pioneer Who Quit Finance to Build Robots

Wang Qian is not a typical robotics entrepreneur. His academic background is in deep learning — specifically, he was among the earliest researchers in the world to introduce the attention mechanism into neural networks, publishing work at the same conference as Google’s early attention paper in 2014, three years before the Transformer architecture would reshape the entire field of AI.

After completing his bachelor’s and master’s degrees at Tsinghua University, he pursued a PhD at the University of Southern California, conducting robotics learning and human-robot interaction research at top American robotics laboratories. But he pivoted mid-career — to quantitative finance. He founded a quant fund in the United States, which by his own account was financially successful. And then he spent many nights unable to sleep.

“I kept thinking: I should have stayed in robotics,” Wang has said in interviews. In 2023, he dissolved the fund and returned to China.

His co-founder and CTO, Wang Hao (王昊), brings a complementary profile: a PhD in computational physics from Peking University, he previously led the algorithm team for the Fengshenbang large language model at the Institute for Intelligent Computing (IDEA Research), overseeing development of China’s first hundred-billion-parameter foundation model and one of the earliest trillion-parameter models, Ziya.

Together, the two founders embody X Square’s core thesis: that building a general-purpose robot brain requires expertise in both large foundation models and robotics learning — and that the two cannot be solved in isolation.

“We Were the Only Variable”

The name “X Square” is not arbitrary. In mathematics, x represents the independent variable — the thing that changes and drives outcomes. The Chinese name, 自变量 (zì biànliàng), carries the same meaning with an additional nuance: zì means “self” or “autonomous.” The company wants to be the variable that changes the world — and it wants that change to be self-generated, not dependent on others.

This framing turns out to be an accurate description of how X Square navigated its early years.

When the company was founded in late 2023, the embodied intelligence landscape in China was already competitive. Galbot (银河通用) and Agibot (智元机器人) had both launched that same year with considerable fanfare, larger teams, and more initial funding. X Square, by contrast, operated with little public visibility and struggled to raise early rounds.

“The biggest difficulty was that nobody trusted us,” Wang Qian has recalled. “Early investors didn’t believe we could actually pull this off. They bet on the team being decent enough that even if we failed at this, we’d find something else to do.”

What changed was output. X Square released its first embodied intelligence model just two months after founding. By October 2024, it had trained WALL-A — at the time, the largest-parameter general-purpose embodied manipulation model in the world. The demos showed something qualitatively different: robots that could hang laundry, prepare shaved ice, wind cables around pegs, and sort parcels of arbitrary shape, all controlled by a single model without task-specific reprogramming. By the time Physical Intelligence (PI), the American benchmark company in this space, released its π0 model in late 2024 — validating the end-to-end VLA approach — X Square had already been on that path for over a year.

“We didn’t need to copy anyone’s homework,” Wang has said. “We were already there.”

By early 2025, the funding picture had transformed entirely. Meituan led a Series A round; Alibaba Cloud and Hongshan (formerly Sequoia China) co-led an A+ round of nearly RMB 1 billion; and by January 2026, a 1 billion yuan A++ round closed with ByteDance, Hongshan. The company has now raised approximately $280 million in total — and is the only Chinese embodied intelligence startup to have simultaneously attracted backing from Meituan, Alibaba, and ByteDance.

The Technical Bet: Why End-to-End Matters

The central debate in embodied intelligence — and the one X Square has wagered its existence on — is architectural. Should a robot’s “brain” be built as a modular stack (perception module → planning module → control module), or as a single end-to-end model that processes raw sensor input and outputs raw motor commands without explicit intermediate representations?

X Square’s answer has been unambiguous from day one: end-to-end, always.

Wang Qian’s reasoning is both technical and philosophical. Modular systems, he argues, impose human-designed abstractions onto a continuous physical world. Each handoff between modules introduces error, latency, and brittleness. The physical world doesn’t segment neatly into “perception” and “planning” — it is a continuous stream of force, contact, spatial relationships, and temporal dependencies. A model that learns to navigate that stream as a whole, rather than as a sequence of sub-problems, will generalize better and fail more gracefully.

“Embodied intelligence is a foundation model for the physical world — independent from, and parallel to, language models and multimodal models for the virtual world,” Wang said at the 2025 MEET conference. “The long-term advancement of robotics intelligence depends on general-purpose AI capability.”

The practical implication is that WALL-A takes language, video, and haptic sensor data as input and outputs velocity, pose, and torque directly — with no intermediate planning layer. The same model controls the company’s wheeled robots, its dexterous hand, and in principle any robotic platform it is trained on. In July 2025, X Square demonstrated what it claims was an industry first: a foundation model controlling a high-degree-of-freedom dexterous hand to perform complex manipulation tasks, including card dealing, without task-specific engineering. The company has also integrated chain-of-thought reasoning natively into the model — enabling robots to break down complex instructions into subtask sequences and reason through them before acting.

Where the Robots Are Actually Working

Read more

There’s a demonstration that Zhou Yong likes to show people who have never thought seriously about robotic hands. He asks them to count how many things they touched in the first sixty seconds after waking up this morning. The alarm. The phone screen. A glass of water. A light switch. The toothbrush. Before most people have finished their morning routine, they’ve executed dozens of complex, multi-fingered grasps that no robot in history has been able to fully replicate.

“The hand,” Zhou told me, “is the last great hardware problem in robotics.”

Zhou Yong — known in English as Alex Zhou — is the founder and CTO of Linkerbot, a company that since 2023 has been building what it describes as some of the most dexterous commercial robotic hands on the market today. At a time when humanoid robots are everywhere in the news and every major tech company from Tesla to Figure AI is racing to build a walking, working machine, Linkerbot has been quietly solving the part of the problem that everyone else has mostly given up on: the hand.

The Problem With Robot Hands

To understand why this matters, it helps to understand what a “degree of freedom” actually means in the context of a robotic hand. Roughly speaking, each degree of freedom is an independent axis of motion — a way in which a joint can move. A human hand has approximately 27 of them. Most commercial robotic grippers have between 1 and 6. The gap between those two numbers is the gap between a robot that can pick up a box and a robot that can tie your shoes.

For decades, the gold standard for highly articulated robotic hands was the Shadow Hand, made by British company Shadow Robot. The Shadow Hand has 24 independently controlled joints and costs roughly between $140,000–$280,000 USD. It is extraordinarily capable and extraordinarily expensive. Tesla’s Optimus robot hand has 22. Both represent the upper ceiling of what global robotics companies have been able to commercialize at any price.

The company’s published and reported product lineup spans up to 42 degrees of freedom, while its more widely discussed commercial products include the L20, a 21-DOF dexterous hand publicly reported at roughly one-twentieth the cost of comparable overseas products.

And critically, Linkerbot is mass-producing them. While overseas competitors are still grappling with load capacity, battery life, and structural durability at the engineering scale, Linkerbot delivered over 10,000 units in 2025. Its products are in use at Cambridge University, Stanford University, Peking University, and Tsinghua University, among dozens of institutional customers.

“We are,” Zhou says with characteristic directness, “the only company in the world that can actually make these at scale.”

The Founder: From Three Hundred Million Users to Robotic Fingers

Zhou Yong’s path to robotic hands is not a straight line — which turns out to be one of his key advantages.

He graduated from the Youth Class at Huazhong University of Science and Technology at age 14, a selective program for academically exceptional teenagers that has produced a disproportionate share of China’s tech founders. He spent the next decade accumulating experience across two domains that almost never overlap: global online communities and embodied intelligence. The dual background gave him something rare in the hardware world — an obsessive focus on the end user alongside rigorous engineering thinking about what is actually achievable.

Before Linkerbot, Zhou’s previous company reached 300 million users — virtually none of them in China. “You can think of me as a Silicon Valley-style company,” he told me. “Less than 1% of my users were Chinese. Three hundred million users, all overseas.” This experience gave Zhou not just a global mindset — he genuinely thinks in terms of international markets from day one — but also a hard-won understanding of what it takes to build at scale for users who have no patience for friction.

“My way of thinking is no different from Silicon Valley,” he said. “I try to see things as they are.”

That perspective shapes how he talks about the global robotics race. Rather than framing the field purely as a national contest, he often returns to questions of manufacturing depth, data accumulation, and skill deployment.

The idea for Linkerbot crystallized around 2018–2019, when Zhou began watching the early stirrings of what would become the embodied intelligence wave. Humanoid robots were still largely academic curiosities, but he could see the trajectory. The fundamental bottleneck, in his reading, was not actuators or walking algorithms or even large language models. It was the hand. He founded Linkerbot in 2023 with a clear and, at the time, audacious mandate: build a dexterous hand that surpasses Shadow Hand in performance, and sell it at a fraction of the price.

From day one, Zhou refused to take the incremental path. “I told the team: our first mass-produced product has to have 20 degrees of freedom. Not 6, not 10. 20. If we’re going to do this, we do it right.”

The Technology: Three Roads, One Destination

What makes Linkerbot’s engineering approach genuinely distinctive is that it has not bet on a single mechanical architecture. Instead, the company has built capabilities across all three mainstream dexterous hand technologies — linkage transmission, tendon drive, and direct drive — a breadth that, according to company disclosures cited in the media, no other company currently matches.

The first is linkage transmission, which uses rigid mechanical linkages between joints. This approach offers high rigidity, strong gripping force, and precise control. Each finger in a linkage-driven hand is an independent module that can be replaced or upgraded individually — a significant advantage for industrial deployment where maintenance downtime is costly. The Linker Hand O6, L6, L20 Lite, and L20 all use this architecture.

The second is tendon drive, which mimics the way human tendons pull on finger bones to create motion. This architecture allows for smoother, more naturalistic movement and is inherently more compact — the reason Shadow Hand, Tesla’s Optimus, and Boston Dynamics’ Atlas all use variants of it. The Linker Hand L30 uses this approach, achieving ±0.20mm repeat positioning accuracy, a maximum movement speed of 440°/second, and a peak fingertip force of 3.5N per finger with a maximum thumb force of 4N — all from a hand weighing just 1.4kg.

The third is direct drive, which emphasizes transmission efficiency, control precision, and fast dynamic response by reducing intermediate mechanical complexity. It is particularly valuable in scenarios that demand high responsiveness, precision, and consistency under repeated operation. Linkerbot’s product portfolio also includes hands built on this architecture.

“We didn’t choose,” Zhou explains. “Different applications need different trade-offs. Industrial scenarios may prioritize rigidity and load capacity. Research and biomimetic tasks may value compactness and natural motion. Other scenarios demand high precision and responsiveness. So we built across all three.”

The current Linker Hand product lineup spans five models:

• O6– 11 joints, only 370g, <±0.22mm repeated positioning accuracy, 50kg maximum load.

• L6 — 11 joints, 623.5g, ±0.20mm repeat positioning accuracy, 0.35-second opening/closing response. Designed for precision assembly and irregular object grasping.

• L20 Lite — 20 joints, 800g, piezoresistive sensors standard. Designed for education, research, piano performance, household assistance, and elderly care.

• L20 — 21 joints (16 Active + 5 Passive DOF), 1.1kg, 18N maximum thumb force. Designed for industrial automation, household assistance, and complex multi-task environments.

• L30 — 25 joints (22 Active + 3 Passive DOF), tendon drive, 1.4kg, CAN FD protocol (up to 5Mbps). Designed for precision industrial tasks and medical assistance & healthcare applications.

The Sensory Layer: Teaching Machines to Feel

One of Zhou’s most emphatic points — one he returns to repeatedly — is that degrees of freedom alone don’t make a useful dexterous hand. What matters equally is what the hand can perceive and learn.

Linkerbot has also integrated a range of advanced sensors into its hands. Pressure sensors allow the system to detect and regulate gripping force with precision, reducing the risk of damaging delicate objects. Tactile sensors, designed to approximate the sensing function of human skin, help the hand perceive surface texture and shape, while the broader sensing system provides real-time feedback on force, form, and temperature. When handling fragile items, the hand can automatically adjust its grip. A high-precision sensor array enables millimeter-level sensing accuracy, and sensor-fusion technology brings these different streams of information together to form a more complete picture of the environment.

That is why Linkerbot frames its strategy not simply around the hand itself, but around what it calls “dexterous hands + a cloud brain.” In practice, this means the company is building not only the hardware, but also the data-collection and model-training system behind it. As the hand interacts with the physical world, it generates the manipulation data that can be used to improve how the system learns and performs.

At the center of this approach is LinkerSkillNet, which the company describes as a large-scale dexterous manipulation dataset built through extensive real-world and simulated learning. On top of that foundation, Linkerbot trains what it calls its cloud brain — with the goal of helping Linker Hand learn new actions faster, execute tasks more accurately, and adapt more efficiently to new scenarios.

The logic is simple, but powerful: the more tasks the system learns, the better it becomes at learning the next one. Linkerbot also says this cloud-based intelligence layer can be adapted across different robotic platforms, giving the system greater flexibility as its range of applications expands.

The Production Bet: Using Robot Hands to Build Robot Hands

There is a detail about Linkerbot’s manufacturing roadmap that Zhou shared in our conversation that hasn’t been widely reported, and it may be the most consequential thing about the company’s medium-term competitive position.

Linkerbot is currently producing more than 1,000 High-DOF dexterous hands per month. Within the near term, Zhou expects to scale that to 4,000 to 5,000 units per month. And the plan for how to get there is elegantly recursive: he intends to build an automated production line that uses robotic hands to assemble robotic hands.

“Once that production line is operational,” Zhou told me, “the barrier to entry becomes extremely high. Think of it this way: one production line is equivalent to a cluster of 1,000 robots working in parallel. Right now, there are almost no companies in the world that can operate a cluster of 1,000 robots. We’ll be building that capability into our manufacturing infrastructure.”

The Market Bet: 1 Million Hands, 2 Billion Robots

Zhou has a theory about how the embodied intelligence revolution will actually unfold, built around a specific number: one million.

“The key question is not when we’ll have better AI models,” he explains. “The key question is where the training data for embodied AI comes from. You can’t synthesize your way to a robot that can work in the real world. You need real-world data from real robot hands doing real tasks.”

His bet is that the deployment of one million dexterous-hand-equipped robots into real environments — homes, factories, hospitals — will generate more manipulation training data in a single day than was collected globally for all of 2025. That data, fed into cloud-based learning systems, will produce the generalization capability that current embodied AI lacks.

Once the generalization problem in robotic manipulation is solved, Zhou predicts that the humanoid robot market will scale from current levels to more than 100 million units, and within three to five years after that, to 2 billion. He delivers this not as aspiration but as engineering calculation.

The Cultural Dimension: Skills as Civilization

There’s a dimension to Zhou’s thinking about dexterous hands that you don’t hear from most robotics founders, and it reveals something important about how he frames the mission internally.

He talks about hands not just as mechanical end effectors, but as carriers of human skill. When Elon Musk talks about Optimus, he has pointed to piano playing as a benchmark for dexterous manipulation. Zhou’s frame is broader: a truly capable hand, in his view, should not be limited to a single benchmark task, but should be able to learn across a much wider repertoire — from instruments and tools to crafts and everyday forms of embodied knowledge.

That idea is not merely cultural. It is also technical. Different industries, traditions, and working environments contain different kinds of fine-motor tasks, and those tasks can become valuable training data for embodied intelligence. A company operating close to dense manufacturing systems and diverse real-world workflows may therefore gain access to a wider range of manipulation scenarios.

This is the other side of the generalization argument: not just more data, but richer data — drawn from more tasks, more contexts, and more forms of human practice. “The hand is the mapping of skill data,” Zhou told me. “And the dexterous hand is the projection of civilization. We have five thousand years of craft traditions to draw on. That’s our biggest advantage.”

The Road Ahead: Industrial Pivot and the Global Market

In 2025, Linkerbot made a decisive pivot toward industrial applications. The company launched its “Industrial Master” series — the L6 Industrial and L20 Industrial — specifically designed for semi-structured factory environments. Both use a newly developed “ultra-strong electric cylinder” drive module that achieves 90% drive efficiency — more than double the traditional benchmark — and integrates drive and control functions within the space of a human palm.

The pivot to industry is not a retreat from the original vision. It’s a recognition that factories are where humanoid robots will actually deploy first, where the data collection can begin at scale, and where the business model is clearest. Semi-structured environments with repetitive precision tasks are the ideal first proving ground for highly articulated dexterous hands.

On the pricing front, the L20 is positioned as an option for customers seeking both high dexterity and strong cost-performance. The O6 is aimed more squarely at accessibility — designed with the goal of making dexterous hands affordable to students and embodied AI researchers alike. Zhou’s strategy reflects a broader logic: the faster dexterous hands proliferate, the faster data accumulates, and the faster embodied systems improve.

Linkerbot currently has offices in Silicon Valley and Canada in addition to its Beijing headquarters. Zhou recognizes that international brand awareness is still a work in progress. “We have a presence,” he said, “but our brand recognition overseas still needs work.”

Why Linkerbot Matters

It would be easy to read Linkerbot’s story as a simple cost story — a company making dexterous hands more affordable than legacy systems. But that reading misses the larger point.

For decades, highly articulated robotic hands have remained largely confined to the ivory tower of robotics. The Shadow Hand, developed in the 1990s, became one of the best-known benchmarks in the field: technically impressive, but also expensive and difficult to scale far beyond research and demonstration settings. The gap between high-end performance and broad commercial viability is precisely the gap Linkerbot is trying to close.

The broader significance lies in where this fits within embodied AI. Many robotics companies— Figure AI, 1X Technologies, Apptronik, Boston Dynamics, Tesla — are all working on the full-stack humanoid, but dexterous manipulation remains one of the hardest problems in the field. A capable hand is not just another component. It is the interface through which a robot performs useful work, collects data, and turns intelligence into action in the physical world.

“The world doesn’t need another simple gripper company,” Zhou said. “The world needs a company that redefines what a hand can be.”

Whether or not his 2-billion-robot prediction is right, the market data is already pointing somewhere interesting. Linkerbot’s 80% global market share in High DoF dexterous hands, achieved not through exclusion but through genuine performance advantage at commercial price points, is the kind of number that tends to create durable competitive positions. More importantly, Linkerbot has done so by positioning dexterous hands not as niche research hardware, but as products increasingly aimed at real deployment and broader adoption. If embodied AI is to move from demonstration to real-world use, that shift may matter more than many people yet realize.

Linkerbot is headquartered in Beijing, with offices in Silicon Valley and Canada. The Linker Hand series is available globally. Its products are used by leading universities, including the University of Cambridge and Stanford University, and have also attracted interest from major industrial customers such as Siemens and Samsung.

This article is the first in Pandaily's "Interview with 100 Founders: China's Next Wave" series.

On July 7, 2025, Hong Kong-based Waton Securities – officially known as Waton Financial Limited (NASDAQ: WTF) – rang the opening bell at the Nasdaq MarketSite to celebrate its recent initial public offering . The company’s ordinary shares began trading on April 1, 2025 , raising about $20 million at a $4 IPO price. Waton is a securities brokerage and financial technology services firm, originally founded in 1989 in Hong Kong . Through its subsidiaries, it provides stock brokerage services (primarily for Hong Kong-listed securities) and also offers software and tech solutions to other financial institutions .

Today, Waton Securities remains relatively small in scale – it recorded roughly $10 million in revenue for the year ending March 2024 with about $2.5 million in net profit . Its market capitalization after the IPO has hovered around $250 million . Yet the company has outsized ambitions under the leadership of its young chairman, Kai Zhou, who acquired Waton Securities in 2021 and now holds a controlling stake. At just 29 years old, Zhou is among the first of the post-95 generation of Chinese entrepreneurs to helm a Nasdaq-listed firm . As both the architect of Waton Securities’s strategic transformation and its public face, Zhou’s vision is to leverage cutting-edge technology – especially artificial intelligence – to reinvent what a brokerage can be.

From Investor to Operator

Kai Zhou’s path to running a brokerage has been unconventional. After studying in China and Canada, Zhou began his career in investment. “I actually returned to China in 2017 and did a year of private equity investment,” he told Pandaily. This was the height of China’s internet finance boom, with startups in peer-to-peer lending and online finance attracting big funding. Zhou observed the frenzy but also recognized the pitfalls. “Internet finance was hot, but I knew there were many hidden risks and regulatory uncertainties,” he recalls . Wanting to participate in the sector’s growth without taking on its full risks, he decided to become a “picks and shovels” provider. In 2018, Zhou founded a financial SaaS (Software-as-a-Service) company, offering technology development and outsourcing services to fintech platforms . This venture proved successful – it gave Zhou his “first bucket of gold” and a network of clients that included numerous online lending and internet brokerage firms .

By 2019, a bull market in Hong Kong stocks was underway and IPO trading was especially popular . Zhou watched his brokerage clients thrive, and a new idea took shape: rather than just sell shovels, why not own a gold mine? “I come from a finance background, so in 2020 I thought, why not acquire a brokerage myself?” Zhou says. He identified a long-established Hong Kong brokerage – Waton Securities – as a suitable vehicle. In 2021, at the age of 26, Zhou acquired Waton Securities, executing a search-fund style entrepreneurship-through-acquisition . This move turned the investor into an operator, putting him in charge of transforming a traditional brokerage business.

Taking over an aging financial firm came with challenges. Industry observers often say “it’s harder to be a ‘takeover guy’ than to start a company from scratch,” and Zhou would agree the integration wasn’t easy . He became the CEO of an existing team with its own legacy culture, many of whom were older than him. Rather than enforce sweeping changes overnight, Zhou focused on winning the team’s trust and gradually aligning them with his vision. “I prefer to frame it as the team accepting me, rather than me changing the team,” he explains . Over time, as colleagues understood his ideas, everyone began “pulling in the same direction,” reducing friction and boosting efficiency . Zhou also wasn’t shy about investing in the business to earn buy-in: “You have to keep pouring in money – if there’s no milk, I provide the milk,” he jokes of his management strategy . With patience, open communication, and fresh capital, the young CEO revitalized Waton’s culture and prepared it for a new era.

AI as a Trading Partner

That new era arrived in 2023, when a technological breakthrough changed Waton’s trajectory. Zhou recounts how the emergence of advanced generative AI made a profound impression on him. “In 2023, GPT-3.5 came out globally. I tried it and was completely blown away – I felt this is the future,” he told Pandaily . Seeing the rapid progress in large language models, Zhou became convinced that artificial intelligence would disrupt most industries, especially investment and trading . He and his team made a bold decision: to go all-in on AI and pivot Waton into an “AI-first” company.

Zhou’s enthusiasm for AI is grounded in his philosophy of what makes a great trader. Even the most skilled human investors are ultimately limited by cognitive biases and emotions. “A well-trained trader still, at times, trades on intuition,” Zhou observes. Humans naturally develop “biases and path dependencies” – for example, the reluctance to buy back a stock you once sold at a lower price, because doing so feels like admitting a past mistake . These psychological factors can hinder long-term performance, and few people ever fully overcome them . AI agents, however, need not suffer from such baggage. Zhou argues that an AI, given the right training, can “absorb a person’s past strategies to refine and strengthen them, while also incorporating other viewpoints without prejudice” . In other words, an AI can learn from human experience without inheriting the human biases that come with it. Moreover, AI systems have vastly superior computational power and can process far more information than any individual. Thanks to these advantages, Zhou is “absolutely convinced” that in the realm of trading and investment, AI will play a critical role – potentially surpassing human traders in performance . Rather than view AI as just a tool, he sees it as a new kind of market participant or partner. “The future belongs to AI,” Zhou says unequivocally .

This conviction led Waton Securities to embark on a sweeping AI-driven transformation. In practical terms, the firm started reorienting its products and services around the idea of AI-assisted and AI-executed trading. By the time of its Nasdaq debut, Waton was branding itself not just as a fintech-enabled brokerage, but as one of the first brokers “for AI”. “Our aspiration is to become a pioneer in offering brokerage infrastructure that supports AI-driven participants,” Zhou proclaimed at the listing ceremony . “We believe AI is emerging as a new category of economic agent, and we are exploring how financial institutions can one day support such entities responsibly, in parallel with human clients.” In Zhou’s view, algorithms and autonomous trading agents may soon join retail and institutional investors as a third type of client in financial markets .

Looking Forward

With its “AI-first” strategy, Waton Securities is positioning itself quite differently from more established online brokerages. Take Futu Holdings, for example, which is one of China’s leading digital brokerage firms. Competitors like Futu have certainly begun integrating AI features – such as smart chatbots or AI-driven analytics – but mostly as tools to enhance service for their human users. Kai Zhou is flipping that model on its head: Waton’s focus is on serving the AIs themselves. “I think it’s very different,” Zhou says of his strategy versus peers. “Our industry peers are mostly looking at how to use AI to provide better service to humans, whereas we are looking at how to use our existing infrastructure and team to provide better service to AI.” In concrete terms, Waton is building out the capabilities for AI clients to act autonomously on its platform. “For example, we want to let AI open brokerage accounts, let AI execute trades, provide AI with market data and backtesting for its trades,” Zhou explains . By empowering AI agents as traders, Waton aims to cater to the trading needs of these non-human clients. “Our target customer is different, so I believe our strategies are completely on two separate paths,” Zhou notes of the contrast with conventional brokers .

This bold vision leverages Waton’s existing strengths in B2B services. Beyond its own retail brokerage business, Waton has been offering a brokerage SaaS platform – a cloud-based service for other small and mid-sized brokerages – known as its “brokerage cloud.” Traditionally, this platform provided client firms with ready-made trading apps and back-end systems, along with access to trading and settlement in Hong Kong, U.S., and mainland Chinese markets (plus related services like margin financing) . Now, Zhou plans to infuse the brokerage cloud with AI capabilities as a differentiating feature. In the near future, Waton will invite AI trading models onto its platform – essentially hosting AI-run trading accounts that can operate under Waton’s brokerage license . The idea is that if Waton manages to cultivate a large number of high-performing AI traders, “legendary trading AIs” as Zhou calls them, their expertise can be leveraged by all the human clients of Waton’s partner brokerages . For instance, a small brokerage using Waton’s SaaS could offer its retail customers the ability to follow or mirror the strategies of these top AIs, or to receive AI-generated trading insights tailored to the Chinese markets. In Zhou’s words, the AIs’ trading “movements” and analysis can be passed downstream to end-users, providing benefits like enhanced research, automated trade reviews, or even personalized trading tutorials . In short, Waton seeks to be the platform that connects cutting-edge AI traders with everyday investors around the world.

Zhou often describes this goal as “helping the world invest in China.” In our interview, he outlined a two-step roadmap to realize that vision . First, Waton aims to onboard a critical mass of global brokerage firms onto its platform, especially targeting those smaller brokerages that lack strong in-house tech – thus extending Waton’s network and distribution. Second, the company will develop a stable of AI-driven trading accounts that excel at delivering returns from Chinese equities and other assets. “When we have enough AI traders that can generate excellent returns investing in China, then end-users and institutions across various regions can, through the brokers on our system, get access to those AIs,” Zhou explains . By combining those steps, an investor in, say, Southeast Asia or North America could use their local brokerage (if it’s a Waton cloud client) to tap into the expertise of AI models specialized in China’s markets. This strategy not only differentiates Waton from the likes of Futu, but also aligns with China’s broader fintech trajectory – integrating AI to open new channels between global capital and Chinese opportunities.

Now that Waton Securities is a public company, Zhou is aware that the market will be scrutinizing the execution of this ambitious plan. The stock’s volatile journey – an initial 400% surge on IPO day followed by a steep comedown – reflects both excitement and skepticism around the “AI brokerage” concept. Yet Zhou remains steadfast in his long-term conviction. He acknowledges that Waton’s path is unorthodox and “not a road with a lot of people on it” . Many investors may not immediately buy into the idea of AIs as clients, and that’s fine by him. “Whether the market or investors recognize our philosophy – we don’t care… or rather, we can’t cater to everyone’s opinion,” Zhou says matter-of-factly . Instead, he insists the company will focus on “continuing down our path and let the results speak for themselves.” It’s a statement of confidence from a CEO who has bet his career on this vision. Indeed, Zhou holds about 86% of Waton’s shares post-IPO , so his personal skin in the game is enormous. He isn’t looking for a quick exit; rather, he talks in terms of decades. “If you ask me 5 or 10 years from now, I definitely plan to be doing this for more than 10 years… maybe much longer,” he says. Zhou even jokes that “if you lined up everyone in the world who’s bullish on our company, I’d be standing at the very front of that line.” Such boldness is characteristic of China’s new generation of fintech entrepreneurs. As Waton Securities steps onto the global stage via its Nasdaq listing, the coming years will test whether Kai Zhou’s AI-powered brokerage can fulfill its promise and perhaps redefine the industry’s future.

XPeng Motors has officially launched its newest electric SUV, the G7, at a July 3 event in China – marking the company’s first new model of 2025. Billed as a “super AI family SUV,” the G7 arrives with cutting-edge autonomous driving tech, a spacious high-comfort design, and aggressive pricing intended to undercut rivals in the fiercely competitive mid-size EV segment. Below we break down the G7’s key features and launch highlights, provide context on XPeng’s recent performance, and analyze how this launch could shake up China’s EV market.

G7 Overview: Design, Specs and Smart Features

Design and Dimensions: The G7 slots between XPeng’s compact G6 crossover and larger G9 SUV, featuring a sleek, futuristic exterior influenced by aerospace elements. It adopts XPeng’s signature second-generation “X Face” front fascia – a clean, aerodynamic nose with a full-width LED light bar – and a wraparound ring-shaped taillight strip emblazoned with the XPENG logo. The body measures 4,892 × 1,925 × 1,655 mm (L×W×H) on a 2,890 mm wheelbase, making it slightly longer and taller than a Tesla Model Y. This yields a very roomy cabin – XPeng claims an 88% space utilization (“得房率”), with front and rear headroom about 45–53 mm greater than Model Y. Cargo capacity is family-friendly: 819 L trunk volume expandable to 2,277 L with rear seats down.

Inside, the G7 is appointed for comfort and tech-savvy convenience. Notably, all four seats come standard with ventilation, heating and massage functions – a rare luxury at this price point. A massive panoramic glass sunroof brightens the cabin, while rear privacy glass and extensive NVH sound insulation (including double-glazed front windows) keep the ride quiet. A 20-speaker Dynaudio sound system and 256-color ambient LED lighting add to the premium feel. XPeng has also paid attention to air quality and wellness, equipping the G7 with dual-zone automatic climate control (plus rear AC), a PM2.5 filtration system, an ionizer air purifier, and even an in-car fragrance diffuser for aromatherapy.

Powertrain and Performance: Unusually, the G7 launches with a single powertrain configuration – a rear-wheel-drive setup with a 218 kW permanent-magnet motor (450 N·m torque) on the rear axle. There is no dual-motor AWD option available (for now), which underscores the G7’s positioning as a practical family SUV rather than a performance crossover. Still, acceleration is brisk: 0–100 km/h in about 6.5 seconds. The chassis features a front double-wishbone and rear five-link independent suspension, augmented by adaptive damping shocks. XPeng touts its new “Taiji AI chassis” system that can scan 200 m of road ahead and adjust suspension within 300 ms, promising a smooth ride on varied surfaces.

Battery, Range and Charging: Two battery sizes are offered. The Long Range Max versions use a ~68.5 kWh LFP battery for 602 km CLTC range, while Ultra Long Range versions upgrade to an 80.8 kWh pack for 702 km CLTC range . The G7 is built on an 800 V high-voltage SiC platform and supports extremely fast charging at up to 5C rates. XPeng claims a 10% to 80% charge can be done in just 12 minutes, given a proper ultra-fast charger – roughly twice as fast as a Model Y, which needs about 24 minutes for a similar charge window. This rapid charging, combined with the ~700 km range, aims to alleviate EV range anxiety for road-tripping families.

“Super AI” Autonomous Driving Tech: The G7’s flagship feature is its emphasis on artificial intelligence and autonomous driving capabilities. XPeng has equipped the SUV with its latest in-house “Xtreme Intelligence” system, including the debut of XPeng’s self-developed Turing AI chips. The top-tier G7 Ultra trim carries 3 Turing chips delivering a combined 2,250 TOPS of computing power – purportedly making it the world’s first L3-level autonomous driving computing platform in a production car. (For context, Tesla’s Full Self-Driving computer is estimated around 720 TOPS and Xiaomi’s new YU7 SUV uses a 700 TOPS Nvidia chip.) Two of the Turing chips are dedicated to XPeng’s autonomous driving “brain,” while one powers the intelligent cockpit. Notably, the lesser Max trims of the G7 do not get Turing chips – they instead use two Nvidia Orin-X chips (~508 TOPS), which is still on par with most competitors’ ADAS hardware. All G7 variants come standard with an array of cameras and sensors (XPeng has actually dropped LiDAR in favor of a pure-vision approach, enabled by high computing power ).

On the software side, the G7 is the first to run XPeng’s full VLA + VLM large-model AI architecture entirely on-board (no cloud needed). The VLA-OL “driving brain” model uses a dual-“brain” paradigm (a big sports brain for decision-making plus a small brain for execution) to allow the car to “actively think, understand and react” more like a human driver in complex scenarios. XPeng claims this yields human-like judgment in tricky situations (avoiding road sinkholes, yielding to ambulances, etc.) and improves driver-assist performance by an order of magnitude. Meanwhile, the VLM vision model serves as an AI “cognitive engine” for the car, handling the cockpit interactions, voice assistant, and learning the user’s preferences. In practical terms, the G7 supports advanced XPILOT features including highway Navigate-on-AP, city driving assist, automated lane changes, and a new “human-car co-driving” mode where the system can hand off smoothly between human and AI control. Automatic parking and remote parking are also built-in, catering to users who struggle with tight parking spots.

Every G7 comes standard with a prominent AR-HUD (augmented reality head-up display) in place of a traditional instrument cluster. Co-developed with Huawei and branded “Zhuiguang Panoramic HUD,” this system projects a massive 87-inch virtual image onto the windshield. The AR-HUD can overlay navigation arrows, safety alerts, and driving assist visuals directly in the driver’s line of sight, making features like AR navigation and lane-change guidance more intuitive . The cabin also features a 15.6-inch central touchscreen (with a Qualcomm 8295P chipset for snappy infotainment), but notably no separate instrument panel – a design enabled by the AR-HUD. Voice controls are enhanced with four-zone far-field microphones and an AI assistant that can understand intent and remember user habits, thanks to the VLM model. Overall, XPeng is pitching the G7 as not just another electric SUV, but a “third space” infused with AI – providing what it calls “AI safety, AI comfort, and AI convenience” to young tech enthusiasts and modern families.

Launch Event Highlights: Pricing Shock and Executive Insights

XPeng unveiled the G7 in a high-profile launch event on July 3, where CEO He Xiaopeng took direct aim at the competition. The biggest surprise was pricing: the G7 is offered in three configurations – 602 Max, 702 Max, and 702 Ultra – starting at just ¥195,800 RMB ($27,000) for the base 602 Max. The mid-range 702 Max is ¥205,800, and the fully-loaded 702 Ultra (with the 3 Turing AI chips) tops out at ¥225,800 ($31k). These official prices elicited cheers at the event, as they came in ¥40,000 lower than the G7’s earlier presale starting price of ¥235,800. By aggressively pricing under the key ¥200k threshold, XPeng has positioned the G7 as one of the most affordable high-tech EVs in its class.

He Xiaopeng openly framed the G7 as an answer to Tesla’s dominance. He referenced Tesla’s recent slogan for its refreshed Model Y (“尽管对比” – “go ahead and compare”) and declared: “The 20–30 万元 segment has been bustling in the first half of this year – akin to several martial arts sects vying for supremacy. The kingly Model Y threw down the gauntlet with ‘compare all you want.’ Today, XPeng G7 has arrived.” On stage, He proceeded to compare the G7 head-to-head with Tesla’s Model Y (the reigning best-seller in China’s mid-size EV SUV market) and also with Xiaomi’s newly launched YU7 SUV:

• Space: The G7 offers a roomier cabin – He noted its front and rear headroom exceed Model Y by 45 mm and 53 mm respectively. This aligns with the G7’s slightly larger body dimensions and prioritization of family comfort.

• Range & Charging: With up to 702 km CLTC range and an 800V/5C fast-charge, the G7 can recharge from 10% to 80% in just 12 minutes. He quipped that “in the time it takes to charge one Model Y, you could charge two G7s”, since a Model Y (even the new version) needs roughly 24 minutes for 10–80%. Tesla’s official CLTC ranges in China span ~593 km (RWD) to 750 km (AWD) after recent battery upgrades, so XPeng is effectively matching the top end of Tesla’s range while beating it on charging speed.

• Autonomous Driving Tech: Here He Xiaopeng showed particular pride. The G7 Ultra’s 3 Turing chips give 2,250 TOPS of localized AI compute – “the highest in the world,” he said – whereas Tesla’s Full Self-Driving computer is ~720 TOPS and also costs ¥64,000 extra as a software add-on in China. (Notably, all G7 trims come with XPeng’s full advanced driver assistance suite standard, no subscription needed.) Xiaomi’s YU7 was cited at 700 TOPS. He emphasized that G7’s Ultra version is 3–28× more powerful in AI computing than other flagship EVs, which XPeng believes will translate to superior self-driving and smart cockpit performance.

• Pricing: The G7’s price undercuts the Tesla Model Y by a huge margin. A base Model Y RWD currently starts at ¥263,500 in China , about ¥67,700 more than G7’s base; even Xiaomi’s YU7 starts at ¥253,500 , over ¥57k higher than G7 . This pricing strategy drew applause as it positions the G7 as perhaps the best value in the segment for the feature set on offer.

He Xiaopeng also acknowledged Xiaomi’s entrance into the EV arena. Xiaomi’s CEO Lei Jun had launched the Xiaomi YU7 only a week prior (June 27) as another Model Y challenger. In a friendly exchange, He publicly congratulated Xiaomi’s launch success and even said he placed an order for a YU7 himself, while vowing to “work hard to bring surprises and new choices” to consumers. Lei Jun reciprocated by mentioning on a livestream that those who can’t wait for the YU7 might consider XPeng’s upcoming G7. Such banter underscored the buzz around this new class of tech-centric EVs. At the G7 event, XPeng revealed that interest was running high: within 9 minutes of the launch, G7 secured over 10,000 orders (with a ¥5,000 deposit each). For comparison, Xiaomi had boasted an almost unbelievable 200,000 orders in 3 minutes for the YU7’s debut (reportedly selling out a year’s production) . These eye-popping preorder figures – if accurate – signal tremendous demand in China for smart EVs around the ¥200k–300k price point.

Overall, the launch event painted the G7 as XPeng’s “cross-generational” flagship for AI driving, aimed squarely at dethroning the Tesla Model Y in China. XPeng executives stressed that the G7 has “no obvious short board” – meaning no major weaknesses – which they see as essential in today’s hyper-competitive market. They also teased what’s next: He Xiaopeng noted that later this year XPeng will roll out an “ultra hybrid power” platform (codenamed Kunpeng) and continue leveraging its full-stack R&D to keep innovating. He even remarked that achieving profitability by the fourth quarter of 2025 “is not a challenging goal” for XPeng, given the expected boost from G7 and other new models.

XPeng’s Current Position in China’s EV Market

The G7 launch comes at a pivotal moment for XPeng. After a difficult 2022–2023 period (when sales slowed and the company incurred heavy losses), XPeng has been mounting a comeback in 2025. The company delivered 34,611 vehicles in June 2025, a 224% year-on-year jump, marking its 8th consecutive month above 30k deliveries. Cumulative deliveries for the first half of 2025 reached 197,189 – remarkably, surpassing XPeng’s total sales for all of 2024. This 6-month volume also put XPeng ahead of its domestic “New EV” startup peers (NIO and Li Auto) in the period, reclaiming leadership among the Chinese EV upstarts. In fact, XPeng’s Q1 2025 deliveries (94,000 units) were up 330.8% year-on-year, a record high for the company.

Stronger sales have begun translating into improved financial metrics. In Q1 2025, XPeng’s revenue more than doubled (+141% YoY to ¥15.81 billion) and vehicle gross margin hit 10.5%, up 5 percentage points from a year prior. Net losses have narrowed substantially – the company lost ¥660 million in Q1, less than half the ¥1.37 billion loss in Q1 2024. XPeng management has indicated that with continued scale and cost optimizations, they anticipate turning profitable by late 2025. The company’s cash reserves are also strong (over ¥45 billion on hand as of Q1), helped by strategic investments (notably Volkswagen took a ~5% stake in 2023) and efficient capital raises.

Driving this turnaround is a refreshed product lineup and XPeng’s “ALL IN” strategy on automotive AI. 2025 has been a banner year for product launches. In the first half alone, XPeng introduced updated versions of its P7+ sports sedan, G9 SUV, G6 SUV, MONA G3i/M3 (compact), and the X9 MPV, alongside software upgrades like City NGP (navigation guided pilot) expanding to more cities. The new P7+ (a facelift of the P7) has been particularly successful – over 62,000 units have been delivered since its late-2024 release, and it has ranked #1 in sales for mid-to-large EV sedans (¥150k–¥200k segment) for 7 months straight. XPeng’s budget-friendly MONA model (an A-segment sedan around ¥150k) is also a hit, averaging over 10k units per month and leading its category. The company’s broader lineup now spans from entry-level smart EVs to premium models, which has helped it capture a wider customer base.

Strategically, XPeng has differentiated itself with a deep focus on in-house technology development. It is one of the few Chinese automakers designing its own autonomous driving chips (the “XPU” Turing chips) and full-stack ADAS algorithms. It has built a large R&D presence (over 40% of staff are in R&D) and invested heavily in AI research – for example, XPeng’s AI team has trained massive driving foundation models with up to 72 billion parameters and is employing reinforcement learning at scale to improve long-tail driving scenarios. XPeng calls itself the “AI Car” pioneer, emphasizing the concept of “AI empowerment for all” – essentially democratizing advanced driver-assist features into lower price segments. Indeed, in May 2025 XPeng rolled out its driver-assist (NGP) features on the ¥150k MONA model, a move hailed as an “AI tech equalization milestone” since such features were previously limited to high-end trims.

Financially, XPeng’s rebounding sales and tech-centric approach have regained investor confidence. The company set an ambitious annual sales target of ~400,000 units for 2025, and by mid-year it had already achieved over 50% of that . With the G7 now launching (expected to contribute significant volume in H2) and another major model – a next-gen P7 sedan – due in Q3, XPeng’s average selling price and margins are projected to rise further. Analysts from CMBI and others have noted that the G7 will be a key swing factor for XPeng’s Q3 performance. One area to watch is whether G7 sales cannibalize XPeng’s own P7+ sedan or G6 crossover, given some overlap in price; XPeng insists each model targets different use cases (P7+ for sporty sedan seekers, G6 as a compact SUV, and G7 as a family SUV). If G7 expands XPeng’s total market without eating into its siblings too much, it could further boost the company’s growth trajectory.

In summary, XPeng enters the G7 era on an upswing – it has reclaimed momentum among China’s EV contenders and is doubling down on its bet that superior AI technology will be a long-term differentiator. The company’s CEO has even stated that XPeng is not merely engaging in a price war or “spec chase,” but rather improving operational efficiency so it can “roll out price cuts from a position of strength” while still offering class-leading tech. Investors and industry observers are now watching to see if XPeng can translate its surging sales into sustained profitability by year-end, validating its strategy.

Market Impact and Competition: G7’s Ripple Effect in China’s EV Landscape

The launch of the G7 is set to reverberate across China’s electric vehicle market, especially in the highly contested mid-size electric SUV segment. This category – roughly the ¥200k–¥300k five-seat SUV class – is arguably the hottest battleground of 2025. It includes Tesla’s Model Y (the incumbent leader), a swarm of new Chinese offerings, and even entrant tech companies like Xiaomi. XPeng’s G7 is diving head-first into this fray with a potent combination of high-end tech and low pricing, which could raise the competitive bar for everyone.

Read more

Hot Sales and Social Media Buzz in China

Xiaomi’s First AI Glasses a Sell-Out: Xiaomi’s newly launched AI smart glasses have become an instant hit in China, shattering initial sales records and dominating online buzz. On launch night (June 26, 2025), eager “Mi Fans” rushed to Xiaomi stores to snap up limited stock, and within 12 hours on JD.com (Jingdong) the glasses sold over 10,000 units, soaring to the No.1 spot on the smart glasses bestseller list . The premium electrochromic-lens editions (which feature tint-adjustable lenses) were completely sold out in minutes . Five days after release, Xiaomi’s JD flagship store had exceeded 10k sales, with another 4,000+ units on Tmall (Taobao) – effectively clearing out initial inventories. At Xiaomi’s retail outlets, staff reported the glasses were nearly out of stock nationwide (standard black-frame units only available in very limited quantities, and tinted versions on back-order for weeks) . Such demand firmly cements Xiaomi’s device as the fastest-selling AI glasses ever in China , a category that until now was considered niche.

Beyond raw sales, social media hype around the product has been immense. Discussion of “Xiaomi AI Glasses” trended across platforms – from Weibo and WeChat feeds to video sites. On Bilibili, Xiaohongshu (Red), and Douyin (TikTok China), countless tech bloggers and everyday users posted hands-on videos and reviews, driving the topic’s popularity even higher . Even content creators who “never wear glasses” have been trying out Xiaomi’s eyewear and sharing their experiences . This viral buzz helped Xiaomi’s glasses “break out” of tech circles into mainstream awareness , signaling genuine consumer curiosity for this new device category.

User Feedback – Pros and Cons: Early adopters generally report that Xiaomi’s glasses deliver on their promises, though with some first-generation trade-offs. Audio quality and comfort are frequently praised – the built-in open-ear speakers produce clear sound and calls, and the 40g frame feels as light and comfortable as ordinary eyewear for most users . Many wearers find it easy to forget they’re using a high-tech gadget at all. The glasses’ 12MP front camera enables convenient first-person photos and videos – a key selling point. Users describe the camera as “good enough” for casual shooting (especially in well-lit conditions) and appreciate the novelty of capturing moments hands-free from a true eye-level perspective . In fact, those who have tried similar products (like Meta’s Ray-Ban Stories or TCL’s Thunderbird glasses) say Xiaomi’s image quality is the best of the trio , with bright, vibrant shots. Parents love being able to record snippets of daily life (e.g. playful children) with a quick tap, without fumbling for a phone . On Chinese e-commerce, the product earned thousands of reviews within days, reflecting strong engagement .

However, feedback also flags notable shortcomings. The glasses lack any augmented-reality display, so all interaction is via voice (Xiaomi’s “Xiao Ai” assistant) or a companion phone app – meaning no visual HUD for notifications or navigation. Some users expected a bit more “sci-fi” and note that Lei Jun (Xiaomi’s CEO) didn’t wear the glasses onstage at launch, presumably because they have no built-in display to double as a teleprompter . Image stabilization and low-light performance are weak points: quick head movements can lead to blurry “shaky” footage, and night videos suffer noticeable jitter . A few buyers humorously complained that pressing the shutter button causes the glasses to tilt (because you’re literally pushing on your eyewear), resulting in slanted shots . Battery life, while improved over earlier products, is still a limiting factor – heavy use drains the glasses in a few hours. Users report mixed results: for example ~10% battery drop for a 7-minute video recording, whereas just idling or listening to music for 30 minutes used only ~1% . Many agree all-day use isn’t realistic yet; you’ll likely recharge midday during intensive use . On the plus side, Xiaomi’s custom dual-chip design (with a Qualcomm Snapdragon AR1 processor and a secondary low-power chip) supports fast charging – about 30 minutes to go from 20% to 93% battery . Unlike Meta and others, Xiaomi notably omitted a charging carry-case, which some see as a drawback for on-the-go use . Finally, some iPhone and non-Xiaomi Android users note that full functionality currently requires a Xiaomi smartphone – the glasses pair to any phone via Bluetooth/WiFi, but certain features (e.g. live-streaming from the glasses’ camera, seamless video calls from glasses) are integrated only with Xiaomi’s own MIUI/Pinecone OS phones . Despite these wrinkles, the overwhelming market response shows that consumers are excited to embrace this new form factor, especially at Xiaomi’s aggressive pricing.

From Concept to Reality: Xiaomi’s Smart Glasses Journey

Xiaomi’s “overnight” success in smart glasses has in fact been years in the making. The company has explored wearable tech eyewear for nearly four years, steadily advancing from early concepts to mass-market product:

• September 2021 – Concept Prototype: Xiaomi debuted its first Smart Glasses “Exploration Edition” concept, signaling its ambitions in augmented reality. Weighing just 51g, the concept glasses used a tiny MicroLED display with optical waveguide technology to project a monochrome image onto the lens . Xiaomi demonstrated basic AR capabilities: hands-free navigation with turn-by-turn directions floating in your view, notification pop-ups, and even real-time translation of text in your line of sight . The concept included a 5MP camera and voice calling, showcasing a future where glasses could handle many smartphone tasks. At the time, it was purely a tech demo with no price or release date announced – but it proved Xiaomi’s R&D teams were serious about AR hardware.

• August 2022 – First Product Test (Mijia Glasses Camera): The company’s ecosystem partner launched the Mijia Glasses Camera, a camera-centric smart glasses device. This was Xiaomi’s first commercial foray into the category (under its Mi Ecosystem sub-brand “Mijia”). The Mijia Glasses Camera featured a dual-camera setup (50MP main + 8MP periscope telephoto) and a small heads-up display for augmented reality translation and shooting tips . Priced at ¥2,699 (crowdfunding offer ¥2,499), it wasn’t cheap – yet it saw an enthusiastic response from Xiaomi’s community. Within 5 minutes of its crowdfunding launch, it raised over ¥1 million in orders , and by the campaign’s end it had amassed ¥10.39 million from over 69,000 backers . This strong showing made it 2022’s top crowdfunded Xiaomi eco-product. The device was a niche experiment, but proved there was real consumer interest in wearable camera glasses for first-person shooting and translation. Xiaomi treated it as a learning experience in optics, battery and user experience for eyewear.

• 2023 – Smart Audio Glasses: As audio-equipped “glasses with speakers” grew trendy, Xiaomi introduced its Mijia Smart Audio Glasses line. The first-gen model (late 2022) and a second-gen in early 2025 (at a budget-friendly ¥999 price ) offered users open-ear Bluetooth audio in a stylish glasses form factor. These devices, similar to Bose Frames or Huawei’s Gentle Monster series, had no cameras or displays – focusing instead on music playback, phone calls, and voice assistant access. While not as high-tech as AR glasses, they further honed Xiaomi’s design of comfortable, lightweight frames and showed Xiaomi’s interest in the “audio glasses” sub-category (essentially wearable earphones). Notably, these were produced via an eco-chain partner (Fengmi Technologies) and branded “Mijia” rather than Xiaomi. This kept expectations modest, but it laid groundwork in components (micro speakers, beamforming mics, battery in temples, etc.) relevant to future smart glasses.

• Late 2024 – Stepping Up Investment: Rumors swirled that Xiaomi was gearing up for a flagship AR/AI glasses launch. Reports emerged that CEO Lei Jun had set an ambitious internal goal of 300,000 units for the first production run – a massive number given the entire Chinese AR glasses market was still nascent. Xiaomi was said to be partnering with Goertek (a leading AR/VR hardware OEM) on design and manufacturing . Goertek’s expertise in optics and micro-displays hinted that Xiaomi’s device could include advanced features like electrochromic lenses and dual processors. By the end of 2024, industry watchers were calling the expected launch a potential watershed for consumer AR in China . Xiaomi’s entrance – as a top smartphone maker with huge retail reach – was anticipated to “break the silence” of big players missing from this arena .

• June 2025 – Xiaomi AI Glasses Launch: Xiaomi officially unveiled its first AI smart glasses on June 26, 2025, pricing it aggressively at ¥1,999 (approx. $280) for the base version . Two premium versions with electrochromic lenses (tint-adjustable sunglasses, in mono or multi-color) were offered at ¥2,699 and ¥2,999 . The product is sold under the main Xiaomi brand, reflecting its strategic importance (unlike past “Mijia” experiments) . Lei Jun introduced the glasses as “a personal smart device for the next era” – essentially a wearable AI assistant . The device indeed integrates Xiaomi’s voice AI (“Xiao Ai”) for continuous dialogue and real-time Q&A about anything in your field of view . It can snap HD photos or record POV video hands-free (“Xiao Ai, start recording” triggers it ), play music, take calls, and even interface with smart home appliances via voice commands . Internally, it packs a dual-processor architecture (Qualcomm’s Snapdragon AR platform plus a secondary BES co-processor) to handle AI tasks and sensor data efficiently . Xiaomi made no secret that it benchmarked this product against Meta’s Ray-Ban Stories collab – emphasizing advantages like better fit for Asian faces, longer battery (up to 8 hours on moderate use), and integration with Xiaomi’s app ecosystem . In features and pricing, Xiaomi is effectively bringing high-end smart glasses capabilities to a much wider Chinese audience, validating a product category that domestic startups had pioneered but not popularized. The immediate sales success and mainstream media coverage indicate Xiaomi’s bet on “AI glasses” as the next big tech accessory may be paying off.

China’s Smart Glasses Arena: Xiaomi vs. the Competition

Xiaomi’s entry arrives amid what local tech media dub the “Hundred Glasses War” – a surge of players big and small racing into smart eyewear . After years of false starts, 2025 is viewed as a breakout year for the industry, with over 100 models of smart glasses (broadly defined) expected to launch in China alone . Companies ranging from phone giants to internet platforms and AR startups are vying to define the market. Here we look at the major Chinese players and how Xiaomi’s offering compares:

• Huawei – Betting on Audio-First Wearables: Huawei has approached smart glasses from a lifestyle angle, focusing on audio and health features. Its flagship HUAWEI Smart Glasses 2 (launched Sept 2023) is essentially an “AI audio glasses” device – glasses with built-in open-ear speakers and microphones, but no camera or AR display . Running HarmonyOS, it supports voice assistant functions, real-time translation and even posture monitoring for spinal health . The emphasis is on seamless integration with Huawei phones and all-day wear. Weighing around 38g with elegant frames co-designed by fashion brands, Huawei’s glasses target users who want a wearable wireless headphone that also looks like stylish eyewear. Priced from ¥1,699 (standard models) up to ¥2,299 for premium titanium frames , Huawei’s models are in a similar range to Xiaomi’s but serve a different use-case. Positioning: Huawei’s lack of camera or AR visuals means it doesn’t directly compete on the “AI vision” functionality of Xiaomi’s glasses. Instead, Huawei is competing to be the go-to audio smart glasses choice – a high-tech successor to Bluetooth earbuds. Huawei has hinted at future AR-capable glasses (and is reportedly investing in AR optics R&D ), but as of mid-2025 its offering remains firmly audio-centric. By volume, Huawei has already sold tens of thousands of these audio glasses in China, indicating a solid niche market. Xiaomi’s new glasses actually include all the features of Huawei’s (music, calls, voice assistant) plus the camera – albeit in a slightly bulkier form – which could sway some consumers away from Huawei’s if they value the camera feature.

• Thunderbird (TCL’s RayNeo) – AR Veteran Turning to Consumers: Thunderbird Innovations – often marketed as TCL RayNeo – is one of China’s pioneering AR device makers. Spun off from display giant TCL, RayNeo has years of experience developing AR display glasses (notably the Thunderbird X2 prototype with color microLED waveguides in 2021). In January 2025, RayNeo launched the Thunderbird X3 (V3) Smart Glasses at CES Las Vegas, marking its first major

Read more

Moore Threads (摩尔线程) – often dubbed “China’s Nvidia” – is a Beijing-based GPU startup rapidly emerging as a key player in China’s semiconductor ambitions. As the company prepares for an IPO on Shanghai’s STAR Market, it has drawn intense interest for its cutting-edge GPU products and strategic importance. Founded in 2020 by Zhang Jianzhong, a former Nvidia executive, Moore Threads has built a full-stack GPU ecosystem from graphics cards to AI accelerators . In this in-depth profile, we explore Moore Threads’ background, technologies, market positioning, funding journey, and its role in China’s drive for tech self-reliance.

Company Background and Founding

Moore Threads was officially established in October 2020, making it one of the youngest among China’s new wave of chip unicorns . Its founder and CEO, Zhang Jianzhong, previously spent 14 years as Nvidia’s Global Vice President and General Manager for China . Under Zhang’s leadership, the startup assembled a core team of veteran GPU engineers – including CTO Zhang Yubo, also an Nvidia alumnus – with the goal of developing a “full-function GPU” entirely in-house . This experienced team enabled Moore Threads to achieve extraordinary early milestones. Notably, the company designed and produced its first GPU chip in under 300 days, a speed hailed as setting a new “China speed” in chip design . By late 2021, Moore Threads had successfully developed a prototype of its first-generation GPU, demonstrating the team’s deep expertise in graphics and parallel computing architectures .

Despite its brief history, Moore Threads quickly gained credibility in China’s tech community. It earned recognition as a national “High-Tech Enterprise” in 2021 and was later designated a “Specialized and New” innovative firm in Beijing . Zhang Jianzhong initially operated behind the scenes – with the company’s early corporate registration under his wife’s name – but by 2023 he had taken full direct control, now holding roughly 36% of Moore Threads’ shares and serving as Chairman and General Manager . This restructuring set the stage for an IPO, solidifying Zhang’s leadership and aligning the company’s ownership for public listing.

The origin of the name “Moore Threads” reflects the company’s vision. It evokes Moore’s Law and multi-threaded processing, signaling a mission to push computing performance forward. As Zhang often emphasizes, the rise of AI and metaverse applications will demand massive computing power, and GPUs – with their parallel processing prowess – are the “computing base” for this future . Moore Threads’ stated vision is to become a globally competitive GPU leader providing advanced acceleration platforms for the AI-driven, digital world . In Zhang’s words, “Today’s large language models are only a starting point for AI. Future applications – from AI for science to immersive graphics – will require enormous compute. Aside from AI acceleration, Moore Threads can provide 3D rendering, video encoding, physics simulation, and scientific computing acceleration. With these diverse capabilities, our GPU clusters can serve a wide range of scenarios for the long term.” This ambitious outlook underpins the company’s founding philosophy.

Technical Capabilities and Products

From the outset, Moore Threads positioned itself as a full-spectrum GPU provider. The company developed its own GPU architecture called MUSA (Moore Threads Unified System Architecture) – a unified platform akin to Nvidia’s CUDA ecosystem . MUSA encompasses a common programming model, runtime libraries, driver framework, instruction set, and chip architecture, enabling software to run across Moore Threads’ various products . In essence, MUSA is the software and hardware backbone that ties together graphics rendering, general compute, AI, and multimedia functions on Moore Threads’ chips. This unified approach is crucial for developers, as applications written for MUSA can be easily ported between cloud and edge environments and across different workloads (graphics, AI, video, etc.) . It mirrors the strategy of Nvidia’s CUDA-unified platform, signaling Moore Threads’ intent to build an ecosystem rather than just standalone chips.

Hardware Architecture: Moore Threads specializes in “multi-functional” GPU chips, meaning each chip is designed to accelerate a diverse array of tasks. The company’s first-generation GPU (code-named “Sudi”**) debuted in 2022 and integrated four major engines: a modern 3D graphics engine, an intelligent multimedia engine, an AI compute engine, and a physics/scientific computation engine . By incorporating all these cores on one chip, Moore Threads became the first Chinese company to design a full-featured GPU covering graphics and AI, and it did so with remarkable speed to market . The emphasis on a general-purpose GPU contrasts with some domestic peers that focus solely on AI accelerators – Moore Threads aims to match the breadth of functionality offered by Nvidia and AMD GPUs.

Moore Threads’ MTT S80 gaming graphics card features the company’s second-generation “Chunxiao” GPU with 4,096 MUSA cores. Launched in late 2022, the MTT S80 delivers up to 14.4 TFLOPS (FP32) and was billed as the world’s first PCIe 5.0 x16 gaming GPU .

Product Lines: In just a few years, Moore Threads has rolled out multiple GPU product series addressing different market segments:

• Gaming and Consumer GPUs: The firm entered the consumer graphics market with cards like the MTT S60, S70, and S80. The MTT S60 (launched in 2022) was a desktop GPU aimed at PCs and workstations, built on the first-gen “Su**di” chip. It packed 2,048 cores on a 12nm process, achieving about 6 TFLOPS FP32 performance . Notably, the S60 supported all major graphics APIs (DirectX, Vulkan, OpenGL) and even Windows 10 OS, a first for a Chinese GPU . Demonstrating decent gaming capability, it could run popular titles like League of Legends at 1080p high settings smoothly on Windows, and titles like Counter-Strike: Global Offensive and Dota 2 on Linux . This showed early on that Moore Threads was serious about compatibility and user experience.

  The flagship MTT S80 card, released in late 2022, upgraded to Moore Threads’ second-gen GPU chip (code-named “Chunxiao”) with 4,096 cores at 1.8GHz. The S80 features 16GB of GDDR6 memory and a PCIe 5.0 interface, offering 448 GB/s bandwidth . With up to 14.4 TFLOPS of single-precision compute, the S80’s raw specs resemble a mid-high-range gaming GPU by global standards . In fact, it was introduced as the world’s first PCIe Gen5 graphics card, beating foreign competitors to that interface technology . Initially, the S80’s real-world gaming performance lagged behind contemporary Nvidia cards due to immature drivers – early tests showed it performing closer to a GeForce GTX 1650 in some games . However, over the past year the company has doubled the S80’s gaming performance through aggressive driver optimizations, achieving playable frame rates in titles like Genshin Impact and League of Legends by mid-2024 . The improved software support has significantly expanded game compatibility to virtually all top games in China . This progress highlights the learning curve for a new GPU entrant, and Moore Threads’ commitment to refining its software stack.

• Data Center and AI GPUs: In parallel, Moore Threads has developed server and accelerator cards for AI, cloud, and high-performance computing. Its first data center GPU was the MTT S2000, launched in 2022 alongside the S60. Based on the same first-gen architecture, the S2000 is a passively cooled, single-slot card designed for dense server deployments . It packs 4,096 MUSA cores (double the S60) and came with up to 32 GB GDDR6 memory, delivering around 10–12 TFLOPS FP32 performance . The S2000 was promoted as a “full-stack” data center GPU supporting 3D rendering, cloud gaming, video encoding/decoding (with multi-channel H.264/H.265/AV1 codec support), AI inference, and scientific computing – all in one package . It also introduced hardware GPU virtualization (MT Mesh 1.0) using SR-IOV technology, allowing one S2000 to be partitioned for multiple virtual machines – a crucial feature for cloud desktop and cloud gaming use cases . By supporting x86 and ARM server CPUs and mainstream Linux OS distributions, the S2000 could integrate into a variety of data center environments . Moore Threads partnered with server OEMs like Inspur, Lenovo, H3C, and Tongfang to validate the S2000 in their systems , signaling confidence that its GPUs could slot into existing enterprise hardware ecosystems.

  Building on that, Moore Threads released the MTT S3000 and MTT S4000 in 2023 as its next-generation accelerators focused on AI training and large-scale cloud clusters. The MTT S3000 (second-gen “Chunxiao” chip) offered around 15.2 TFLOPS FP32 and 32 GB memory, while the latest MTT S4000 (third-gen architecture, code-named “Quyuan”) pushes up to 25 TFLOPS FP32 with 48 GB GDDR6 memory . For context, Nvidia’s A100 data center GPU (2020 generation) provides ~19.5 TFLOPS FP32, and the current flagship H100 delivers ~67 TFLOPS . This means Moore Threads’ S4000 is roughly on par with Nvidia’s 2020-era high-end in raw single-precision compute, though only about one-third the performance of Nvidia’s latest top chip – highlighting a gap that Chinese GPUs still need to close at the ultra-high-end . The S4000 also introduced Moore Threads’ MTLink interconnect (240 GB/s bandwidth) for multi-GPU clustering, which, while improving multi-card scaling, remains far slower than Nvidia’s NVLink (900 GB/s) used in H100 systems . Despite these differences, the S4000 is a significant step up for China’s GPU capabilities. It is tailored for large AI model training and inference, supporting 8K HDR video output and multi-card distributed computing for models with tens of billions of parameters .

  Crucially, Moore Threads invested heavily in its software stack and developer tools to support these GPUs. The company provides drivers compatible with standard AI frameworks (TensorFlow, PyTorch, PaddlePaddle, etc.) and libraries for graphics and multimedia . In a notable move, Moore Threads created a tool called MUSIFY, which allows developers to port CUDA code to MUSA with “zero cost” . In other words, existing CUDA-based applications can run on Moore Threads GPUs with minimal changes, leveraging the vast NVIDIA CUDA software ecosystem on a new platform. This strategy of CUDA interoperability is seen as vital for adoption, given Nvidia’s dominance in AI software. “Using our MUSIFY tool, the MTT S4000 can fully leverage the existing CUDA software ecosystem, achieving zero-cost migration of CUDA code to the MUSA platform,” the company announced . Such capabilities lower the barrier for cloud providers and researchers to experiment with Moore Threads hardware without rewriting all their code. Additionally, Moore Threads’ GPUs support all major operating environments – Windows and several Chinese Linux distributions (Kylin, UOS), x86 and ARM CPUs – ensuring they can slot into government and enterprise IT settings where domestic software stacks are used . This broad compatibility and “full feature” design ethos have been Moore Threads’ technical calling card.

Market Position and Industry Impact

Moore Threads has quickly risen to prominence in China’s nascent GPU industry. Among the handful of Chinese GPU startups founded around 2017–2020, Moore Threads is the youngest but has achieved the highest valuation and arguably the most comprehensive product lineup . Its approach closely mirrors Nvidia’s business model, earning it the nickname “domestic Nvidia” in Chinese media . Like Nvidia, Moore Threads offers both consumer GPUs (for gaming/graphics) and enterprise GPUs (for AI and data center), and it has established separate divisions for each. In 2023, the company created two strategic business units: AISG for AI chips and MCSG for consumer graphics, paralleling Nvidia’s split between its data center AI business and GeForce gaming division . This dual focus differentiates Moore Threads from peers such as Biren Technology or Iluvatar (燧原科技), which have concentrated mostly on AI training chips. By covering the full spectrum, Moore Threads aims to be a one-stop GPU provider for China’s needs – from powering video games and metaverse applications to training large neural networks and running cloud services.

In China’s GPU ecosystem, Moore Threads is seen as a front-runner alongside a few other well-funded startups. It is one of the country’s first firms to produce a “full-function” GPU (capable of graphics and compute), whereas earlier domestic efforts in GPUs were either limited to mobile graphics (like smartphone GPUs) or AI accelerators without strong graphics support. This unique positioning has made Moore Threads a showcase project for China’s broader tech industry. According to the Hurun Global Unicorn Index 2024, Moore Threads was valued around RMB ¥255 billion (~$35 billion) – far above domestic GPU peers like Iluvatar CoreX (¥160B) or Biren (¥155B) . Industry watchers have noted that both Moore Threads and Biren were co-founded by former Nvidia China executives, and are considered the leading Chinese candidates to eventually rival Nvidia’s products . This has not escaped international attention: in October 2023, the U.S. government added Moore Threads to its export Entity List, explicitly identifying it (along with Biren) as a threat to U.S. dominance in AI chips . The listing is intended to hobble Moore Threads’ progress by cutting off access to advanced chip manufacturing using U.S. technologies . Paradoxically, the U.S. sanctions underscore Moore Threads’ importance – it has been deemed “China’s Nvidia challenger” on the global stage .

Domestically, Moore Threads has aligned its products with China’s push for indigenous technology in critical infrastructure. The company’s GPUs have been adopted or tested by major state-owned enterprises, particularly in sectors with “国产化” (localization) demand where dependence on foreign tech is a concern . For example, telecom giants China Mobile, China Telecom, and China Unicom have partnered with Moore Threads on cloud computing and data center projects. In late 2022, Moore Threads joined with operators (China Mobile Beijing, China Telecom Beijing) and OEM Lenovo to form the “PES-KUAE Intelligent Computing Alliance”, aiming to deploy GPU-accelerated computing centers with domestic hardware . The company unveiled a computing cluster solution called KUAE, and by July 2023 it expanded this from a 1,000-GPU cluster to a 10,000-GPU cluster design, signaling an ambition to reach “万卡” (10k-card) supercomputer scale in the near future . Banks have also emerged as key partners: in October 2023, SPD Bank, Postal Savings Bank of China, and Beijing Bank signed an agreement with Moore Threads to build a 10,000-GPU general-purpose computing cluster for fintech research, representing the first such large-scale GPU cluster by Chinese banks . Moore Threads has similarly signed cooperation deals with China Construction Bank’s Beijing branch and worked with state-owned energy and telecom firms on regional “zero-carbon” data center projects (for example, in Qinghai province and Guangxi) that plan to utilize Moore Threads GPUs .

This focus on state-backed clients and strategic industries shows Moore Threads carving out a protected niche: providing domestic GPUs for government, finance, and telecom applications where secure, locally-sourced technology is prioritized. It contrasts with Nvidia’s customer base, which is largely global cloud providers and AI firms. (In one recent quarter, Nvidia reported $26.3 billion in data center revenue, largely from U.S. and international tech companies .) Moore Threads, on the other hand, has found its early traction with Chinese enterprises that have mandates to reduce reliance on foreign chips . For instance, Moore Threads revealed that its GPUs and MUSA software stack have been optimized for China Mobile’s internal cloud platform, and its hardware is being used in indigenous AI computing centers built by local governments and companies . In the public sector, Moore Threads teamed up with GIS software firm SuperMap to create a GPU-accelerated 3D GIS solution entirely on domestic technology, lowering dependence on Nvidia or AMD GPUs . Another example is cybersecurity firm Qi-Anxin’s 360 Smart Brain AI appliance, which uses Moore Threads GPUs as the underlying compute engine . By embedding itself in such projects, Moore Threads is gradually seeding an ecosystem of applications tuned for its hardware.

To support these partners, Moore Threads has been proactive in building out its software and developer ecosystem. It claims to have hundreds of ecosystem collaborators – from CPU vendors (like Loongson, Phytium, Zhaoxin) to middleware and application developers . During product launch events, the company has showcased domestic software running on its GPUs, including WPS Office (Kingsoft’s office suite), the Little Fish video conferencing platform, Taiji Graphics and ZWCAD design software, Glodon BIM, Supremind GIS, and even game engines and sports games . By demonstrating support for popular Chinese software and workflows, Moore Threads is signaling to enterprise and government buyers that a viable “all-Chinese” computing stack is emerging. This broad approach – spanning hardware, system software, and industry solutions – is why analysts often compare Moore Threads directly to Nvidia, which similarly cultivates an ecosystem around its GPUs.

At the same time, Moore Threads faces challenges in truly rivaling the international giants. Observers note that Nvidia and AMD benefit from decades of IP accumulation and a vast patent portfolio in GPU technology – a formidable barrier for newcomers . “Foreign GPU vendors have accumulated a massive number of patents over decades; this is the reality domestic GPU products must face,” a veteran Chinese semiconductor engineer told Yicai Global . Additionally, while Moore Threads’ presence in protected domestic markets is growing, it has yet to prove itself in open competition against Nvidia/AMD for commercial cloud business. As of 2024, there have been relatively few announcements of China’s internet giants (like Alibaba, Tencent Cloud, or Baidu) deploying Moore Threads GPUs at scale; those companies still rely heavily on Nvidia (though U.S. export restrictions may force a change). Moore Threads appears aware that true market validation will require expanding beyond policy-driven sales. The company has recently cut prices on its gaming GPUs – for example, its mid-range MTT S70 card launched in mid-2023 saw a steep price drop from ¥2499 to ¥899 RMB within months – in an effort to attract more mainstream consumers and developers. It is also continually improving performance and software support, as seen in the S80’s driver progress, to make its products more competitive on pure merit.

Nonetheless, within China’s “GPU rush”, Moore Threads stands out. It has been frequently identified as a likely candidate for “China’s first GPU stock” once it IPOs . Other domestic GPU startups, like Biren and Celestial Semiconductor (沐曦科技), are also pursuing listings, but Moore Threads’ blend of strong financing, a star founder, and its full-stack strategy give it a leading profile. The coming years will test whether the company can translate its unicorn valuation into real market share in China’s booming AI computing market – projected to grow over 27% annually, with China’s GPU server market reaching $6.4 billion by 2024 .

Funding History and IPO Plans

Moore Threads’ rapid rise has been fueled by several hefty funding rounds, backed by an all-star roster of investors from both China’s state funds and big tech venture arms. In the span of three years, the company completed five rounds of financing totaling over ¥6.5 billion RMB (approximately $1 billion) . This war chest has supported its aggressive R&D and talent recruitment despite operating losses. Below is a brief timeline of Moore Threads’ funding journey:

• Angel & Pre-A Rounds (2020–Q1 2021): Shortly after its October 2020 founding, Moore Threads secured angel investment (amount undisclosed). By February 2021 (just ~100 days in operation), it raised a Pre-A round reportedly worth “several billion yuan” – an unusually large sum at that stage, reflecting investors’ high hopes. Notably, ByteDance (owner of TikTok/Douyin) participated in the Pre-A round , signaling interest from China’s tech giants in a domestic GPU alternative. Pandaily described the pre-A as a “mind-boggling” round for such a young startup . This early influx of capital allowed Moore Threads to sprint through chip development in 2021.

• Series A (November 2021): Moore Threads announced a ¥2 billion RMB Series A on November 25, 2021 . The round was jointly led by Shanghai Guosheng Capital (a state-backed fund), 5Y Capital (a major VC firm, formerly Morningside), and the BOC International Bohai fund, with participation from CCB International, Qianhai Fund of Funds, Sequoia Capital China, GGV Capital, and others . This mix of investors included both government-guided funds and market-oriented VCs, reflecting a strategy to blend policy support with commercial savvy. The A round proceeds were earmarked to accelerate mass-production of the first GPU chip, development of GPU SoC IP, and building the software ecosystem . Around this time, Moore Threads revealed that its inaugural GPU chip had successfully taped out, validating the investors’ bet.

• Series B (December 2022): In late 2022, Moore Threads closed a Series B of ¥1.5 billion RMB, which brought its company valuation to about ¥240 billion RMB (over $34 billion) . This round was led by the China Mobile Digital New Economy Fund (investment arm of the state telecom operator), along with existing stakeholders. The infusion came amid a global chip boom and China’s intensifying interest in AI infrastructure. By this point, Moore Threads had completed five funding rounds (Angel through B+), with cumulative fundraising exceeding ¥6.5 billion . Other known investors across these rounds include Shenzhen Capital Group (深创投) – a major government venture fund, Tencent (via its investment arm Tencent Venture Capital) , Sequoia China, GGV Capital, Lenovo Capital, and CAS-linked funds, among many others . The cap table is strikingly diverse: Moore Threads attracted tech giants, local government funds, insurance capital, and traditional VCs alike, all eager for a slice of the “China GPU” story.

• Series B+ (Late 2023): Despite geopolitical headwinds, the company managed to raise an additional ¥2+ billion RMB in a B+ round by November 2023 . This round, reportedly completed after Moore Threads was added to the U.S. Entity List, underscored Chinese investors’ confidence in its prospects. It also marked the fifth fundraising round since inception. Right before this, on October 17, 2023, the U.S. sanctions hit – Moore Threads was blacklisted for its role in advanced computing chips. In an internal all-hands letter, CEO Zhang Jianzhong addressed the sanction, announcing some “routine position optimizations” (interpreted as minor layoffs or restructuring) but emphatically saying China’s GPU industry is not in its “darkest hour” . “Our mission remains unchanged – we are dedicated to building China’s highest-quality full-function GPU, and we are determined to carry this undertaking to its ultimate goal,” Zhang wrote to reassure employees . The successful B+ raise soon after indicated that capital continued to back Moore Threads as a long-term play in spite of U.S. pressure.

By April 2024, Moore Threads’ valuation was listed at ¥255 billion on the Hurun Global Unicorn Index . To put its valuation in perspective, at ~$35 billion, Moore Threads is valued higher than many established semiconductor firms, yet its revenues are still nascent. According to the IPO prospectus, Moore Threads generated RMB ¥46 million in 2022, ¥124 million in 2023, and ¥438 million in the fiscal year ending April 2024, while incurring net losses of roughly ¥1.8 billion, ¥1.67 billion, and ¥1.49 billion in those years respectively . The exploding revenue (nearly 10× growth from 2022 to 2024) shows traction, but the company remains in heavy investment mode, spending aggressively on R&D and ecosystem development. These figures also highlight the high cost of designing competitive GPUs – requiring sustained funding until volumes scale up.

IPO Plans: Moore Threads is now charging toward an IPO on China’s STAR Market (Shanghai’s Sci-Tech Innovation Board). It officially initiated IPO counseling in November 2024 by filing with the Beijing Securities Regulatory Bureau, a required step to go public in China . By mid-2025, the company had submitted its IPO prospectus to regulators . The prospectus indicates Moore Threads aims to raise approximately ¥8 billion RMB (over $1.1 billion) in the offering . The fresh capital is earmarked for ambitious new chip R&D projects: roughly ¥25 billion for a next-generation AI training/inference GPU, another ¥25 billion for a next-gen graphics GPU, about ¥19.8 billion for a new AI SoC chip project, and the remainder (~¥10 billion) for working capital . These projects, described as “new generation autonomous and controllable” chips, align with China’s goal of self-reliance – implying the new GPUs will be domestically designed and hopefully manufactured without critical U.S. technology. The pre-IPO valuation was around ¥246.2 billion RMB (about $34B) , as per the last private financing round, and that was used to price the shares. If successful, Moore Threads’ IPO could become the first pure-play GPU listing in China, marking a milestone for the country’s semiconductor sector.

The listing process is expected to be closely watched. The company’s sponsor is CITIC Securities, and it reportedly completed the IPO tutoring and guidance phase by early 2025 . Barring regulatory delays, Moore Threads could debut on the STAR Market later in 2025. Given the hype around AI chips, some analysts anticipate a warm reception from domestic investors, though much will depend on market conditions and the company’s continued growth. It’s worth noting that other Chinese AI chip startups are lining up to IPO as well – for instance, Biren Technology filed for counseling in September 2024, and another GPU startup, Celestial Semiconductor (Muxi), did so in early 2025 . This suggests a race to the public markets as firms seek capital to fund next-gen chip development (which is ever more capital-intensive under tech export controls). Moore Threads, with the largest valuation, could set the benchmark.

Strategic Role in China’s Semiconductor Push

Moore Threads’ story cannot be separated from the wider context of China’s tech self-reliance campaign. GPUs are a linchpin technology – essential for AI, supercomputing, gaming, and military applications – and until recently, China had virtually no domestic alternatives to Nvidia or AMD. The rise of Moore Threads has thus been both a product of, and a contributor to, Beijing’s concerted efforts to foster homegrown semiconductor champions.

Read more

Launch Event Highlights

Xiaomi’s highly anticipated electric SUV YU7 made its debut on June 26, 2025, during a late-night “Xiaomi People–Car–Home Ecosystem” launch event in Beijing. As the second model from Xiaomi’s automotive arm (following the SU7 sedan), the YU7 was the headline reveal and positioned squarely as a direct challenger to Tesla’s Model Y . Xiaomi’s CEO Lei Jun took only “five minutes” the night before launch to decide on undercutting the Model Y’s price by ¥10,000 (∼$1.4k) , unveiling the YU7 with a starting price of ¥253,500 for the base model – deliberately just below Tesla’s entry Model Y in China . The new SUV comes in three trims (Standard, Pro, and Max) priced at ¥253.5k, ¥279.9k, and ¥329.9k respectively , each corresponding to Tesla’s RWD, Long Range AWD, and Performance variants.

The hype and initial demand were extraordinary. Within minutes of opening reservations (with a ¥5k deposit scheme), Xiaomi’s Weibo account announced over 200,000 orders in the first 3 minutes – roughly half of Tesla Model Y’s entire Chinese sales last year, secured in practically no time. This “3-minute 200k” feat sent Xiaomi’s stock surging and underscored the intense interest from buyers. Lei Jun quipped during the presentation that the SUV market is the “main battlefield” of the auto industry and that with YU7’s launch, Xiaomi is stepping into that arena against its “strongest competitor” yet . Clearly, Xiaomi staged the YU7 introduction not just as a product launch, but as a bold contest of strength with Tesla.

Xiaomi YU7: Key Specs and Features

Xiaomi pitches the YU7 as a “luxury high-performance SUV,” loaded with top-tier hardware and smart tech. It’s a mid-to-large coupé-style electric SUV, measuring about 4.999 m long, 1.996 m wide, 1.600 m tall, with a 3.0 m wheelbase – notably larger in footprint than the Model Y. The design is sleek and sporty: a long hood (classic 1:3 proportion) and low-slung stance give it a “wide-body and low-profile” look . The styling has drawn comparisons to high-end sport SUVs (Chinese media nicknamed it “Farami,” suggesting a Ferrari Purosangue resemblance ). Xiaomi offers 9 vibrant colors (from Pearl White and Deep Sea Blue to Lava Orange and Emerald Green) to appeal to style-conscious buyers .

Under the skin, YU7 is built for performance and range. All versions share an advanced 800 V silicon-carbide (SiC) high-voltage platform , enabling ultra-fast charging and high power output. The Standard and Pro models use a 96.3 kWh battery (LFP chemistry), while the Max upgrades to a 101.7 kWh ternary (NMC) pack . Range is a major bragging point: the base RWD Standard YU7 can run 835 km on a charge (CLTC rating) – dramatically higher than a Model Y RWD’s ~593 km in China . Even the dual-motor YU7 Pro manages 770 km CLTC , and the YU7 Max (Performance version) about 760 km – all beating their Tesla counterparts by dozens or even 200+ km of range in Chinese test conditions . This is partly due to sheer battery size (the YU7 Standard carries 33.8 kWh more than a Model Y RWD, an advantage roughly worth ¥20k in battery cost alone) , but also thanks to efficient design – the YU7’s drag coefficient is as low as 0.245 Cd , very impressive for a SUV.

Performance is equally headline-worthy. The Max trim is a dual-motor AWD “beast” with a combined output of 508 kW (691 hp) and 866 Nm torque . It launches 0–100 km/h in just 3.23 seconds with a top speed of 253 km/h – squarely in sportscar territory and a hair quicker than a Model Y Performance (≈3.5–3.7 s). The YU7 Pro (dual-motor long-range version) does 0–100 in 4.27 s , also outpacing Tesla’s Long Range Model Y (~5 s 0–100). Even the Standard RWD YU7, with a single 235 kW rear motor, can sprint 0–100 in about 5.9 s – not far off Tesla’s base model. All YU7 variants feature Xiaomi’s V6s Plus high-speed motor technology capable of 22,000 rpm, and even the base model has a healthy 528 Nm of torque for strong acceleration .

Xiaomi has “maxed out” the configuration on the YU7 . Key hardware across all trims includes adaptive air suspension (dual-chamber air springs on AWD models) with continuous damping control, a variable-ratio steering system, and large 4-piston Brembo brakes for confident handling . Inside, the cabin is packed with tech: a Snapdragon 8 Gen3 automotive-grade SoC runs the infotainment on Xiaomi’s own “Pengpai OS” (Surge OS), providing a snappy UI and rich connectivity . Notably, the YU7 introduces the “Xiaomi Tianji Screen” panoramic HUD – a wide heads-up display projected across the base of the windshield . This acts as a secondary display spanning the driver’s view, showing key info in a more intuitive way (and possibly replacing a traditional instrument cluster). In addition, rear passengers get a detachable 6.7-inch control tablet in the back to adjust climate, seats, etc. . Luxury touches abound, especially in the Max trim: options (some as limited-time free perks) include Nappa leather upholstery, “zero-gravity” reclining seats with 10-point massage for driver and shotgun (both power adjustable to near-flat position) , a 25-speaker premium audio system, soft-close doors, a smart dimming panoramic sunroof, HEPA filtration, and even a built-in mini fridge for rear passengers . Xiaomi clearly aimed to deliver a high-end experience that can match or exceed what Tesla offers in comfort and features.

On the battery and charging front, YU7 leverages its 800 V architecture for exceptional charging speeds. It supports up to 5.2 C fast charging, meaning a high-power DC charger can add on the order of 600+ km of range in 15 minutes under ideal conditions . For example, the YU7 Pro can replenish ~425 km in 15 minutes on a fast charger , and the Max (with a 5.2C rate) can go 10% to 80% in just 12 minutes, or about 620 km in 15 min . This outpaces Tesla’s peak charging rates; the Model Y’s 400 V system is limited to around 250 kW (which adds ~270 km in 15 min under Tesla’s Supercharger V3). One caveat: unlike Tesla (or Nio and others), Xiaomi lacks a dedicated charging network of its own, so YU7 owners will rely on public charging infrastructure or third-party stations . Some Chinese media note this could be a convenience shortfall versus rivals with proprietary charger networks , though Xiaomi is investing in partnerships and emphasizing that its universal 800 V support can take advantage of the fastest chargers available.

In terms of advanced driver assistance, the YU7 comes standard with a full sensor suite and high-power computing on every trim. It’s equipped with one forward-facing LiDAR (with 200 m range) plus a 4D millimeter-wave radar, 11 HD cameras, and 12 ultrasonic sensors, feeding into an NVIDIA DRIVE Thor autopilot chip delivering 700 TOPS of AI compute . Lei Jun announced that all YU7 vehicles will include Xiaomi’s “HAD” end-to-end assisted driving system (Xiaomi’s in-house answer to Tesla Autopilot), enabled by that hardware . This is an upgrade from Xiaomi’s first model (SU7), where not all trims had the full AD kit. Xiaomi has poured resources into this domain – Lei Jun revealed an initial ¥5.79 billion investment and a dedicated team of 1,800+ engineers for autonomous driving R&D . The YU7 will soon get an XLA large-model AI upgrade later in 2025 to enhance its driving intelligence . That said, reviewers who test-drove it note that while its ADAS is much improved over the first-gen SU7 and gives a solid sense of safety, it’s still not as mature or polished as Tesla’s Full Self-Driving (FSD) or the systems from Li Auto, Xpeng, or Huawei’s Harmony cars . Lei Jun himself admitted on stage that Tesla leads in areas like energy efficiency and autonomous driving software – “Model Y’s energy management and FSD are indeed ahead of us, we still have a lot to learn” he conceded . Nonetheless, Xiaomi is striving to catch up quickly on software while trying to “overtake on hardware” by including sensors like LiDAR that Tesla famously omits.

YU7 vs Tesla Model Y: How Do They Compare?

The Xiaomi YU7 was explicitly crafted to take on the Tesla Model Y, and the comparisons span everything from design to specs. Here’s a breakdown of the key face-offs between YU7 and the Model Y in the Chinese market:

• Design & Appearance: The YU7 is a larger vehicle with a more aggressive sporty aesthetic. It has a long, elegant hood and a sloping fastback roofline, giving it a coupe-SUV silhouette that many find very eye-catching . In fact, commentators note it looks “better than good”, with sleek lines exuding both athleticism and style – one even dubbed it “法拉米” (“FaraMi”, a play suggesting a Ferrari-like Xiaomi) . By contrast, Tesla’s Model Y has a more understated, rounded crossover shape that prioritizes function over flamboyance. The YU7’s exterior also incorporates active aerodynamic elements (e.g. a 100-position active grille shutter, front air ducts, dual rear spoilers) to reduce drag , whereas the Model Y relies mainly on its smooth egg-like form for aero. Both cars have flush door handles, but Xiaomi uses a novel power pop-out hidden handle design for a cleaner look . YU7 offers far more color choices (nine factory colors vs. Tesla’s five) to cater to personalized tastes . In essence, the YU7 aims to appear more luxurious and “fun” than the relatively minimalist Model Y. (It’s worth noting the YU7 is Xiaomi’s first SUV design – whereas the earlier SU7 sedan took inspiration from Porsche, the YU7’s styling echoes Italian super-SUV cues. Whether one prefers the flashier Xiaomi or the familiar Tesla shape is subjective, but Xiaomi clearly wanted to make a visual statement.)

• Size & Space: Physically, the YU7 is bigger in every dimension compared to the Model Y (except a slightly lower height) . It stretches about 202 mm longer, 76 mm wider, and has a 110 mm longer wheelbase than Tesla’s SUV . This extra length and width translate to a roomier interior. Xiaomi even bragged that using a 1.88 m dummy, they measured front and rear headroom, rear knee room, and rear shoulder room all surpassing the Model Y (and even edging out the larger Porsche Cayenne in some metrics) . Reviewers who sat in YU7 agree it has very comfortable seating, with well-bolstered seats and ample legroom . The rear seats can recline to 135° for lounge-style comfort, and the cargo area is generous (678 L, expanding to 1758 L) despite the sloping tail . The Model Y, while quite roomy for its size, offers a slightly more utilitarian cabin with upright seating and less lavish rear accommodations (no reclining rear or executive perks). Both cars seat five, but YU7 targets a more premium seating experience – for instance, YU7’s optional zero-gravity massage seats and extra rear screen have no Tesla equivalent. In summary, space and comfort tilt in YU7’s favor, courtesy of its larger body and luxury focus.

• Battery Range & Efficiency: In pure range figures, the YU7 decisively beats the Model Y’s Chinese specs. The YU7 Standard’s 835 km CLTC rating dwarfs the 593 km of a China-market Model Y RWD . Even comparing AWD versions: YU7 Pro gets 770 km vs Model Y Long Range’s ~720 km, and YU7 Max ~760 km vs Model Y Performance 615 km . However, it’s important to note Xiaomi achieved this by using much larger batteries – e.g., 96.3 kWh vs Tesla’s ~78 kWh in AWD, and over 100 kWh in YU7 Max vs ~80 kWh in Model Y Performance . Tesla still holds the edge in energy efficiency per kWh; its powertrain and weight optimizations mean it can go far on a smaller battery. Lei Jun acknowledged Tesla’s lead in energy management – Model Y’s lower consumption is a benchmark Xiaomi chased . Indeed, Xiaomi claims the YU7’s best energy consumption is 13.3 kWh/100 km in ideal conditions , which is excellent and comparable to Model Y’s ~13–14 kWh/100 km (given the size). So while YU7 provides more total range, it comes at the cost of a heavier battery. For consumers, though, the headline is clear: YU7 alleviates range anxiety by offering hundreds of extra kilometers and faster charging to top it off. Tesla’s counter might be its extensive Supercharger network and slightly better real-world efficiency – but on paper, Xiaomi wins the range duel.

• Acceleration & Performance: The YU7 Max is a genuine performance SUV. At 3.23 s 0–100 km/h, it actually out-accelerates a Model Y Performance (which does 0–100 in ~3.7 s in China) . The Xiaomi’s top speed (253 km/h) is also higher than Model Y’s governed ~217 km/h . With nearly 700 horsepower on tap and high-performance tires (up to 275 mm wide in back) , the YU7 Max is positioned as a “million-yuan level” performance car in a ¥330k package . It even has racetrack cred – Xiaomi touts that the YU7 prototype ran a continuous 24-hour endurance test (in a Le Mans style challenge) and that the sibling SU7 sedan set an EV lap record at Nürburgring . Tesla’s Model Y Performance is no slouch, known for its punchy dual motors and agile handling, but Xiaomi is clearly pushing for an image of greater horsepower and driving fun (giving buyers all the power they can handle) . Moreover, the YU7’s chassis tuning is co-developed with performance in mind: double-wishbone front suspension, multi-link rear, adaptive damping and air springs on AWD models . Early testers report it rides comfortably while still feeling sporty and planted – Xiaomi aimed to balance “sporty and comfortable” in the suspension setup . The Model Y is relatively firm and sporty too (especially the Performance edition), but doesn’t offer adjustable suspension or high-end dampers at this price. Braking is another win for YU7: its Max trim gets performance brakes and all models use large 4-pot calipers , whereas Model Y has more basic brakes (some owners even upgrade Tesla brakes aftermarket for heavy track use). Overall, in raw performance and driving dynamics, the YU7 appears to one-up the Model Y on paper – though we’ll see if that translates to a noticeably superior feel on the road.

• Technology & Features: Xiaomi has stuffed the YU7 with tech features that the Model Y either offers only as paid upgrades or not at all. For instance, every YU7 has a LiDAR sensor and surround sensors for driving assist, where Tesla famously relies only on cameras and doesn’t have LiDAR. The YU7’s in-cabin technology also arguably outshines Tesla’s minimalist approach: besides the expansive HUD (Tianji Screen) , Xiaomi’s SUV has dual wireless phone chargers, an AR heads-up display, a second-row tablet controller, and is deeply integrated with Xiaomi’s consumer electronics ecosystem. The idea of “人车家” (People–Car–Home) integration means your YU7 can interface with your smart home and smartphone seamlessly . For example, you could use the Xiaomi Home app or Mi AI assistant to control home devices from the car, or use the car to display phone content and respond to your Xiaomi smartphone’s commands. Tesla, on the other hand, has its strengths in software updates, a robust central OS, and a rich app ecosystem (games, Spotify, etc.), but it doesn’t integrate with home IoT in the same way. Both cars support OTA (over-the-air updates) for continuous feature improvements. Another interesting feature: Xiaomi developed a special “motion sickness alleviation mode” for the YU7’s drive control, using adaptive suspension and power modulation to smooth out accelerations and turns – a very user-centric touch for family comfort. This kind of feature shows Xiaomi leaning on its consumer electronics DNA to think of the user experience details. Meanwhile, Tesla’s focus has been on self-driving software (FSD beta) and infotainment, where it still might have an edge in polish and capabilities (like the Model Y’s sophisticated AI driving features and larger ecosystem of third-party apps). In summary, Xiaomi YU7 leads in hardware and creature comforts, packing in features that make the car feel futuristic and premium, whereas Tesla leads in software maturity and simplicity.

• Price & Value: Perhaps the most crucial comparison in the eyes of buyers: the YU7 is cheaper than the equivalent Model Y at every tier in China . As Lei Jun proudly pointed out, despite YU7 often leading Model Y in specifications, Xiaomi only charges ¥1k less for the base model (¥253.5k vs ¥263.5k) . The gap widens at higher trims – the YU7 Pro is about ¥33.6k (~$4.6k) cheaper than a Tesla Model Y Long Range AWD , and the YU7 Max undercuts the Model Y Performance by ¥25k (~$3.5k) . Xiaomi is also aggressive with launch promotions: early YU7 buyers get free upgrades worth over ¥30k (including leather and zero-gravity seats, power frunk, etc.) , further sweetening the deal. Tesla has responded in China with frequent price cuts and even a 5-year 0% interest loan offer on Model Y , but Xiaomi’s strategy of “using our base model to beat their top model” in features is clearly aimed at delivering better bang-for-buck. Lei Jun calls “value for money” a core part of Xiaomi’s auto playbook – much as it was in smartphones . For consumers cross-shopping, the YU7 offers a more luxury and performance for less money proposition. The Model Y, however, carries the weight of the Tesla brand, a proven track record, and higher resale value, which some buyers may factor in. Still, in pure pricing terms, Xiaomi has thrown down the gauntlet: YU7 gives you more car for your yuan. Chinese media have noted that this cost-performance focus – “use your standard config to challenge competitors’ highest config – is exactly how Xiaomi disrupted the smartphone market, and now they’re applying the same playbook to EVs .

In summary, Xiaomi’s YU7 emerges from the comparison looking like a spec-sheet champion – offering a bigger, longer-range, feature-packed SUV for less money than the segment leader. It decisively checks the boxes in areas Chinese consumers care about: range anxiety, tech features, and value. However, the real-world contest will depend on factors beyond specs: Tesla’s advantages in brand cachet, software, charging network, and production scale will all influence how the YU7 ultimately fares in the market.

Lei Jun’s Take: “Challenge Model Y – We Dare to Compare!”

From the outset, Xiaomi has been unabashed in framing the YU7 as a Model Y fighter, and no one emphasized this more than CEO Lei Jun himself. At the launch event, Lei Jun recounted how Tesla issued a sort of challenge earlier in the year when promoting the refreshed Model Y. “At the start of this year, Tesla put out a message – ‘feel free to compare to the new Model Y’, which really surprised me,” he said . While other domestic automakers stayed silent, Lei Jun immediately responded, “I saw no one else react, so I directly replied: ‘OK’.” . On stage, he made it clear that Xiaomi was answering Tesla’s “battle invitation” head-on: “Tesla previously said, ‘go ahead and compare’, and we at Xiaomi just won’t accept defeat. Today we officially take up Tesla’s invite to compare.” . This bold proclamation set the tone – Xiaomi is not shy about stacking the YU7 against the Model Y feature-by-feature.

Lei Jun unveiling the Xiaomi YU7 and directly comparing its specs to Tesla’s Model Y at the 2025 launch event. He repeatedly asserted YU7’s advantages in range, power, and pricing over its Tesla rival.

Throughout the presentation, Lei Jun mentioned “Model Y” at least 10 times , systematically highlighting where YU7 comes out on top. He showed comparison slides: for example, noting YU7 Standard’s 835 km vs Model Y’s 593 km range, the inclusion of LiDAR, 700 TOPS chip and premium suspension standard on YU7 vs none of those on Tesla, all while YU7 costs less . When summarizing, Lei Jun declared that in many metrics YU7 leads the Model Y, yet it’s still cheaper by ¥10k – a point of pride for Xiaomi’s pricing strategy . He also frankly acknowledged Tesla’s strengths. “Tesla is really formidable,” he said. “Model Y has been the global best-seller for years and it’s indeed a very good car – they are extremely confident.” He even admitted, “Model Y’s efficiency and FSD are ahead of us, we still need to learn from them.” . But that didn’t temper his competitive fire: “Tesla daring to send out that challenge – I dare to compare. We Xiaomi are simply not afraid.” He added, “Even if we can’t beat them, it’s okay – but we won’t back down.” .

Lei Jun had been beating the war drums even on social media leading up to the event. Many followers asked him if the YU7 could ever outsell the Model Y in China. His response: “Xiaomi YU7 will continue to challenge the Model Y! I’m especially confident in our product strength.” . He tempered it by saying, “As for sales, we’ll see how everyone feels after launch.” – acknowledging that winning over the market may take time. Internally, he even labeled the goal of overtaking Model Y as “an outrageous target”, but one worth pursuing. To that end, Lei Jun revealed one secret of the pricing decision: just the night before release, he decided the YU7 Standard must start at ¥1 万 below the Model Y’s price , calling it a small sacrifice for greater market impact. He also joked how rumors of YU7’s price had run wild (some expected as low as ¥199k), and that he had to disappoint those dreaming of a sub-¥20 万 SUV – “with Model Y’s level of specs, it has to be around ¥300k”, he had said in May . Indeed, Xiaomi ultimately priced it in the mid-¥20 万 range, aligning with his philosophy of “compete on product, tech, and user value, not just price” .

Notably, Lei Jun’s aggressive stance was partly inspired by Tesla’s own provocation in China. In early 2024, Tesla China’s marketing for the Model Y refresh included the phrase “尽管对比” (“feel free to compare”), which many saw as Tesla challenging domestic EV makers to a spec duel . By picking up that gauntlet, Xiaomi got to cast itself as the brave upstart willing to take on the giant. This narrative plays well in China’s tech community – reminiscent of how local smartphone brands would openly compare their features to Apple’s or Samsung’s in launch events. Lei Jun is a master of this approach, and with YU7 he made the Tesla comparison explicit at every turn. He even invited industry peers to witness it: executives from Nio, Li Auto, Lotus and other carmakers reportedly attended the YU7 launch in the audience , underlining how it felt like a milestone moment in the domestic EV scene. At the end of the event, Lei Jun proclaimed: “The competition in this market is fierce, but I still have confidence in the Xiaomi YU7.” The message was clear – Xiaomi is all-in on this fight.

Market Positioning and Strategy

Xiaomi is positioning the YU7 as the flag-bearer of its automotive ambitions – a model that firmly plants Xiaomi in the hotly contested electric SUV segment dominated by Tesla. Lei Jun stated that SUVs are the true battleground of the auto industry, given their popularity , and that with YU7, “Xiaomi is finally entering the main battlefield, facing the strongest opponent and the most brutal competition.” . This framing shows Xiaomi’s strategic aim: use the YU7 to prove itself against the benchmark, much as the earlier SU7 sedan was aimed at Tesla’s Model 3. (In fact, Xiaomi’s first model, the SU7, was deliberately benchmarked to Model 3’s size and performance; Lei Jun proudly noted the SU7 went on to beat the Model 3 in Chinese sales for several consecutive months .) With YU7, Xiaomi zeroed in on the mid-sized pure electric SUV segment (¥250k–350k range), which is not only huge (over 50% of China’s passenger EV market ) but also crowded with competition. Many domestic rivals – Nio’s sub-brand (Lekker L60/L90), Li Auto’s upcoming pure EVs (L7, L8), Xpeng’s new G7 SUV, Geely’s Zeekr, Huawei’s AITO models, etc. – are all vying to “kill the Model Y.” In 2024, there was even talk of a “encirclement of Model Y with a wave of six new Chinese SUVs launching in the same week, all claiming to target Tesla . Yet, by year-end, Tesla “still sat atop Bright Summit”, outselling each challenger by a large margin . This history isn’t lost on Xiaomi. The company is entering a fierce market melee, but it brings some unique advantages to the table.

First, Xiaomi leverages its background as a consumer electronics powerhouse. It has strong cash flow and a tight-knit supply chain from its smartphone business, which can help sustain the car venture . The YU7’s aggressive pricing despite high-end components hints at Xiaomi possibly accepting razor-thin margins initially – something it famously did with phones – to gain market share. Xiaomi also emphasizes its sprawling ecosystem: the “人车家全生态” (People–Car–Home full ecosystem) vision means Xiaomi wants the car to be an extension of one’s digital life . They’ve partnered with companies like BYD, GAC Toyota, and Nissan to integrate Xiaomi’s platform and services, aiming for a more open, connected ecosystem instead of a closed garden . This could attract tech-savvy customers who already use Xiaomi gadgets, as the YU7 can seamlessly interface with their phones, smart home devices, and cloud services.

Secondly, Xiaomi’s “specs-first” product strategy – focusing on design, power/performance, and value– is its play to differentiate in the EV arena . Chinese media note that these three “cards” are exactly what made Xiaomi phones popular, and now YU7 is following suit . Design-wise, Xiaomi made sure the YU7 stands out with a sporty luxury vibe to capture young buyers and even female consumers (Lei Jun joked that Xiaomi doesn’t want to be seen as only a “geeky male” brand – “this time we want to win the hearts of the ladies”, he said about making cars look good) . Performance-wise, Xiaomi isn’t holding back – even benchmarking lap times and endurance tests to show the car’s mettle . And on value, as discussed, Xiaomi is undercutting and out-equipping competitors, embodying the “honest pricing” reputation it cultivated in electronics. As a result, YU7 is marketed as a car that redefines the standard for luxury high-performance SUVs at this price, essentially attempting to move the goalposts for what buyers expect for ~¥300k.

The market positioning of YU7 is thus a bit of a tightrope: it’s sold as a premium product (with the word “luxury” used frequently in promotions ), yet it’s also pitched as a price-performance champion. Xiaomi likely wants to lure would-be Tesla buyers by saying “why settle for a barebones Model Y when you can have a loaded YU7 for less?” At the same time, they want to convince premium-minded customers that YU7 belongs in the conversation with higher-end brands. This dual positioning is reflected in how Xiaomi talks about competitors: they consider Model Y the main rival, but also acknowledge others like Xpeng G7, Nio’s offerings, Geely’s Zeekr 7X, Avatr 07, etc., as being cross-shopped . In one analysis, an expert noted “some might say YU7’s competitors are Xpeng G7, Zhijie R7, Avatr 07, etc., but I believe the true opponent is only one: Tesla Model Y.” – reinforcing that Xiaomi itself primarily measures success by whether it can steal significant sales from Tesla.

Crucially, Xiaomi is aware that having a great product is only half the battle. The other half is execution: manufacturing quality and scale. As a newcomer to car manufacturing, Xiaomi faces a learning curve in quality control and production ramp-up . The initial SU7 had some hiccups and long wait times, prompting Xiaomi to invest in expanding factory capacity. They currently have one factory in Yizhuang, Beijing, which can produce maybe 150k cars/year per phase, and they’ve set a bold target of 35万 (350k) vehicle deliveries in 2025 . Achieving this would require nearly matching Tesla’s volume in China. Xiaomi is already building a second phase for the factory (another 150k capacity) and even acquired land for a third factory for future expansion . However, capacity bottleneck is a real concern: even before YU7’s launch, the SU7 had ~30-week delivery backlogs due to limited production capacity . Additionally, YU7, despite sharing the Modena EV platform with SU7, has 90% new parts and differs in assembly processes, meaning ramping it up will not be trivial . Chinese media have pointed out that production ramp is the biggest variable in Xiaomi’s bid to challenge Model Y . If Xiaomi can’t build YU7s fast enough or with consistent quality, impatient buyers might still opt for readily available Model Ys (Tesla’s Shanghai Gigafactory is a well-oiled machine turning out thousands per week).

Lei Jun has publicly addressed this, saying the factory is highly automated and currently able to do 20–30k cars per month at full tilt . He was visibly excited by the huge YU7 order influx and said Xiaomi will “go all out to expand production and speed up delivery” to meet the demand . Clearly, scaling up is part of Xiaomi’s strategy to drive costs down and hit that aggressive price. The company has strong financials (over ¥1113 亿 revenue and ¥107 亿 profit in Q1 2025 across all businesses ) to support initial losses or heavy investment in the auto division. In essence, Xiaomi is prepared to play the long game: it sees cars as another smart device, one that can lock users into its ecosystem, and it is willing to disrupt the market by competing on multiple fronts – specs, price, integration – simultaneously.

Reception and Reactions in China

The reaction in Chinese tech and auto circles to the Xiaomi YU7’s launch and its open Model Y rivalry has been intense and mixed – a blend of excitement, skepticism, and patriotic pride. Here are some key takeaways from media and public commentary:

Overwhelming Interest: There’s no denying the YU7 created a huge buzz. The fact that Xiaomi secured 200k orders in minutes blew many people’s minds . Social media was flooded with discussions, and Xiaomi’s own community forums had snatch-up guides circulating even before the launch – fans sharing tips on how to place orders fastest, as if it were a new flagship smartphone release . This kind of enthusiasm is rare for a new entrant in autos, and it signals that Xiaomi’s brand power is successfully carrying over to its vehicles. Many users expressed admiration for how Xiaomi delivered on the promised specs: *“真香!” (“really fragrant/good!”) was a common phrase in forums, joking about how initial doubters are won over by the final product’s value. The nine paint colors, the luxury interior, the high specs – these all played well with a consumer base that loves feature-rich products. On Zhihu (China’s Quora), when asked “Why does a luxury-oriented YU7 still compare itself to Model Y?”, one answer noted that Xiaomi’s strategic calculus is spot on: Model Y is the segment benchmark in consumers’ minds, so by publicly measuring up to it, Xiaomi precisely communicates YU7’s positioning and grabs attention . In other words, Xiaomi tying itself to Tesla in the narrative has already succeeded in generating free publicity and a perception of rivalry – which for a newcomer, is valuable in itself.

That said, almost every analyst followed their praise with caveats about challenges. First, there’s Tesla’s entrenched lead. As 21st Century Business Herald put it, Model Y is a “mountain” that countless have tried to climb . They recounted how a flurry of Model Y challengers in 2024 still couldn’t dethrone Tesla . Xpeng’s CEO He Xiaopeng, upon Lei Jun’s launch, even commented hopefully about a “双7组合” – possibly referring to Xpeng G7 and Xiaomi YU7 both being new 7-series SUVs – that together might change the SUV market dynamics . This camaraderie among Chinese EV makers highlights that beating Tesla is seen as a collective challenge. Nio’s Li Bin and Li Auto’s Li Xiang each have their strategies (Nio with a new brand, Li Auto by building more charging stations than Tesla) . So observers are viewing Xiaomi as joining this broader “Team China vs Tesla” contest. Many consumers with nationalist leanings cheer for Xiaomi, saying things like “Finally, a Chinese SUV that can go head-to-head with Tesla”. Some Tesla owners even commented on forums that YU7’s arrival will push Tesla to further improve or cut prices, which benefits everyone.

Skepticism and Critiques: On the skeptical side, commentators pointed out Xiaomi’s lack of experience in car building. Some initial SU7 owners faced long waits, and a few quality issues (typical new automaker woes). A Jiemian article bluntly stated, “the biggest variable is production capacity”, cautioning that Xiaomi must overcome manufacturing bottlenecks to truly challenge Model Y’s volume . Others question whether Xiaomi can maintain quality while ramping up output to tens of thousands per month – Tesla’s advantage has been exactly that: scaling with acceptable quality. Additionally, while YU7’s spec lead looks great now, Tesla is not standing still. In late 2024, Tesla introduced a refreshed Model Y (nicknamed “Project Juniper” globally) with improved range (~707 km CLTC on the new long-range variant) and new features like an 8-inch rear screen and better interior materials . If those upgrades reach China in 2025, Tesla will narrow the gap in some areas (for example, offering a rear display, which was previously a Xiaomi selling point). Tesla has also shown a willingness to adjust prices and leverage its superior profit margins to stay competitive. Some financial analysts warn that Xiaomi might be entering a price war which could hurt its margins long-term if Tesla cuts prices further or if other Chinese brands undercut Xiaomi in turn.

Chinese tech media also highlighted two notable shortcomings of the YU7 despite all the strengths: charging infrastructure and autonomous driving . As mentioned, without its own fast-charger network, Xiaomi cars rely on third-party stations – not a deal-breaker, but when rivals like Nio provide battery swap stations and exclusive charging lounges, Xiaomi might need to innovate on charging solutions (Xiaomi has said it will offer some charging subsidies and home charger packages to YU7 buyers, but details are scant). On autonomous driving, while Xiaomi’s hardware is top-notch, the software needs to prove itself. Tesla’s FSD beta (though not officially in China) and Xpeng’s Navigation Guided Pilot have set high bars. If Xiaomi’s “HAD” system lags significantly in real-world performance, tech-savvy buyers could be less impressed. Early reviews say it’s decent but still second-tier at the moment – something Xiaomi is racing to improve with its upcoming XLA AI model update.

Quality and Service: Another aspect discussed on forums is after-sales service and reliability. Tesla has years of service center expansion and over-the-air diagnostic experience. Xiaomi is new to this – will Xiaomi’s famed customer service in phones translate to cars? They are building stores and service centers (some saw the YU7 displayed at Xiaomi’s flagship stores attracting crowds). There’s cautious optimism because Xiaomi has a loyal fanbase and a reputation for listening to user feedback. The company has already integrated car support into its Mi Home app. However, as one Zhihu user pointed out, “a car is not a phone – you can’t just reboot it or replace it easily if something goes wrong”, emphasizing that Xiaomi must uphold higher standards of quality control and safety for its automotive venture . In response, Lei Jun in the launch event repeatedly stressed “安全是一切” (safety is everything) . He noted that Xiaomi invested heavily in vehicle safety features: YU7 uses ultra-high-strength 2200 MPa steel in key areas, and all trims include the full ADAS sensor suite not just for autonomy but also for active safety (collision avoidance) . This was partly to allay concerns after a series of EV safety incidents in China’s EV industry news. Tech bloggers acknowledged these moves, saying Xiaomi is addressing the learning points from its first model and competitors’ experiences.

Competitive Impact: The arrival of YU7 has certainly put competitors on notice. Xpeng notably launched its G7 (a coupe SUV priced from ¥239k, slightly under YU7) just a month prior, and their CEO’s public engagement with Lei Jun’s posts shows they see Xiaomi as both an ally in expanding the EV market and a rival for customers . Li Auto – which so far only sells extended-range hybrids and larger SUVs – might not compete directly in price, but Li Xiang’s emphasis on building a charging network to “surpass Tesla” could also serve to entice customers of cars like YU7, since Xiaomi doesn’t have its own network . Meanwhile, Tesla China’s response after YU7’s launch has been low-key publicly, but one can expect Tesla to highlight its software and efficiency advantages in marketing going forward. Chinese consumers now have an interesting comparison: “特斯拉 Model Y还是小米 YU7？” (“Tesla Model Y or Xiaomi YU7?”) – a question that would have seemed bizarre a few years ago, yet here we are in 2025, with a gadget company’s car legitimately challenging the industry’s EV champion.

In the court of public opinion, there’s a sense that Xiaomi has upped the ante. Many enthusiasts are excited to see a fresh contender push Tesla to innovate more. Some neutral observers caution that initial hype (and orders) don’t always translate into long-term success – citing how other Tesla rivals had strong starts but fell behind when Tesla slashed prices in early 2023, triggering a price war. But others argue Xiaomi’s entry is different because it’s not a niche startup; it’s a cash-rich tech giant with a huge domestic fanbase. “If anyone can pull a huge fan following to buy EVs, it’s Xiaomi,” wrote one commentator, noting how some Xiaomi phone loyalists waited to buy a car until Xiaomi made one. Indeed, anecdotal posts exist of users saying “I didn’t buy a Tesla because I wanted to support Xiaomi’s car once it comes.” That kind of brand loyalty could give Xiaomi a boost in converting sales.

In conclusion, Xiaomi YU7’s launch has been received as a milestone moment – a major tech company bringing its full force into the EV battle, directly targeting Tesla. The car itself has largely impressed onlookers with its specs and value, giving credence to Lei Jun’s confidence in its “product strength”. The strategy of boldly comparing to Model Y has earned Xiaomi both applause (for ambition) and scrutiny (for the daunting task ahead). As one media outlet succinctly put it: “Xiaomi YU7 aims to climb the Mount Model Y. It has three strong cards – looks, power, and value – and Xiaomi’s resources to back it up. But to truly summit, it must also overcome quality control and production bottlenecks.” The coming months will test whether Xiaomi can deliver on its promises at scale. If it succeeds, the YU7 might not only “knock the Model Y off its pedestal” in sales, but also mark a turning point where a Chinese tech firm proves it can stand toe-to-toe with Tesla. Even if it falls short, the consumer benefits of this showdown – better cars, better prices – are already being felt. For now, the Xiaomi vs. Tesla story has begun a new chapter, and the EV world is closely watching how this bold gambit plays out in 2025.

What Is Community Group Buying and Why Did It Boom in China?

Community group buying (CGB) refers to a model where residents in a neighborhood collectively purchase groceries and daily necessities at bulk discounts, with a local coordinator (a “group leader”) handling orders and next-day pickup for the group. This model took off spectacularly in China around 2018–2020, driven by the promise of ultra-low prices and convenience. Platforms eager to tap lower-tier cities and suburban markets offered steep subsidies – for example, one platform sold a 500g pack of eggs for only ¥1.99 (about $0.30), below wholesale cost, exemplifying the fierce price wars . The strategy worked to rapidly attract users (including many elderly and rural consumers new to e-commerce) with the lure of “low price + convenience.” By 2020, China’s community group-buying platforms were fulfilling over 100 million orders per day, and the market size surpassed ¥100 billion (∼$14B) . The COVID-19 pandemic further accelerated this boom – during city lockdowns, group buying became a lifeline for household supplies, cementing its popularity.

Behind the frenzy, however, cracks were already showing. The explosive growth led to a “hundred regiment battle” of competing startups and tech giants, all burning cash to grab market share. Homogenized offerings and subsidy-fueled expansion meant platforms competed not on service or quality, but on who could offer deeper discounts . This set the stage for an eventual reckoning once regulators and economics caught up with the industry.

Meituan’s Entry with Meituan Youxuan: Rapid Rise and Strategic Intent

Meituan – known for food delivery and local services – jumped into community group buying in mid-2020 with a new unit called Meituan Youxuan (“preferred selection”). The strategic intent was clear: Meituan saw CGB as a massive new growth avenue beyond food delivery, especially to penetrate China’s smaller cities and rural markets. Community group buying was initially touted as a “perfect solution” for bringing fresh grocery e-commerce to low-tier cities , aligning with Meituan’s mission to meet everyday consumer needs. By leveraging its huge existing network of delivery riders and merchants, Meituan integrated Youxuan into its ecosystem. For example, many Meituan food delivery couriers and mom-and-pop shop owners became group leaders for Youxuan, and the Meituan app funneled traffic to group-buy deals. This synergy gave Meituan an edge – its customer acquisition cost for group buying was reported to be about 30% lower than competitors, thanks to the efficiencies of piggybacking on Meituan’s high-frequency service platform .

Meituan Youxuan expanded rapidly nationwide, and at its peak it was handling a significant volume of orders. In fact, by 2021–2022 Meituan Youxuan and Pinduoduo’s Duoduo Maicai emerged as the two dominant players (a duopoly) in China’s community group buying arena . Meituan’s group-buy GMV (gross merchandise volume) reportedly reached the “hundred-billion-yuan level” at its height. However, after the initial land-grab phase, Meituan began pulling back from aggressive expansion. According to company disclosures, Meituan’s new initiatives segment (which includes Youxuan) accumulated ¥77.7 billion (~$11B) in losses from 2020 to 2022, and Meituan Youxuan’s annual transaction scale shrank from the peak ¥100+ billion range down to about ¥70–80 billion . In other words, growth had stalled and the cash burn was becoming difficult to justify.

Why Meituan Is Shutting Down Most of Its Group-Buying Business

After pouring resources into community group buying for several years, Meituan is now dramatically downsizing and refocusing Meituan Youxuan – effectively exiting most regions. Several key factors drove Meituan to this decision:

• Sustained Losses and Unviable Economics: The community group buying model proved extremely hard to profit from. Meituan’s financials show massive cumulative losses – over ¥80 billion (>$11B) lost on its group-buying operations over a few years . Despite scale, the unit economics remained poor: “the profit per order was under ¥1”, meaning break-even was nearly impossible . High operational costs (from cold chain logistics, labor, warehousing) and thin margins on cheap produce made the business a money pit. After investing huge sums, Meituan saw no clear path to profitability for Youxuan.

• Regulatory Pressure on Subsidies and Pricing: China’s regulators intervened once the subsidy wars spiraled out of control. In March 2021, the State Administration for Market Regulation issued new rules (“Nine Nos”) explicitly banning below-cost selling and other unfair practices in community group buying . This was a turning point that “dropped the regulatory sword of Damocles” on the industry . With predatory pricing curtailed by law, platforms could no longer offer absurdly cheap deals to hook users. Meituan and its rivals abruptly lost their key user acquisition tool (ultra-low prices), and the price advantage vanished, causing many bargain-seeking customers to lose interest . In short, the regulatory crackdown ended the era of “burn cash for growth,” forcing a pivot from growth-at-all-costs to a sustainability focus .

• Shifting Consumer Behavior: Chinese consumers’ habits around online grocery have evolved. During the boom, many consumers flocked to whatever platform had the cheapest offers. But once heavy discounts subsided, loyalty proved thin – users would not stick around without subsidies. Moreover, post-pandemic, consumers started valuing quality and immediacy more. A recent survey found that 60% of users now prioritize product quality when choosing grocery services, while those prioritizing lowest price fell to 30% . The core promise of community group buying – next-day pickup in exchange for slightly lower prices – became less appealing as on-demand grocery delivery (in under 1 hour) became more widespread. The convenience gap narrowed, and users grew less willing to plan purchases a day in advance just to save a small amount.

• Internal Cannibalization by Instant Delivery: Meituan faced a strategic clash between its own businesses. In parallel with Youxuan, Meituan has a thriving instant delivery service (via Meituan app’s “Meituan Flash” and other on-demand retail partners) that delivers goods from nearby stores within 30–60 minutes. Both services ultimately target the same grocery and FMCG needs, but one promised next-day pickup (group buy) and the other offered same-hour home delivery. Notably, around 80% of the product categories popular in community group buy (fruits, vegetables, snacks, daily essentials, etc.) overlap with those in Meituan’s on-demand retail sales . And from a consumer perspective, waiting until tomorrow morning versus getting it now is not a huge difference for most routine needs . This meant Meituan’s two models were effectively competing with each other for the same customers and orders. As one analysis put it, Meituan Youxuan’s shutdown is fundamentally to avoid “left hand vs right hand” internal competition between next-day group buying and instant delivery . Continuing to run parallel grocery businesses with such overlap would only cause resource “internal friction” and duplication. Shutting down the weaker one (Youxuan) to focus on the higher-growth, more efficient instant retail business became the logical choice .

• Structural Flaws in the CGB Model: Beyond Meituan’s specific situation, the community group buying model itself showed structural weaknesses. It turned out to be a heavy-asset, complex operation rather than the light model originally envisioned. To fulfill next-day orders, Meituan and others had to set up multi-layered warehouses (central depots feeding city “grid warehouses” down to local pickup points), incurring high fixed costs in rent, cold storage, and manpower . Night-time sorting of fresh produce and last-mile distribution to hundreds of pickup spots added to the cost. Spoilage and waste were significant – fresh fruits and veggies had high loss rates, which ate into margins . Moreover, community buy platforms found it hard to move beyond just produce and a limited SKU selection, capping each order’s value . And unless an area had very high order density, delivery efficiency stayed low and costs high . All these factors meant that even after the initial user growth, the model struggled to scale profitably. Meituan’s retreat is an acknowledgment that this “heavy” model is unsustainable without endless cash burn.

In essence, Meituan Youxuan became a story of good intentions colliding with harsh economics. The company announced in June 2025 that it would shut down or suspend Meituan Youxuan operations across most of China, exiting all but a few core regions, and reassigning staff to other units . Meituan framed it as a strategic restructuring: consolidating resources into more promising ventures (like instant retail and its core delivery business) and pulling out of persistently unprofitable markets . This marked a definitive end to Meituan’s community group buying experiment at national scale.

How Competing Platforms Approached (and Exited) Community Group Buying

Meituan was far from the only tech giant that jumped on the community group buying bandwagon. In the 2020–2021 boom, virtually every major Chinese e-commerce or on-demand platform launched a CGB service – and most have since scaled back or shut those efforts. Here’s how Meituan’s key competitors fared:

• Pinduoduo (Duoduo Maicai): Pinduoduo, which built its success on group purchase deals in e-commerce, naturally expanded into community grocery buying with Duoduo Maicai. Pinduoduo’s approach leveraged its deep roots in agriculture and rural supply chains. It established a farm-to-table logistics network – for example, working directly with cooperatives so that fresh mushrooms picked in Yunnan could reach customers the next day . This direct sourcing gave Pinduoduo an edge in controlling fresh produce spoilage to under 5%, versus ~8% industry average . Pinduoduo also ran a leaner operation: it adopted a relatively asset-light model, outsourcing much of its warehousing and delivery to third parties, which kept operating costs ~15% lower than Meituan’s more labor- and warehouse-intensive model . By 2023, Pinduoduo’s Duoduo Maicai and Meituan Youxuan were the last two giants standing in an otherwise shaken-out sector . Pinduoduo has continued to persevere in community group buying, but even it has become more cautious. Notably, in 2022 Pinduoduo redeployed significant staff and resources from Duoduo Maicai to build its new international e-commerce app Temu – indicating that Pinduoduo sees higher growth opportunities elsewhere . (According to a HSBC report, Temu already contributed 23% of Pinduoduo’s total revenue in 2023 .) Still, Duoduo Maicai remains operational nationwide, serving the users it acquired, though Pinduoduo is no longer expanding the service aggressively at all costs.

• Didi (Chengxin Youxuan): Ride-hailing company Didi Chuxing made an ambitious foray into community group buying in 2020 with “Chengxin Youxuan” (also known as Orange Heart). Didi had hoped to capitalize on its vast user base and driver network to diversify into local commerce. Chengxin Youxuan expanded rapidly in 2020, operating in dozens of provinces at its height. However, it became one of the earliest casualties of the CGB bubble. In early 2021, regulators’ admonishments about the price wars (and Didi’s own regulatory troubles in its core ride business) put heavy pressure on Orange Heart. By September 2021, Didi began shutting down Chengxin Youxuan’s operations across cities, and by the end of 2021 it had effectively exited the community group buy sector entirely . Didi’s short-lived experiment ended with heavy losses and layoffs, illustrating how difficult it was for even a cash-rich tech firm to crack this market. Orange Heart’s quick demise also foreshadowed the broader shake-out of smaller and non-core players once the golden days passed.

• Alibaba (Taocaicai / Taobao Maicai): Alibaba approached community group buying through several iterations. Initially, its grocery-focused arm Freshippo (Hema) dabbled in community group purchases (via “Hema Jishi”), and Taobao launched a grocery group-buy service (“Taobao Maicai”). In 2021, Alibaba merged these initiatives into a dedicated business called Taocaicai (“淘菜菜”), under its Taobao Deals division, signaling a serious push to challenge Meituan and Pinduoduo. Taocaicai tried to emulate the Meituan/Pinduoduo playbook, but Alibaba faced internal coordination problems. The company’s sprawling structure meant overlapping teams and channels all trying to do online groceries. According to one former Taocaicai employee, a single produce supplier had to interface with three different Alibaba divisions for similar grocery programs – an inefficient and costly setup . This lack of internal synergy kept Taocaicai’s costs high and its execution sluggish . By 2022, Taocaicai was losing momentum. Alibaba began retrenching the business, and in early 2023 it shut down Taocaicai’s next-day pickup service in many provinces, marking a retreat from community group buying. Alibaba has since shifted focus to other grocery retail models (for instance, shipping produce from origin wholesale markets to consumers, and leveraging its Ele.me platform for on-demand grocery delivery) . The Taocaicai brand still exists but on a much smaller scale after these pullbacks.

It’s worth noting that other players also fell by the wayside. JD.com launched “Jingxi Pinpin” for community group buys but folded it by 2021–2022. Startup contenders like Shihuituan, Tongcheng Life (Tongcheng Maicai), and others either went bankrupt or were acquired when the bubble burst . Xingsheng Youxuan, a regional pioneer from Hunan province, managed to survive by focusing on its home turf, but it too faced immense pressure during the price wars. In summary, virtually all the “community group buy” initiatives that exploded in 2020 have either exited or scaled way back by 2023. Pinduoduo’s grocery service is the lone big player still standing in a significant way – and even Pinduoduo has diversified its bets beyond this sector.

A New Era for China’s Retail Platforms: Instant Retail and the End of Subsidy Wars

Meituan’s retreat from community group buying signals broader shifts in China’s online retail landscape. The rise and fall of CGB has ushered in a new phase marked by two major trends:

Read more

In June 2025, Unitree Robotics completed its latest Series C financing round, drawing investments from prominent Chinese tech giants including Tencent, Alibaba, Ant Group, China Mobile, and Geely, pushing its valuation beyond $1 billion. Just months earlier, in February, a troupe of life-sized robots spun and leapt across China’s biggest televised stage, dancing in sync with human performers at the annual Spring Festival Gala . The humanoid machines – sleek black-and-white figures moving with uncanny agility – stole the show and captivated a nation. It was a defining moment for Unitree Robotics, the young company behind those robots. Born in a Hangzhou lab less than a decade ago, Unitree has rocketed from obscurity to the forefront of a booming robotics revolution. Its journey, marked by audacious engineering and viral showcases, reads like a tech industry fable – a scrappy Chinese startup taking on giants, betting on low-cost innovation to bring sci-fi robots into the real world.

Humble Origins of an Aspiring Innovator

Wang Xingxing, Unitree’s founder and CEO, did not follow a typical prodigy-to-founder script. Far from excelling in school, the Ningbo-born Wang struggled academically – repeatedly flunking English exams and languishing at the bottom of his class . But what he lacked in grades, he made up for in obsessive tinkering. As a child, Wang spent his meager allowance scavenging parts to build model airplanes, homemade batteries, even a makeshift turbine. One youthful experiment nearly ended in disaster – he forgot an electrolysis rig in the family home overnight, filling the air with chlorine gas . This hands-on ingenuity, however reckless, set the stage for his true calling.

At Zhejiang University of Science and Technology, and later in a master’s program at Shanghai University, Wang found his passion in robotics – specifically, compact four-legged robots powered by electric motors . In 2015, he entered a Shanghai robotics design contest with a clunky prototype canine he cobbled together from scrap metal and hobby motors. That creation, dubbed “XDog,” won second prize and a RMB 80,000 award (around $12,700) . More importantly, XDog validated Wang’s belief that small, electrically actuated quadrupeds could be the future of robotics. “Most innovations in society are amalgams,” he told Chinese media later. “You can combine the latest ideas from various industries to be cutting-edge – and actually be the best in the world” .

By 2016, Wang had proven the concept in academia but worried the market wasn’t ready. He initially took a job as an engineer at drone-maker DJI, shelving his startup dreams . Then fate intervened: videos of XDog performing trots and jumps went viral online, drawing attention from tech circles and even offers from companies to buy his design . Instead of selling out, the 26-year-old quit DJI and struck out on his own. In August 2016, with help from an angel investor, he founded Hangzhou Unitree Technology (the English name is a blend of “universal” and “tree,” reflecting Wang’s vision of “growing” technology in all directions ). The one-man project was now a company – albeit just a three-person team in a tiny office – embarking on an ambitious quest to democratize high-performance robotics .

Building an Army of Affordable Robot Dogs

From the outset, Unitree set itself apart with a laser focus on cost-effective engineering. In late 2017, the startup unveiled its first commercial product: Laikago, a mid-sized quadruped robot named after the Soviet space dog Laika . Laikago’s appearance immediately drew comparisons to the famed canine robots of Boston Dynamics. But unlike the American firm’s DARPA-funded prototypes, Unitree’s dog was designed for mass production and sale. It used off-the-shelf electric motors and clever mechanical design to keep costs down – a radically different philosophy from Boston Dynamics’ pricey, lab-bound machines.

Timeline of Unitree’s core robots: Starting with Wang’s graduate project “XDog” (2013–2016), the company launched a series of quadrupeds – Laikago (2017), AlienGo (2019), A1 (2020) and Go1 (2021) – before venturing into humanoids with H1 (2023) . Each generation improved performance while lowering costs, reflecting Unitree’s vertically integrated engineering strategy.

Crucially, Unitree invested in vertical integration early on. With few suppliers making affordable parts for cutting-edge robots, Wang’s tiny team started developing their own core components – high-torque electric motors, compact gear reducers, controllers, even LiDAR sensors . “We optimized everything from mechanical structure to control algorithms, and kept key hardware and supply chain under our control,” Wang told investors in 2021 . This approach gave Unitree a manufacturing edge: by 2021, it could sell a capable robot dog for as little as RMB 16,000 (~$2,500) – roughly 1/20th the price of Boston Dynamics’ Spot (which cost $75,000+) . The flagship Unitree Go1, launched in mid-2021, was billed as a “consumer-grade” quadruped that anyone could buy, complete with a low-cost camera and AI-powered follow mode for jogging or strolls .

The bet on affordability paid off. Within a few years, Unitree became the world’s first company to mass retail high-performance quadrupeds, shipping hundreds of robots to customers worldwide . As of 2021, the startup had delivered nearly 1,000 units globally – a figure that dwarfed the output of older Western rivals. Researchers, tech enthusiasts, and corporations from North America to Europe jumped at the chance to own an agile robot dog at a fraction of the usual cost. The Wall Street Journal even hailed these low-cost robotic dogs as a major achievement for Chinese tech . Many dubbed Unitree “the Chinese Boston Dynamics”, marveling at the similar-looking machines emerging from Hangzhou. Wang bristled at the comparison. “When I was making XDog in 2013–2015, [Boston Dynamics] hadn’t even figured out electric actuators,” he noted, alluding to the U.S. firm’s reliance on hydraulics back then. “We were also the first to market with four-legged robots… We might look similar to Boston Dynamics, but we’re on our own path.”

Indeed, Unitree’s path diverged by prioritizing pragmatism over perfection. Robotics insiders point out that the company’s real innovation was “leveraging the Chinese supply chain to create low-cost, high-performance, highly reliable” versions of known robot designs . In other words, Unitree productized what was once laboratory research. This ethos – more Tesla than DARPA – meant accepting some trade-offs in cutting-edge capability in exchange for immediate scalability. It also fostered a culture of constant iteration. Every new model built on lessons from the last, gradually closing the performance gap with more expensive competitors. By 2023, Unitree claimed over 60% of the global market share in quadruped robots , and (remarkably) had become profitable each year since 2020 – a rare feat for a robotics startup. In Wang’s words, “quadruped robots have the best chance to be the first bio-inspired robots to enter everyday life”, thanks to their balance of techiness and utility . Unitree was bent on making that prophecy come true.

Viral Sensations: Robot Bulls, Olympians, and a Super Bowl Cameo

While Unitree quietly sold robots to researchers and early adopters, it was mass media moments that vaulted the company into public consciousness. One of the first big breaks came on Chinese New Year’s Eve 2021. That night, hundreds of millions tuned in to CCTV’s Spring Festival Gala – the country’s most-watched show – and were treated to an upbeat skit featuring 24 prancing robot “oxen” in matching costumes . These horn-adorned performers were Unitree’s A1 quadrupeds in disguise, nicknamed “Benben the Ox,” sharing the stage with celebrities like Andy Lau and Guan Xiaotong . Their coordinated dance to ring in the Year of the Ox was a novelty that won China’s heart, earning Unitree instant name recognition as the maker of the famous “robot bulls.” The Gala appearance triggered a surge of interest and orders for the young company’s robots .

It was only the start of Unitree’s showbiz adventures. In February 2022, at the opening ceremony of the Beijing Winter Olympics, a fleet of Unitree quadrupeds marched out as part of a high-tech display, symbolizing China’s technological prowess on the world stage . A year later – improbably – Unitree’s robot dogs popped up in the U.S. spotlight. During the 2023 Super Bowl pre-game show, pop star Jason Derulo was flanked by a crew of dancing robots that bopped and shuffled beside human dancers. Those nimble electronic backup dancers were none other than Unitree Go1 units, specially choreographed for the NFL’s biggest party . For a Chinese hardware startup to feature in “America’s Super Bowl” was a surreal milestone, and it wasn’t lost on Unitree’s team. The company proudly shared clips of its Go1s grooving on the Super Bowl stage, noting how far their creations had come – from Chinese factory floors to the global pop culture arena.

Unitree’s robot dogs in action: Two Go1 quadrupeds retrieve a thrown discus and javelin during the athletics events at the 2023 Hangzhou Asian Games. Unitree’s four-legged robots were deployed as field assistants to fetch equipment – a task that delighted spectators and showcased how robots can step into mundane jobs at large-scale events .*

Not all of Unitree’s fame has been about fun and games. The company’s robots have also found practical use in high-stakes environments. Throughout 2022–2023, Unitree worked with Chinese authorities on pilot programs using its quadrupeds for industrial inspections and emergency response . In one case, its larger B2 model (an industrial-grade, waterproof robot dog) patrolled remote power grid stations for the State Energy Group, navigating 500 kV substations to check equipment readings . In another, modified Unitree robots were trialed by a provincial fire department, scurrying into hazardous zones that would be too dangerous for firefighters . These deployments, though early, signaled that Unitree’s machines were not just viral video fodder – they were maturing into serious tools.

By late 2023, the Unitree brand had frequent flyer status at tech expos and arenas worldwide. Its robots appeared at the World Artificial Intelligence Conference in Shanghai and the World Robot Conference in Beijing. They even made a cameo in a pre-show demonstration at the 2024 Paris Olympics handover event (according to Chinese media) . Each public outing brought a mix of astonishment and debate – were these performances mere gimmicks, or glimpses of a robotic future? Wang has acknowledged that showcase events serve a dual purpose: they demonstrate real technical progress while also generating buzz and “interim commercial value” in the long road to everyday robotics . In his view, having robots dance, flip, or even spar in robot combat tournaments (another spectacle Unitree has explored ) helps fund and inspire the next phase, where utility will take center stage.

Nothing, however, topped the spectacle Unitree pulled off entering 2024. At the January CES tech show in Las Vegas, the company unveiled its first humanoid robot to an international audience – and let attendees kick it to test its balance. The Unitree H1 humanoid, a glossy biped standing 1.8 meters tall, was showcased walking and reacting to force, impressing onlookers as one of the only life-size humanoids operating live on the CES floor . Tech vloggers eagerly stress-tested H1 (one proudly declared “I kicked a life-size humanoid robot!”) and found it remarkably robust . That same month, back in China, Unitree’s humanoids truly arrived in pop culture – appearing as comedic “guest stars” during the 2024 Lunar New Year Gala on CCTV . In a playful skit, an H1 robot nicknamed “Fuxi” introduced itself as a Gala comedian, cracking scripted jokes and dancing alongside human actors . The sight of a humanoid robot bantering on live national TV was equal parts eerie and exciting for viewers. It also symbolized a bold evolution for Unitree: from robot dogs trotting in the background to humanoid robots taking center stage.

From Four Legs to Two: Enter the Humanoids

Unitree’s leap into humanoid robotics was as rapid as it was daring. Internally, Wang had contemplated bipedal robots for years – he even built a crude biped as a college freshman in 2009 during winter break . But the company held off until it had honed its craft (and bank balance) with quadrupeds. By 2023, that moment arrived. “We launched our humanoid project in 2023,” Wang said, developing the first bipedal prototype in just six months by leveraging years of quadruped R&D . The result, unveiled in mid-2023, was Unitree H1, a full-size humanoid robot built with the same philosophy as its mechanical dogs: keep it relatively low-cost, iterate fast, and don’t reinvent the wheel unnecessarily.

H1’s specifications were ambitious. Standing about 1.8 m (5’11”) tall and weighing 47 kg, it was designed to walk, run, and even sprint at up to 3.3 m/s (7.4 mph) – a speed that would set a world record for full-size humanoids . Internally, H1 used custom high-torque electric joints and an array of depth sensors, benefiting directly from Unitree’s experience in building agile quadruped limbs. In late 2023, Unitree quietly began small-batch production of H1 and shipped a few units to early clients . The company wasted no time pushing the envelope: within months, software updates enabled H1 to pull off feats like in-place backflips and aerial cartwheels, thanks to reinforcement learning algorithms and motion-capture training . Videos of a robot the size of a person doing somersaults – and “kip-up” jump-to-stand moves – went viral, blurring the line between science demo and Kung Fu film. By early 2024, Unitree declared that its humanoids had achieved “superhuman flexibility”, mastering martial arts-like sequences that garnered widespread attention online .

To broaden its reach, Unitree took a page from its quadruped playbook: it introduced a smaller, cheaper humanoid model aimed at research and education markets. In spring 2024, the company unveiled Unitree G1, a 1.27 m tall “mini-humanoid” priced around $16,000 – dramatically lower than the six-figure sums competitors were quoting for human-sized bots . The G1, essentially a scaled-down cousin of H1, made its debut at the ICRA academic conference in 2024 . With 13 degrees of freedom per leg and a 35 kg body, G1 isn’t meant to replace human laborers, but it offers labs and developers a robust bipedal platform at the cost of a mid-range car. Unitree began mass producing G1 by late 2024 , instantly positioning itself as a leading supplier of humanoid research robots by volume.

While heavyweights like Tesla have also promised general-purpose humanoids (Tesla’s Optimus project is famously in the works), Unitree’s head start in actually delivering units has made it one of the most watched players in this nascent field . By 2025, industry data showed Unitree’s humanoid shipment volume among the global leaders . In China, it is certainly the poster child of the trend. Wang believes the timing is right. “Within the next year or two, robots will acquire generalized capabilities for both commercial and household tasks,” he predicted in mid-2025, “such as tidying rooms and delivering items” . The company has already upgraded its bots with manipulator arms to, say, open doors or assemble simple components – incremental steps toward real utility . Still, Wang is realistic that the “ChatGPT moment” for humanoids – a breakthrough that truly brings robots into daily life – “still requires some time” . In the meantime, Unitree is content to push the envelope one demo at a time, confident that each spectacle (be it dancing or boxing robots) moves the needle on technology and public acceptance .

The Tesla of Robotics? A Battle of Strategies

As Unitree scales up its humanoid ambitions, observers often draw parallels to other industry pioneers. One comparison is Tesla, in how both companies approach innovation through vertical integration and aggressive cost targets. Much like Elon Musk’s automaker rethought car manufacturing, Unitree has rethought how advanced robots can be built affordably. By designing nearly every part in-house – motors, chips, software – and owning its supply chain, Unitree controls costs and can iterate quickly . This strategy, combined with a willingness to embrace “good enough” components, allowed it to undercut competitors and sell robots by the thousand. “Our goal is to keep prices competitive while maintaining reasonable profit margins,” Wang told one interviewer, noting that as technology advances, the cost of quadrupeds (and now humanoids) will keep falling .

Another inevitable comparison is Boston Dynamics, the decades-old U.S. robotics lab famed for its jaw-dropping (and eye-wateringly expensive) machines. Wang has a nuanced respect for Boston Dynamics – admiring their technical feats but deliberately choosing a different path. “Boston Dynamics has been making robots for many years,” he said, “but I believed even before 2013 that hydraulic [actuation] couldn’t be commercialized – costs will never come down, and there’s always oil leakage.” For consumer and workplace robots, Wang insists electric motors are the only viable route, and on that front he feels Unitree had a head start. When Boston Dynamics announced in 2023 that it would retire its famous hydraulic humanoid (Atlas) to focus on an all-electric design, Wang wasn’t surprised – he wondered what took them so long . In effect, Unitree has tried to be to Boston Dynamics what a lean startup is to a government contractor: faster, cheaper, and more focused on real-world deployment.

Industry experts note that Unitree’s humanoids, while impressive, are not yet as advanced in autonomy or agility as Boston Dynamics’ showcase robots or Honda’s legacy Asimo. However, Unitree is narrowing the gap with astonishing speed, driven by China’s booming AI capabilities and its own engineering grit. “Previously, it took one to two years for a humanoid to learn to walk,” Wang said in 2024, “but now with large AI models, this can be achieved in a month.” He predicts that by the end of 2025, at least one company (perhaps his own) will unveil a general-purpose AI model for robotics, combining vision, language, and decision-making into a package that truly empowers autonomous helpers . If that happens, Unitree’s mix of good-enough hardware and improving AI could prove formidable, flooding the market with capable robots before more expensive rivals catch up.

Crucially, Wang’s vision extends beyond quadrupeds and humanoids: he speaks of an ecosystem of robots “of alternative forms” working together . Already, Unitree has branched into robotic arms (for fixed automation tasks) and even consumer gadgets like a motorized “fitness pump” device. In Wang’s ideal future, humanoids may build entire cities (he muses that governments could deploy 100,000 robots to construct a metropolis) while swarms of smaller bots handle microscale tasks, “even shrinking down to the size of cells” to transform our environment . It’s a sweeping techno-optimist view, one that aligns with his lifelong dream of “advancing society and human happiness through technology.”

Market Traction: From Chinese Factories to Global Tech Labs

For all the futuristic talk, Unitree’s credibility stems from tangible market traction – both in China and abroad. The company has proven adept at turning R&D into revenue. Since 2017, Unitree has deployed robots across dozens of industrial projects in China, carving out niches in energy, utilities, and public safety . Its robots patrol power plants and electrical substations, often in harsh outdoor conditions, transmitting real-time data and performing routine inspections that free up human technicians . In one pilot with State Grid, a Unitree dog autonomously navigated a 220 kV substation, using thermal cameras to check gauges and detect hotspots . And in a recent 5G-enabled “smart factory” trial, multiple humanoid robots were coordinated for assembly tasks over a wireless network – a demonstration of what multi-robot collaboration could look like on tomorrow’s shop floors . These early deployments, though limited, give weight to Wang’s claim that humanoids will first gain traction in industrial and commercial applications (not immediately in people’s homes) .

Internationally, Unitree has punched above its weight in reaching customers. It was the first company to sell quadruped robots directly online to the public, shipping to overseas buyers through its website and distributors as early as 2018 . By 2021, the Go1 model had pre-orders from over 30 countries . Academic labs from MIT to ETH Zurich bought Unitree robots as affordable research platforms; tech hobbyists in Silicon Valley and Berlin proudly unboxed their own robot dogs, sharing feedback that in turn helped Unitree refine its designs. The company’s global footprint is evident in its media coverage: Unitree devices have been featured by BBC and CNN, and reviewed by countless YouTubers. This grassroots adoption helped Unitree quietly dominate unit sales. According to an early investor, Unitree’s quadrupeds by 2025 made up more than 60% of the world’s installed base of legged robots . In other words, out of every 10 four-legged robots out there, at least six are likely from Unitree – an astonishing statistic for a firm barely nine years old.

One reason for this success is that Unitree cultivated multiple market segments. It sells slightly different robot variants for consumer, educational, and industrial use, tailoring price points and support accordingly . A tech enthusiast can buy a Go1 or Go2 dog and have it running within minutes of unboxing . A university might opt for an AlienGo or B1 model with open APIs for research, accepting a bit more setup time. For industry clients, Unitree offers on-site integration services (hence why an industrial deployment can take weeks to implement) . This segmentation ensures that from a high school robotics club to a state-owned enterprise, there’s a Unitree product to fit. The company’s aggressive pricing has forced others to respond – a dynamic much like how Chinese drone makers undercut competitors in the 2010s to dominate that market.

Unitree’s growing clout hasn’t gone unnoticed by China’s leadership and tech establishment either. In early 2025, Wang Xingxing was invited to Beijing for a high-profile private-sector summit hosted by President Xi Jinping, sharing the front row with titans like Huawei’s Ren Zhengfei and Xiaomi’s Lei Jun . The 35-year-old Wang was by far the youngest of the bunch, a symbolic inclusion signaling the government’s support for next-generation innovators. After the meeting, Wang spoke to media about the rapid progress of AI-powered robots and emphasized the need for AI models tailored to robotics . Around the same time, Hong Kong’s Chief Executive made a point to visit Unitree’s Hangzhou headquarters, even encouraging the company to consider a future IPO in Hong Kong . These nods from officials underscore how strategic China views humanoid robotics – as a frontier to conquer in the broader tech race.

Big Tech Backing and Billion-Dollar Bets

The market momentum and strategic importance of Unitree have translated into a flood of investor interest, especially in the past two years. The company’s funding journey reads like a who’s-who of Chinese tech finance. It reportedly raised a small angel round in 2016 (just RMB 2 million) to get off the ground . By 2019, it secured a pre-Series A led by Sequoia Capital China’s seed fund, injecting several million RMB to ramp up production . A larger Series A followed in mid-2021 – Shunwei Capital (Xiaomi founder Lei Jun’s fund) led a multi-million dollar investment that valued Unitree as a rising star in frontier tech . At that point, Unitree had already delivered on its Gala fame and global shipments, making it one of the most promising robotics startups in China.

Read more

Ladies and gentlemen, good afternoon.

I’d like to thank GSMA for inviting me. It’s a great pleasure to join you here at MWC Shanghai. Since the founding of GSMA in 1987, the telecom industry has experienced nearly 40 years of rapid growth. These days, however, we have entered an era of technology oversupply: in major markets, basic consumer needs have been fully met and revenue growth has stagnated. The whole industry is now in a stable development period and facing growth challenges. Industry players are working hard to explore opportunities and pathways for growth. I want to use this opportunity to share a few of our findings — which may suggest a few pathways to drive growth. I will focus on four main points.

1. Ramp up for changes and meet new demands with high growth potential

The telecom market has become very mature, but even a mature market is not set in stone; change is everywhere. One effective pathway to achieve growth is to start from the end users, gain insight into the changes in their needs, and then, through network build-out and optimization, create products and services to meet those changing needs — especially needs with high growth potential. Let me give two examples of such growth-oriented demands.

The first example is delivery riders, including food delivery couriers, parcel couriers, and ride-hailing drivers. Globally, this is a new and fast-growing user group: from 30 million in 2020 to 70 million in 2024, and it is expected to reach 160 million by 2030, about 5% of the global workforce. They bring great convenience to our lives, and for carriers, they represent a rapidly growing segment of new high-value users. For example, when food delivery riders are delivering orders, they need to call the customer upon arrival or if any issues need communication. Their average monthly call duration — MOU (Minutes of Usage) — is four times that of a typical user. In the Chinese market, food delivery riders’ MOU is around 800 minutes per month. When they are not taking orders and have downtime, they often watch online videos to relax, so their data usage — DOU (Data Usage per user) — is twice that of a typical user. As a result, a rider’s ARPU (Average Revenue Per User) is 1.6 times that of a normal user.

Another example is livestreamers. With the rapid rise of mobile Internet, livestreaming has become a new profession that is growing quickly worldwide. The number of professional livestreamers increased from 10 million in 2022 to 50 million in 2024, and by 2030 it is expected to reach 130 million, roughly 4% of the global workforce. The content that streamers create allows people to see a more exciting world and a richer daily life. For carriers, they too are a fast-growing group of new high-value users, and the network has become their production system. A strong network is essential for livestreamers: their DOU is five times that of a typical user. In the Chinese market, a livestreamer’s monthly DOU can reach 100 GB, and their ARPU is four times that of a typical user.

In fact, many new user groups, new devices and applications, and new user behaviors have emerged. In addition to delivery riders and livestreamers, there are e-sports players, smart glasses, trendy smart gadgets, working on high-speed trains, binge-watching TV series on the subway, and so on. These changes are developing rapidly and bringing new demand characteristics — for example, different types of services, different usage locations, sudden surges in demand, real-time requirements, and ubiquitous connectivity. Looking at the global market, those carriers who have caught these growth-oriented needs are seeing relatively better growth, while those who haven’t are facing greater challenges in achieving growth.

2. Boost HD video supply and consumption through coordinated effort across the ecosystem

With the explosive growth of short videos worldwide, they now account for about 50% of mobile data traffic. However, the supply and consumption of high-definition video still haven’t been handled well. On one hand, some consumers can’t fully use up their data plans, causing some of them to downgrade their plans; on the other hand, we still don’t have an adequate supply of HD video for users to consume.

We all know that the value of high-definition video is definite. For consumers, it means a better viewing experience. For OTT platforms, HD video can improve their business conversion rates. For carriers, 1080p video brings about 5× the traffic of 360p. Suppose a user watches 50 hours of video in a month and it’s all in 1080p — that would mean consuming roughly 31.5 GB of data. Meanwhile, the average monthly DOU in China in 2024 was 18.2 GB, which shows there is huge potential yet to be tapped.

But looking at the current share of HD video in overall consumption, the gap is obvious. According to statistics from typical major cities in China, content at 1080p and above accounts for only 22% of mobile video traffic. There are multiple reasons behind this gap. From the OTT perspective, the price of outbound Internet bandwidth is still somewhat high, so during peak hours they lower video bitrates to control costs. For carriers, the ToB (enterprise) business is an important growth avenue, and if they lower outbound bandwidth prices, it would impact their enterprise revenue. From the device perspective, continuously watching HD video drains a lot of power, and when the battery level falls below around 30%, video apps automatically reduce the bitrate.

So the challenges are multi-faceted and require systematic solutions. First, for carriers: they should price outbound Internet bandwidth appropriately to encourage OTT providers to increase the supply of HD content, thereby boosting HD content consumption and overall net revenue. Perhaps we can try this experiment: On the premise of maintaining the current outbound bandwidth revenue for OTT content providers, have all OTT providers supply videos in HD (1080p) and see how much more traffic consumers use and how much DOU increases, and then calculate the cost-benefit. For OTT content providers, they should increase their supply of HD content, enhance the user experience, and achieve higher conversion rates. At the same time, network equipment vendors need to keep innovating to improve network capabilities and reduce the per-bit cost of bandwidth. Finally, device manufacturers need to reduce the power consumption of video streaming and improve battery life, in order to ensure a better experience for users.

3. Bring 5G to every car for new growth in intelligent connected vehicles

Letting 5G connect every car is a common aspiration of car owners and carriers, but automakers don’t necessarily think so. This is why currently only about 30% of vehicles are 5G-connected. Why are automakers unwilling to use 5G connectivity for cars? The main reason is that the costs are too high — both the 5G IPR (intellectual property rights) fees and the price of the T-Box module are high. Therefore, we hope GSMA can play a role, and that through joint efforts across the industry we can keep the 5G IPR fees for vehicles within a reasonable range. Meanwhile, we also need to ensure healthy competition among carriers in the connected-vehicle market: we should avoid a scenario where the price of connecting each car gets pushed too low, turning it into a low-value battleground. Additionally, I want to emphasize two points:

• Both the car and its cockpit need network connectivity, but their needs and business models are different, so they should have separate connections. For the cockpit’s connectivity, a B2C business model can be used, sharing the smartphone ecosystem — letting consumers choose the type of connectivity and purchase data plans on their own. For the vehicle’s T-Box connectivity, a B2B business model should be used, with the car company continuing to decide the connectivity method.

• When driving autonomously, cars must be able to sense their surroundings and make decisions on their own, without relying on network connectivity as a prerequisite.

4. Bring FTTR to micro and small businesses to seize opportunities in the intelligent era

FTTR (Fiber To The Room) has already demonstrated its value in meeting the needs of high-end home broadband users, effectively boosting the ARPU of high-end broadband customers and increasing carriers’ revenue. By 2025, the number of FTTR users in China will reach 75 million, whereas outside of China there are only just over 500,000, so there is enormous potential for growth. With the advancement of FTTR technology, it’s now possible to use a single fiber for access and extend it via indoor optical splitters to up to 128 fiber drops to different rooms or locations. This kind of FTTR solution can effectively address the issues currently faced by street-side shops, small offices, or campus venues: for example, poor Wi-Fi experience for customers and employees, no guarantees for critical applications (such as point-of-sale systems), an insufficient number of device connections, and lack of O&M (operations & maintenance) support when network faults occur. An FTTR network can meet the requirements for high-speed, stable, low-latency connectivity with full Wi-Fi coverage. Globally there are over 500 million micro and small businesses. If carriers can leverage FTTR to meet these needs, it will represent a significant opportunity and pathway for growth.

These are the few findings I’ve shared today, which may also serve as a few pathways for growth for carriers. Of course, different countries and different carriers operate in different environments and face different competitive landscapes, so opportunities and pathways for growth will differ as well. Huawei is willing to work together with carriers to jointly explore opportunities and pathways for growth, and to help carriers achieve sustained growth.

Pioneering Open Models: From Abab to M1

Shanghai-based AI startup MiniMax has quickly emerged as a model-driven company, pushing the boundaries of large language models (LLMs). In January 2024, it launched Abab6, China’s first Mixture-of-Experts (MoE) LLM , and followed up in April with Abab 6.5, a trillion-parameter model supporting 200,000-token context lengths . The Abab 6.5 series (including an optimized 6.5s variant) was shown to approach the performance of leading models like GPT-4 and Claude-3 on core tasks . For example, Abab 6.5 perfectly handled “needle-in-haystack” long-context tests, correctly answering 891/891 queries where a single irrelevant sentence was hidden in lengthy text . These early bets on MoE – at a time when most peers stuck to dense models – helped MiniMax leapfrog into the top tier of China’s AI model startups .

Now MiniMax has open-sourced its most advanced model yet: MiniMax-M1, unveiled on June 17, 2025. Billed as the world’s first open-weight large-scale hybrid attention reasoning model, M1 combines an MoE architecture with a novel “lightning attention” mechanism . The model is massive – 4.560 trillion parameters in total – though only about 459 billion are active per token thanks to MoE gating . M1 boasts an industry-leading context window of 1,000,000 tokens for input and can generate up to 80,000 tokens in output . This ultra-long context (eight times longer than competitor DeepSeek’s R1 model) is a milestone for handling lengthy documents and complex reasoning .

Performance and efficiency: Despite its size, M1 is designed for efficiency. Its “lightning” attention greatly reduces inference cost for long outputs. Generating 80k tokens with M1 consumes only ~30% of the compute required by DeepSeek-R1 (similarly, 100k tokens need ~25% of R1’s FLOPs ). This efficiency and sparse activation mean M1 can perform “deep thinking” with far less hardware overhead. Early evaluations show M1’s overall capabilities are on par with top global models . Notably, in public benchmarks, M1 outperforms DeepSeek-R1 and Alibaba’s Qwen-3 on tasks involving AI agents and complex context processing . Its chain-of-thought reasoning is strong, albeit sometimes overly lengthy – a trait observed in cutting-edge reasoning models like OpenAI’s O1 and DeepSeek’s latest versions . On math and coding challenges, M1 currently trails the very latest tuned version of DeepSeek (R1-0528) , highlighting that each model has strengths and weaknesses. Still, M1’s coding ability is comparable to the first-tier models in the market, and its unprecedented memory depth is a major differentiator .

Technical highlights: MiniMax credits two core innovations for M1’s leap: an optimized linear (“lightning”) attention and an improved RL training algorithm called CISPO. Lightning attention breaks the traditional Transformer memory bottleneck, enabling context lengths in the millions of tokens with manageable latency . In fact, MiniMax-Text-01 (a precursor model released in January) used linear attention to handle up to 4 million tokens – 32× the context of GPT-4 and 20× that of Claude’s long-context version . For M1, this means even at 1M tokens it remains effective, unlocking use cases in long documents and conversations. The CISPO algorithm (“Clipped Importance Sampling Policy Optimization”) doubled the efficiency of reinforcement learning fine-tuning, cutting training time and cost dramatically . Training cost was a mere $534,700 (USD) – and only three weeks of time – to refine M1’s reasoning abilities . Such a low cost for a model of this scale is astonishing, demonstrating MiniMax’s focus on “cost-efficiency” innovation. The company achieved this by leveraging large-scale RL with minimal human labels, plus extensive parallelism and pipeline optimizations .

To summarize M1’s specs and features:

• Architecture: MoE + “lightning” attention hybrid (sparse expert model with efficient long-context attention) .

• Scale: ~4.56 trillion parameters (with ~45.9B active per token) .

• Context Window: 1,000,000 tokens input and 80,000-token output, the longest context in the industry .

• Efficiency: Only ~25–30% of competitor model’s compute needed for long outputs ; lightning attention cuts latency to 1/2700 of standard methods for large contexts .

• Training: Enhanced RL (CISPO) yielding 2× efficiency; full training cost just ~$0.53M .

• Performance: Among the top tier of open models – excels at tool use and long-context QA, while remaining competitive (if slightly behind the state-of-the-art) in code and math .

MiniMax has open-sourced the complete weights and technical report for M1 on GitHub and HuggingFace for the community . This openness makes M1 one of the most advanced models globally to have freely available weights. By contrast, many Chinese LLMs (and even OpenAI’s GPT-4) remain closed-source, so M1’s release is a noteworthy event for researchers and developers worldwide.

Why Open-Source? Strategy and Impact

MiniMax’s decision to open source M1 is both a technical and strategic move. Officially, the company describes M1 as aiming to provide a “high-performance, low-barrier” alternative for developers and enterprises . By sharing the model, MiniMax hopes to catalyze innovation in the AI community – allowing users to fine-tune it, deploy it privately, and build new applications on par with those powered by closed models . “Compared to closed models, open models give users full control, including the ability to fine-tune and advantages in data privacy,” noted a Hugging Face engineer at MiniMax’s open forum . The complete M1 weights and documentation are available on HuggingFace and GitHub, and MiniMax is working with the national supercomputing centers and open-source inference frameworks like vLLM to ensure developers can easily run M1 efficiently . In short, MiniMax is actively nurturing an open ecosystem around its model.

Beyond community good-will, there is a competitive rationale. Releasing a model of M1’s caliber counters the narrative that only tech giants have cutting-edge AI. Analysts say MiniMax is “proving through innovation that one can break the compute-and-capital barrier”, shifting the AI race from raw parameter count and cash burn to more meaningful metrics like efficiency and real-world value . This approach of “卷成本、卷效率” (competing on cost and efficiency) could pressure other model vendors to follow suit, focusing on genuine technical breakthroughs instead of engaging in a “parametric and valuation vanity fair” . Indeed, within China there has been a growing open-source movement: e.g. Baichuan Intelligence opened 7B/13B models in 2023, and Tsinghua’s Zhipu released smaller ChatGLM variants. But MiniMax-M1, as the first open model with trillions of parameters and a novel architecture, is poised to “have a profound impact on the domestic and even global AI large-model market” . Observers predict it will inspire more openness and collaboration, accelerating the overall progress of AI.

It’s worth noting that M1’s launch is part of “MiniMax Open Source Week”. The company planned five days of consecutive announcements following M1 . These include new tech and product updates each day, signaling a broader commitment to transparency and developer engagement. Such a campaign not only galvanizes MiniMax’s reputation among developers, but also serves as a bold challenge to its rivals: essentially saying “we will compete in the open.” By seeding an ecosystem of users and contributors, MiniMax could build network effects around its technology, something typically enjoyed by open-source projects. This strategy aligns well with China’s trend of model registration – as of mid-2025, over 430 AI models have been registered with regulators for public deployment – indicating a crowded field where differentiation and community support are key. MiniMax’s open-sourcing move helps distinguish it as a technology leader, not just another AI startup chasing hype.

Company Background: Unicorn with SenseTime DNA

MiniMax was founded in December 2021 by Yan Junjie, who previously was a Vice President and Research Director at SenseTime (one of China’s “AI Four Little Dragons” of the CV era) . Yan holds a PhD from the Chinese Academy of Sciences and is a respected expert in deep learning and computer vision, with 100+ papers and over 10,000 citations to his name . At SenseTime he led the development of core AI toolchains, face recognition systems, and general AI R&D . Co-founding MiniMax with him is Zhou Yucong, another early SenseTime engineer and accomplished algorithm specialist . Zhou, a Beihang University graduate, won international supercomputing competitions (ASC15, ISC17) during college and managed an AI lab team at SenseTime, bringing valuable experience in large-scale computing . The core team is heavily composed of alumni from SenseTime and other top AI labs, giving MiniMax a strong pedigree in AI research and engineering .

Interestingly, the company’s name “MiniMax” comes from the minimax algorithm in game theory . Yan Junjie chose it to reflect “finding the minimum of the maximum loss” – symbolizing an approach of managing risk and optimizing outcomes . This philosophy, of focusing on the best possible result in a worst-case scenario, seems to have guided MiniMax’s development strategy.

From the outset, MiniMax combined ambitious R&D with a product-focused mindset. The startup’s first product, Glow, launched in October 2022 as an AI companion app where users could create their own virtual agents with customized personality, appearance, and voices . Glow was a hit – within 4 months it amassed almost 5 million users, especially among young people . However, in March 2023, Glow was taken down from Chinese app stores due to regulatory clearance issues (AI services in China must undergo security assessment and filing) . This could have been a major setback, but MiniMax quickly pivoted to overseas markets. By June 2023, it launched Talkie, essentially Glow’s counterpart for global users . Talkie is an AI social/chat companion app available on Google Play and Apple App Store internationally. Thanks to its advanced AI character customization (users can design unique AI personas with specific looks, voices, and behaviors) and improved conversational naturalness, Talkie gained significant traction. By mid-2024, Talkie ranked among the top AI apps in the US, exceeding 3.8 million downloads in the US by June 2024 and reaching 11 million monthly active users globally . In fact, Talkie’s user base was reported to be about 60% of industry leader Character.AI’s, and in the U.S. its download numbers even surpassed Character.AI at one point – an impressive feat for a Chinese startup “going global.” Revenue-wise Talkie has been modest (under $1 million as of late 2024, reflecting a focus on growth over monetization) , but its success validated MiniMax’s product-driven approach and gave it an international footprint.

In September 2023, MiniMax brought a similar concept to the domestic market with Xingye (星野), an AI companion content community . Xingye allows users in China to create and share AI characters and interact with them, providing “emotional value” and entertainment. It saw strong uptake, topping domestic AI app charts (26th in social on the App Store) and reaching ~900k downloads in the first half of 2024 . By June 2024, Xingye had ~500k daily active users, making it the leading AI companion app in China . This dual focus on AI companion products – Talkie abroad and Xingye at home – distinguished MiniMax early on, demonstrating an ability to build engaging AI applications on top of its models.

Meanwhile, behind these apps, MiniMax invested heavily in building a full-stack multi-modal model suite. By late 2023 the company had developed not only large text models (the Abab series) but also models for speech, music, image, and video generation . In August 2024 it released abab-video-1, its first AI text-to-video model able to generate high-definition video from prompts . This model powered the app Hailuo AI, a creative platform for AI-generated videos. Following the release of abab-video-1, Hailuo AI’s usage skyrocketed – its web traffic in September 2024 jumped 860%, reaching the top of growth charts globally . Users marveled at the quality and realism of the AI-generated videos (including human-like characters and performances), which were considered a breakthrough in stability and detail . By 2025, Hailuo AI evolved into a multi-modal creative suite, supporting text, image, audio, and video generation for users and content creators.

To support third-party development, MiniMax also built an API open platform. This platform offers flexible cloud API access to its models for enterprises and developers, with enterprise-grade security and reliability . As of early 2025, the MiniMax open platform has attracted over 40,000 developers and corporate users worldwide . It provides services across more than 20 countries and numerous industries, enabling functions like text generation, customized chatbots, voice synthesis, image creation, and more . Partnerships span office software (e.g. integrating MiniMax’s model into Kingsoft WPS for smart document summaries, spreadsheet Q&A, and slide generation ), online communities (powering natural language search and multi-turn QA for a major forum ), healthcare (AI assistants for pharmacists and doctors via Gaosi Health, including medical Q&A with domain fine-tuning ), and beyond. These use cases showcase MiniMax’s push to commercialize its AI through B2B solutions while its own apps capture consumers.

MiniMax’s rapid growth and R&D haven’t gone unnoticed by investors. The company has raised several funding rounds since 2022, bringing in a who’s-who of tech VCs and strategic backers. Early rounds saw participation from IDG Capital, Hillhouse (Gaorong) Ventures, ZhenFund, Yunqi Partners, and MiraclePlus, among others . In mid-2023, MiniMax reportedly raised over $250 million (Series B) led by Tencent, which valued the company at around $1.2 billion . By March 2024, it secured a Series C of more than $600 million led by Alibaba, lifting its valuation to roughly $2.5 billion . Notably, Chinese gaming giant MiHoYo (maker of Genshin Impact) also invested, reflecting cross-industry interest in MiniMax’s tech . The hefty Alibaba-led round signaled strong confidence in MiniMax as one of China’s leading independent AI startups. According to Chinese media, by early 2025 MiniMax’s valuation had further climbed to over $4 billion (around RMB 25–30 billion) – making it one of the most valuable AI unicorns in the country. This valuation puts MiniMax ahead of its immediate peers; industry insiders note that its “full-stack technology value is higher” than others in the cohort .

Today, MiniMax is headquartered in Shanghai’s Xuhui District, which has become an AI innovation hub. The company is considered one of the “AI dragons” of Xuhui, alongside SenseTime (which is also in Shanghai) and other rising players . The local government highlights MiniMax as a leading example in a cluster of AI firms that together form a robust ecosystem of “tech R&D – application – industry synergy” in the district . With a team of top scientists and engineers and strong backing, MiniMax appears well-positioned at the intersection of China’s vibrant AI research scene and its booming demand for AI applications.

IPO on the Horizon: Plans and Speculation

With its rapid ascent, MiniMax is now reportedly eyeing the public markets. According to a recent Bloomberg report, the company is considering an IPO in Hong Kong as soon as late 2025, at a target valuation of around $3 billion . The plan, still in early discussion, has led MiniMax to hire financial advisors to prep the offering . Sources close to the company confirmed that there is indeed an internal notion of going public, though details like timing and exact valuation remain fluid . If it proceeds, MiniMax would be among the first of China’s new-wave large-model startups to IPO, marking a significant milestone for the industry.

MiniMax isn’t alone – rival Zhipu AI (another Alibaba- and Tencent-backed AI startup) is also reportedly preparing for an IPO . This suggests a broader trend of China’s AI unicorns seeking capital from public markets to fuel the next stage of growth (and perhaps to satisfy investors after several hefty private rounds). For MiniMax, an IPO would bring not just funding but also public credibility on a global stage, which could help it compete against both domestic peers and Western AI firms.

Media and analysts have reacted to the IPO news with a mix of optimism and caution. On one hand, MiniMax is seen as a leading contender in the “大模型六小龙” (the six prominent Chinese LLM startups) – now often distilled to a “four little giants” group due to shakeouts . Its strong valuation and backing by tech giants (Alibaba and Tencent both on the cap table) signal investor belief in its prospects . Successfully open-sourcing M1 just ahead of these IPO plans also boosts MiniMax’s story; it shows the company can deliver cutting-edge innovations that garner international attention and community adoption. This could bolster MiniMax’s pitch to public investors as not just another AI company, but one with a differentiated, technologist edge. Indeed, earlier this year some observers questioned whether MiniMax’s high valuation was sustainable without a breakthrough in “reasoning” models to match competitors like DeepSeek . The launch of M1 (a powerful reasoning model now filling that gap) likely alleviates those doubts and demonstrates that MiniMax can keep up with – or even surpass – its peers technologically.

On the other hand, going public will put MiniMax under greater scrutiny. Profitability remains a big question for AI model companies, which incur heavy computing costs. As a recent commentary noted, the “burn rate” of large models is high and commercialization is still nascent, so the path to sustainable revenue is not yet proven for any of these startups . MiniMax will need to show how its product ecosystem (Talkie, Xingye, Hailuo, enterprise APIs, etc.) can generate steady income, and how open-sourcing models translates into business value. Regulators will also watch closely – Chinese authorities have supportive yet stringent policies for AI, including security reviews and licensing for models, which an IPO prospectus would need to detail as a risk. Additionally, listing in Hong Kong means MiniMax would be subject to international investor expectations regarding transparency and governance, a new environment for a young startup.

As of June 2025, MiniMax’s team has acknowledged the IPO consideration but emphasized it’s in “very preliminary preparation” . There is no fixed timeline or exchange finalized, and the valuation target could change with market conditions . For now, the IPO buzz underscores MiniMax’s stature: less than four years old, the company has vaulted into the upper echelon of AI firms poised to test the public markets. It will be a closely watched debut if and when it happens, potentially setting a benchmark for the value of AI model innovation coming out of China.

Product Ecosystem: From AI Companions to Multi-Modal Platforms

MiniMax’s product strategy is a dual engine: consumer-facing AI applications and B2B platforms/tools. This strategy not only helps monetize its AI, but also serves as a feedback loop to drive model improvement (a philosophy the founder calls “product-driven model evolution” ).

AI Companion and Social Apps: As discussed, Talkie (overseas) and Xingye (domestic) are core products providing AI companionship and content creation. These apps leverage MiniMax’s language and voice models to create interactive conversational agents. Users can customize virtual characters’ appearance and voice (leveraging MiniMax’s image and speech synthesis models), set their backstory and personality, and then chat with them or have them generate content. Talkie even added gamified features like collectible/tradable character “cards” to boost engagement . The popularity of these apps indicates a real user appetite for personalized AI “friends” and creators. They also serve as rich data sources for MiniMax to refine dialogue and alignment – every user conversation can help identify model weaknesses. Notably, Yan Junjie chose to focus on such chatbot and community products first (rather than a generic ChatGPT-style tool) because they are more forgiving of errors and better at spurring model iteration . As he explained, a chat app’s tolerance for occasional mistakes means a not-yet-perfect model can still provide value without user backlash, whereas a content creation community demands higher quality outputs to keep users interested . This approach allowed MiniMax to deploy models early, gather user feedback, and improve rapidly.

Hailuo AI (Conch AI) Creative Suite: Hailuo AI is MiniMax’s multi-modal content generation platform. Initially, “Hailuo AI” referred to a chatbot product, but in a March 2025 product line reorganization, MiniMax rebranded its main chatbot app as “MiniMax” (aligning it with the company name, focusing on text tasks) and reassigned the “Hailuo AI” name to a dedicated video generation product . In other words, Hailuo AI became the brand for MiniMax’s AI video creation tool, competing with the likes of Kuaishou’s Koli AI and ByteDance’s Ji Meng AI in China . This came after the success of the abab-video model; MiniMax realized a standalone video creation app had huge potential. Using Hailuo, users can input a piece of text (such as a script or idea) and get an AI-generated video complete with characters, motions, and even voice-overs. The results in many cases are comparable to professional animation or video editing, produced in a fraction of the time and cost. By early 2025, Hailuo AI’s overseas version drew global attention for its quality. It has been showcased at events like the China International Import Expo, where MiniMax demonstrated AI-generated videos and even an in-car AI agent built with a partner for smart vehicles . This hints at the wide applicability of Hailuo’s tech beyond just fun videos – from entertainment to advertising to automotive user interfaces.

MiniMax isn’t stopping at video. The March 2025 update also revealed plans for a new audio generation application (a separate app purely for AI-generated voice/music) to accompany its abab-speech models . By then, MiniMax had already released a state-of-the-art text-to-speech model (Speech-02) that topped certain leaderboards, beating even OpenAI in some TTS benchmarks . The forthcoming audio app will likely allow users to create lifelike voices or music on demand, tapping into use cases in podcasting, voice acting, audiobooks, and more. With text, video, and audio products, MiniMax would operate at least five core products (MiniMax Chat, MiniMax Agent, Talkie, Xingye, Hailuo AI, plus the new audio app) covering the gamut of AI content creation .

MiniMax Agent: Another exciting development is the MiniMax Agent, a general-purpose AI agent that the company began testing in mid-2025 . Integrated into the MiniMax app (the main chat platform), this Agent mode can perform complex tasks for users through multi-step reasoning and tool use. Early testers found the Agent could, for example, search information, generate and self-test code, create a website, or even produce a presentation (one demo had it generate an OpenAI history PPT) autonomously . Users simply give a high-level instruction and the Agent plans the steps, calls external tools or APIs as needed, and returns a result – an approach akin to emerging “AI assistant” platforms and similar to models like OpenAI’s Plug-in enabled GPT or projects like AutoGPT/Manus. MiniMax’s Agent is still in beta (accessible to invited users, especially overseas) , but it demonstrates the company’s intent to be at the forefront of the 2025 wave of AI Agents. Insiders say MiniMax views the convergence of long-context reasoning models (like M1) with agent architectures as a key opportunity to regain any ground lost to early movers like DeepSeek . With M1’s 80k-token output capacity and strong “reflection” abilities, it is well-suited to power an agent that can handle extended decision sequences and incorporate large knowledge context. MiniMax appears to be strategically leveraging its unique model strengths (like huge context windows and multi-modality) to carve out a niche in the agent space, potentially a high-growth area if AI copilots become mainstream for productivity.

Open Platform & Industry Collaborations: On the enterprise side, the MiniMax Open Platform (API) has made steady progress in integrating AI into traditional industries. One standout domain is smart hardware. MiniMax has been quietly providing large-model solutions to consumer electronics and IoT manufacturers. Industry reports indicate that most major Chinese smartphone makers – including OPPO, Xiaomi, and Honor – have been procuring MiniMax’s LLM as the AI assistant foundation for their devices . These phone brands take MiniMax’s base model, fine-tune it, and embed it as their on-device AI, rather than building completely from scratch . The only exceptions are said to be Huawei and Vivo, which have invested in more in-house model development, but even they collaborate with external AI firms in some ways . This means that when Chinese consumers use the voice assistant or AI features on many new phones, it’s often MiniMax’s technology under the hood enabling natural language understanding, conversations, and content generation. The scale of this is significant: China’s top smartphone vendors ship tens of millions of devices, so MiniMax’s model could be running on a huge number of endpoints, albeit usually in a cloud-assisted mode (some have 7B–70B parameter “on-device” models for offline use, supplemented by cloud for heavier tasks) .

In February 2025, MiniMax took this a step further by forming the “MiniMax Intelligent Hardware Innovation Alliance.” This coalition brings together various hardware and AI companies to explore AI integration in consumer devices, from smart home and wearables to automotive systems . Founding members included AR glasses startups, headphone and VR device makers, chip/component firms (like Allwinner Technology), and more alongside MiniMax . The alliance’s goal is to promote open collaboration and standards for embedding large models into hardware, addressing challenges like model compression, on-device inference, and user experience. At the launch forum, MiniMax noted that AI is transforming hardware, citing that China’s consumer AI hardware market had already reached ¥1.17 trillion (~$160B) in 2024 and is growing ~10% YoY . Use cases discussed ranged from education devices to smart car “AI co-drivers” . For instance, one partner demoed an in-car AI agent platform (built with MiniMax) that has been implemented by several domestic automakers in production vehicles . Another showcased AI hearing aids that use long-context memory to adapt to user preferences . These examples illustrate how MiniMax’s models are finding their way into diverse real-world products beyond just chat apps – a testament to its broad applicability.

MiniMax is also collaborating with media and education sectors. It partnered with The Paper (Pengpai News), a major digital newspaper, to integrate AI in journalism . Together they launched an “AI Spring Festival greetings” feature where users could generate custom New Year greeting videos using MiniMax’s video model, which attracted tens of millions of views and positive feedback . The company has engaged academia too, running AI video creation workshops at universities like Fudan and comm media labs, to both recruit talent and seed its tech in creative communities . Such outreach helps MiniMax build an ecosystem of users and developers who are familiar with its tools and can offer improvements or new content.

All these product and ecosystem efforts feed back into MiniMax’s core mission: “to co-create intelligence with users” (as the company slogan puts it ). By having end-user applications, MiniMax ensures its R&D is guided by real demand. By offering open APIs and forging partnerships, it gains distribution and specialized feedback (e.g., how well does the model handle medical queries or embedded scenarios). And by open-sourcing models like M1, it invites the world to improve and adapt its technology in ways a single company might not envision. This interplay between research, product, and community is intentionally cultivated by MiniMax’s leadership. As VP Liu Hua remarked at a developer forum, “Open-source co-survival is our ecosystem philosophy – we will explore tech innovation with global developers and partners, keeping an open attitude to push the boundaries of AI.” .

Competing in China’s Model Arena: A New Open-Source Tide?

In China’s burgeoning large-model arena, MiniMax is often grouped among the top startups challenging the established tech giants. Along with Zhipu AI, the Moonshot , and Jiyue Xingchen, it’s considered one of the “四小强” (four little strong players) in 2025, an evolution of the earlier “六小龙” cohort . Each has a different approach: for instance, Zhipu (spawned from Tsinghua University) built the GLM series and has close ties with academia; Baichuan Intelligence initially open-sourced smaller models and then focused on a medical-domain model; DeepSeek came seemingly out of nowhere with a cutting-edge reasoning model. Among these, MiniMax had gained a reputation for being technically solid yet somewhat conservative – it was often called the “most steady” of the group . That steadiness was seen in its methodical build-up of a full stack (text, image, video, etc.) and not rushing out features just for hype. However, when DeepSeek-R1 burst onto the scene in early 2025 as an open-source reasoning model with superior logic skills, MiniMax briefly appeared to lag in the “reasoning model” race . Tech pundits noted that Tencent, ByteDance, and others quickly integrated DeepSeek’s model into their products to give users advanced Q&A and problem-solving, whereas MiniMax refused to integrate a competitor’s model in its flagship apps (at least in China) . Instead, MiniMax used DeepSeek’s tech only in some overseas experiments, holding out for its own solution . This deliberate patience mirrored ByteDance’s strategy (ByteDance also avoided using DeepSeek in its main apps, choosing to wait for its self-developed model) . By mid-2025, that self-reliance paid off: MiniMax’s M1 provided the answer, delivering a “homegrown” reasoning model to rival or exceed DeepSeek-R1’s capabilities in key areas .

Now, with M1’s open-source release, many in the industry see MiniMax as leading a new wave of openness and competition. DeepSeek R1 was already a trailblazer – it proved that a smaller team could replicate some of OpenAI’s techniques (R1 was said to be inspired by OpenAI’s O1 model) and open-source them . MiniMax has upped the ante by open-sourcing an even larger and more novel model (hybrid MoE+Attention). This one-upmanship can spur Chinese AI labs to venture further into open source. We may witness a domino effect where, to keep up, others release their own models openly. Indeed, just a couple of weeks before M1, Alibaba’s DAMO Academy released Qwen-14B and other models with open checkpoints, and Baidu has hinted it might open smaller versions of its Ernie model. A competitive dynamic is forming: closed versus open, heavyweight tech firms versus agile startups.

MiniMax’s bet is that openness and cutting-edge performance are not mutually exclusive – you can be at the frontier and still share your work. The positive feedback from global AI circles on M1 supports this: the model’s HuggingFace repository and paper garnered praise from prominent researchers and even venture capital AI specialists abroad . By demonstrating that Chinese startups can innovate (M1’s 1M-token context is unmatched) and contribute to the global open research community, MiniMax is boosting China’s image in AI R&D. It also potentially attracts international talent and collaborators to engage with its projects, which could create a virtuous cycle of improvements.

In terms of raw model capability, MiniMax-M1 has firmly put the company in the conversation with top-tier models globally. While OpenAI’s GPT-4 remains closed and dominant in many benchmarks, M1’s long-memory specialization gives it an edge in scenarios GPT-4 cannot handle (GPT-4’s context length is reportedly 32k tokens, whereas M1 handles 1,000k) . For enterprises dealing with large documents or technical codebases, this is a huge selling point. In China’s domestic market, M1 appears to outshine most local closed models; MiniMax asserts its performance “is among the best of open models and even superior to some domestic closed models, approaching international leading levels.” . Tests show M1 beating competitors like DeepSeek-R1 in agent tool use, as noted, and it nearly matches GPT-4o-1120 (an open replica of GPT-4) on many tasks . Of course, AI capability is multi-faceted – for example, on math word problems or programming, DeepSeek’s latest distilled versions might still have an edge due to heavy fine-tuning, and giants like Baidu’s Wenxin (Ernie Bot) or Alibaba’s Tongyi Qianwen have the advantage of massive proprietary training data and integration into ubiquitous services (search, e-commerce). But the gap is closing rapidly. The playing field in China is now crowded with high-caliber models: Tencent, ByteDance, Alibaba, Baidu all launched their own “reasoning LLMs” between March and May 2025 . MiniMax managed to arrive fashionably late with M1 in June, and by open-sourcing it, effectively threw down the gauntlet.

Analysts consider MiniMax’s open model strategy as a way to “break out amid formidable rivals” . Surrounded by bigger companies (“big factories”) and fellow star startups, MiniMax is using openness and technical boldness as its weapons to stand out. There is a sense that 2025–2026 will be a make-or-break window for these AI startups – a “短暂窗口期” that is shrinking as the tech giants accelerate and as the market begins to demand real-world revenue . By fully leveraging its long-context and multi-modal expertise, MiniMax aims to carve a solid niche (e.g. being the go-to solution for any application requiring processing of very large texts, or powering agentic AI that can interact with complex environments). If successful, this could allow it to compete not by size alone, but by differentiation.

In summary, MiniMax’s recent flurry of activity – open-sourcing the M1 model, ramping up product lines, and exploring an IPO – reflects a company moving with urgency and confidence. It is attempting a delicate balance of being research-driven and product-centric, of cooperating (with open-source communities and industry partners) while competing fiercely (with big tech competitors). The outcome of this approach will have wider implications. A MiniMax IPO, for instance, would benchmark how global investors value an AI startup that prioritizes open R&D. Similarly, if M1’s open release leads to rapid improvements or spawning of new applications, it could validate open innovation as a faster route to AI advancement than closed propriety approaches.

For now, MiniMax stands as a leading light in China’s AI landscape – a unicorn with academia-grade tech ambitions and a startup’s nimbleness. As the CEO Yan Junjie remarked, “the AI industry is full of vitality and potential, but to stand on the world stage, we cannot rely on shortcuts” . MiniMax is clearly not shortcutting; it is playing the long game of pushing technology’s frontier (from million-token memory to multi-modal agents) and doing so in a way that invites the world to join in. In the fast-evolving AI arena, such an approach might just minimize the risks and maximize the chances of long-term success – true to the company’s very name.

Introduction: Wang Ning is the founder of PopMart, a Chinese art toy company that has grown from a small Beijing store into a cultural phenomenon and business empire. In a recent speech at a private forum, he reflected on PopMart’s 13-year journey and the philosophy behind its success . Wang shares how the company turned niche collectible toys into an enduring source of joy for adults , and why no real competitor has emerged to challenge PopMart’s dominance. He explains the value of “useless” art in a world of utility, the importance of patience and fundamentals in business, and his vision of taking PopMart global. Readers can expect to learn how PopMart blends art and commerce, taps into consumer emotions, and continually innovates to celebrate life in a changing world.

01. The Use of Uselessness: The Truly Eternal

I often face doubts like: “If you’re making these useless things, how high can the growth ceiling really be? People are rational—surely they’ll spend more money on necessities, not on these non-essential items.” Even up to now, many people don’t understand this industry; they can’t fathom why anyone would buy these art toys. The market is filled with debates between the “shorts” and the “longs” – skeptics who see it all as empty hype versus believers who see real potential.

In fact, all consumer behavior boils down to satisfying two things: a sense of fulfillment and a sense of existence. Fulfillment means a person’s material needs and basic spiritual needs are met. Existence means showing others who you are – whether you’re wealthy, have taste, or possess an artistic flair. Fulfillment or existence – which of the two is a necessity, and which is not?

Let me give an example. Imagine that overnight we all became super-rich and lived in a grand estate, and today you’re home alone. Picture the following two scenarios:

• Scenario 1: You drive out and suddenly recall you forgot to turn off a faucet. Left running all day, the faucet would waste about 40 yuan worth of water. Would you feel anxious about that?

• Scenario 2: While you’re away, the huge fountain at your estate’s entrance keeps running. One day of water + electricity for the fountain costs 400 yuan. Would you feel anxious about that?

I’ve posed these scenarios to many people. Most answer that in the first scenario – the open faucet – they would feel anxious, but in the second scenario – the constantly running fountain – they would not feel particularly bothered.

Isn’t that interesting? It’s all just water, flowing away, so why do we worry about the running faucet but not the running fountain? I’ve analyzed the reason behind this.

We assume our spending habits are highly rational – that we’d fret over each drop of water wasted. But in reality, we’re far more emotional than we imagine. The fountain actually uses more water and costs more money, yet nobody cares about it. So when we make consumer products, the question becomes: Do we lead people to the “faucet,” or to the “fountain”?

Consider this: if Molly’s head could be pulled off and used as a USB flash drive, would you still buy so many of them? Of course you wouldn’t. Because every time you bought a Molly USB drive, you’d second-guess yourself: “Why did I buy yet another flash drive? I already have several; I can’t even use them all!” If the cute figurine had a utilitarian function, it would prompt that rational scrutiny each time.

So many people think to add some practical function to a product – but they haven’t figured out whether they are leading people to the faucet or the fountain. It’s just like with music. Many years ago, when people had only just solved basic food and clothing needs, a lot of folks couldn’t understand why anyone would pay for albums to listen to music. It was hard for them to imagine that music would later become a whole industry.

Similarly, people often say that pure visual art seems to have been in decline in recent decades. I think that’s because art no longer exists much in a purely “art for art’s sake” form. Pure art is like when we go to a museum to admire human civilization – ancient murals, sculptures, and so on. But nowadays art has already integrated into our lives – printed on our clothes, on our phone cases.

I once asked some luxury brand executives why they love collaborating with artists. They understand it very clearly: art is eternal, and “useless” things are truly eternal. Any product that has a functional attribute automatically comes with a short life cycle and an inherent decay. No matter if today you buy the most advanced phone, computer, or car on the market – by tomorrow it’ll already start depreciating. This is unavoidable. Only those “useless” things can endure across time. That is the charm of art, and it’s the charm of our industry .

02. Industry Barriers: PopMart as the “Record Label” of Art Toys

From the day PopMart opened its doors, we’ve been doing business in plain sight. We ask: What are we selling? How do we sell it? What products are the hottest?… All that information is transparent – we play with our cards face-up. And yet 13 years on, we still haven’t encountered a competitor of similar scale. Why is that?

It’s true that “useless” things have enduring appeal – but getting people to actually spend money on a purely non-functional product is extremely difficult. The barriers to this industry are much higher than outsiders imagine.

People often ask me, “What is PopMart’s moat? What’s your barrier to competition?” I believe we have two major barriers – one soft barrier and one hard barrier.

1. Soft Barrier: Artists are a scarce resource that money cannot buy. PopMart is actually a lot like the “record company” of our industry. The artists are a bit like those virtuoso pianists or violinists from centuries ago.

Before record companies existed, those artists could only perform for a few people by selling expensive tickets to lavish ballroom shows. But after the invention of records, music could be recorded and sold all over the world, achieving true scale and commercialization.

Our role is akin to finding those brilliant “pianists” and “violinists,” then helping them achieve full commercial success. We were very fortunate to have discovered such talented artists early on.

The 1980s–90s were a golden era for the cultural industry in Hong Kong. Back then, a group of designers wanted to do pure art, but they found that fewer and fewer people were buying paintings or sculptures. So they pivoted, exploring whether art could be made into something like prints – smaller pieces, maybe editions of 50 or 100, to sell at toy conventions. Eventually they discovered that soft vinyl material (also known as “tangjiao” in Chinese) could capture perfect line work and vibrant colors simultaneously – like a combination of painting and sculpture – so many artists gradually got involved in making figures. (Note: “tangjiao” is a process of coating the inside of a mold with molten PVC, achieved by slowly rotating the mold in a furnace, so that the plastic forms an even layer.)

Art, by its nature, strives for uniqueness – every artist has their own “school” or style. By creating characters, artists can quickly establish a distinctive identity, and thus IP based on art toys was born.

In the early days, our method for scouting IP was pretty simple: we’d attend toy shows around the world. These art toys are often released in limited runs, and many enthusiasts would line up overnight to buy them. So our standard for choosing artists was literally to pick the ones with the longest queues! If people are lining up in droves, we want to work with that artist.

Over time, you realize that truly outstanding artists are very scarce resources. They’ve often spent many years honing their art, and they have strong recognition within those circles. As our platform grew, we also started attracting more up-and-coming artists with potential. Just like how the biggest, best record label draws in new talent – a kind of siphoning effect – it creates a positive cycle.

Now that our company’s scale is much larger, we can even begin leading tastes and trends. It’s a bit like the film industry: initially you need a famous star to get box office hits, but once you become a top-tier studio, it can flip so that anyone cast in your film becomes famous. As the platform grows, we can use our resources to push an IP into the spotlight and make it well-known.

So, artist talent is a soft barrier. It’s not something you can overcome just by fighting on speed, money, management, or market share. It’s a moat that can’t be crossed simply with brute force or copycats.

2. Hard Barrier: In the end, doing business comes back to the basics. I always emphasize eight characters: “尊重时间，尊重经营” – respect time, respect running the business. Over the years, I’ve seen many companies rise and fall. It usually comes down to two big mistakes: either not respecting time, or not respecting business operations. These are the two easiest errors to make.

To respect time means understanding that anything which truly takes ten years to accomplish cannot be rushed in two or three years.

To respect operations means that no matter how great your business model is, no matter how flashy the concept, in the end it all comes down to the nitty-gritty of rice, oil, salt, soy sauce, vinegar, and tea – the everyday details. It comes down to how you deal with people, tasks, money, and the continual troubleshooting of complex issues.

I often say we’re like that cyborg girl in the film Alita: Battle Angel – on the surface she looks like a beautiful young woman, but beneath the skin her core is made of precise mechanical parts. Our business involves a lot of fields: supply chain management, team management, strategy, storefront operations, merchandise management, and so on. For a single store, we think through extremely granular details: how should the products be displayed? How should the customer service be? We even think about something as small as who should open the store in the morning and who should close at night, to ensure the shop runs safely and smoothly.

Retail is all about details. It’s inventory, it’s cash flow cycles, it’s supply chain, it’s shelf placement – it’s all the boring, trivial bits and pieces. Running a business ultimately returns to that long-term accumulation of basic, everyday work. In my view, this is a hard barrier. If something truly takes ten years to accomplish, we’ve spent ten years on it – and any would-be competitor will likely need to spend about the same amount of time as well.

03. Latest Insight: “Happiness” Might Be a Bigger Market

In the early years, we faced a lot of misunderstanding. People were completely unfamiliar with this market, so they didn’t believe adults would still buy toys – like how a bald man wouldn’t buy shampoo. Some people might also wonder, why are we called “POPMART” (literally “Popular Mart” or “Trendy Supermarket”)? It’s because at the very beginning, we were a retail channel brand – PopMart’s original idea was to gather all the products young people liked and put them together. When we first started, we sold all sorts of things: home goods, electronics, stationery, cosmetics, etc.

That actually ties back to my entrepreneurial experience in college. In my junior year, I opened a “grid store” – I divided a physical store into many small grid-like compartments which I then rented out to different people. Those people would use their rented grid to sell various small goods on consignment. It was like an offline Taobao shop. At that time, the grid store business was extremely popular. Because it offered a wide variety of items and a fun browsing experience, it was a big hit among young people.

Looking back though, the grid store model had a lot of problems behind it – things like the management model, operational efficiency, etc. But it planted an entrepreneurial seed in my heart. I wanted to keep iterating on a model like the grid store.

Around 2010, I traveled to many places for research. I saw a retail chain brand in Hong Kong called Log On, which was very similar to the business model I had envisioned – it was also an aggregation of all the categories young people like. That inspired me when creating PopMart.

However, once we got PopMart up and running, we discovered we’d fall into a kind of “vicious cycle”: If we wanted better revenue numbers, we needed more SKUs (Stock Keeping Units – unique product styles and items), which meant bigger investments, bigger stores, more staff, more complicated inventory, more complex management and training…which in turn meant efficiency would drop lower and lower, and it became very hard to form a clear consumer impression of what our brand was.

So we decided to simplify – we raised a slogan at the time: “Reduce breadth, increase depth.” That meant cutting some product categories and going deeper into others. By 2015, we discovered that out of all our categories, the one growing fastest was art toys (潮玩). So we made a very important decision—to axe everything else and focus solely on art toys .

In the earlier days, many people compared us to Lego. But now, we’re more akin to Disney . What’s the thinking and logic behind that shift?

We did in fact borrow a lot of ideas from Lego initially. I believe Lego’s greatest brilliance is that it created its own language and system, almost like a tech company would invent a new coding language or platform. (Note: The “Lego play system” was a strategy proposed by Lego’s leader Godtfred at a 1954 toy fair, when he heard buyers complain that toy makers shouldn’t just make one-off products that briefly grab the market, but rather develop an interconnected system of toys to drive repeat sales. Inspired by this, Lego’s strategy rose from “making individual toys” to “creating an entire play system,” i.e. every Lego model kit can connect and interact with pieces from other kits – trains, cars, characters, etc.)

The result is, if you want to collaborate with Lego, you have to adapt everything into Lego’s language system. That proprietary language and ecosystem carries enormous business value – far beyond the toy products themselves. This is Lego’s core competitive asset built over so many years, the most valuable thing it has.

So, from early on PopMart also set out to build our own language and system. Following that idea, we began to define the term “art toy” (潮玩) itself in our context, and to define our way of doing things. For example, we insisted that art toys should have as many standard features as possible – standard sizes, materials, “hidden edition” ratios, etc. Over time this spawned a whole set of industry lingo, like “blind box,” “shake the box,” “chase (hidden) figure,” and so on.

In just five short years, take the term “blind box” for example – it went from an obscure phrase, to a trendy buzzword, to an overhyped cliché, and now it’s basically returned to a neutral term again.

As the company grew, we found that in some ways we diverged from Lego’s path. Beyond creating a language, we also built up a roster of intellectual properties (IP). And through our “language” system, we enabled everyone to recognize these IP characters. Now, prior to this, lots of people had tried for years to create a nationally beloved super IP in China, and the success rate was very low. But our model really has managed to incubate a batch of super IPs.

After that, we started to develop an entire art toy ecosystem centered around IP. In recent years, we’ve tried enriching the products and business lines around our IP characters – including games, films, spin-off merchandise, even theme parks.

Previously, our company was more heavily a “Lego model,” but now the “Disney model” portion is starting to grow. I’ve noticed many people are taking interest in our theme park business. That segment has only just begun – we’ve built a very small theme park as an experiment in this industry.

A lot of people also ask me, “Why are you getting into games? Why build a theme park? Are you trying to learn from Disney?”

That brings me to a gradual shift in our own thinking.

When we were still very niche, focused on making trendy collectibles, what we were doing was actually quite limited in scope. We were doing something about being trendy – it was a very small circle.

And what does it mean to do something about “trendiness”? It’s basically figuring out how to appear trendy, how to make young people think we’re cool.

At that time, our understanding of “trend” was: Trend is a form of superiority – I have what you don’t, I know what you don’t. It’s that sense of being ahead or above.

Back then, we spent a lot of effort cultivating that feeling of superiority – for example, culturally positioning ourselves as avant-garde, or using limited editions of our figures to create exclusivity. We wanted everyone who bought PopMart products to feel like they were so trendy and cool.

Later on, we realized that to truly expand the market – to “grow the pie” – we had to break out of that little trend circle. So we expanded from the trend circle to the fashion circle. (In Chinese, “trend” and “fashion” can both translate to fashion, but you’ll find Supreme and Zara are fundamentally different things.) We essentially went from being like Supreme and started moving towards Zara, to reach a much larger scale .

However, as we kept moving forward, I discovered that actually we can go even bigger – tap into an even larger circle, a bigger market, and fulfill an even more universal need. And what need is that? It’s the circle of “happiness.”

I feel that “happiness” might be a far bigger market. So around this broader market, we’ve been creating many products and experiences that center on companionship and joy. For example, our theme parks, and a lot of our licensing deals, are all related to bringing people companionship and happiness.

At present, aside from Disney, we are the largest domestic company in China when it comes to exporting IP through licensing – meaning we license out our characters for all kinds of products and partnerships. (This is another way our business mirrors Disney’s).

04. Looking Ahead: Our Business Has Only Just Begun

Speaking of the future, our biggest strategy is globalization and diversification. By “group diversification,” I mean developing a richer array of businesses centered around our IP; by “globalization,” I mean increasing the proportion of our revenue that comes from overseas markets.

In one respect, we have been quite lucky: we started preparing to “go overseas” in 2018, making us one of the earlier companies in China to plan for international expansion . In truth, a lot of preparation is needed for going abroad: from meeting various product specifications and standards, to adapting content to local cultures and doing translations, and then to building an entire overseas team.

In the past few years, our overseas markets have been growing over 100% annually – this year we expect triple-digit percentage growth again. I think we’re probably one of the Chinese consumer product brands – especially as a culture-based company – that’s performing the best overseas.

These last two years, no matter if you go to Southeast Asia, Japan, Korea, Singapore, or to the UK, France, the US… you can find our stores. We currently have 50+ overseas stores, and by the end of this year it should be close to 100 stores .

Our approach to overseas markets is different from some brands that just set up a token presence abroad to “look good” and then use it mainly for marketing back home (the old “export to domestic sales” trick). All of our overseas operations are driven by the goal of making money and turning a profit. In fact, our overseas business is already nearly 20% net profit margin now.

Another thing that sets us apart from many other brands is our insistence on directly-operated stores. A lot of companies, to expand quickly, will choose franchising. But we run all our stores directly.

At the moment, our overseas revenue is still not a huge portion of the whole, but by our estimates we’ll achieve at least 1 billion RMB in overseas revenue this year, which would be a bit over 10% of the company’s total revenue. What’s more important is that the overseas segment is growing very fast. So I believe it won’t take many years for our overseas income to reach 50% of the total. The U.S. market in particular is performing exceptionally well – we’re likely to invest even more heavily this year and next year into the U.S. market’s growth.

That’s our expectation for the future as a whole.

Earlier this year, I was strolling down a street in the UK when I was suddenly hit by a feeling: our business is only just beginning. I saw these streets teeming with people, saw all those globally famous brands. It struck me how much more we still could do overseas. There is so much potential out there in the wider world waiting for us to explore.

05. Bonus: Every Company Must Find a Way to Sustainably Invest in IP

A lot of people question: If an IP doesn’t tell a story, how long can its vitality really last?

Let me first ask everyone: What important IPs live in the minds of our generation? How were they formed?

For example, Journey to the West, My Fair Princess, Legend of the White Snake – these all became major IPs. Maybe it’s because they had great stories and characters, but we believe a crucial factor was the element of time – time spent accompanying the audience.

When we were kids, every summer vacation the TV stations would start broadcasting those shows, and our generation watched them over and over. That long period of repeated companionship imprinted them in our minds as IPs – even super IPs for a whole generation.

Now the biggest problem is that time has become fragmented – young people just don’t have the time anymore. If you open a video streaming app now, you’ll find most of the TV series listed you’ve never watched (and don’t plan to), or you just watch a three-minute plot summary. So using traditional television shows to nurture an IP has become harder and harder – the show might not even get the chance to catch fire before the next one is already up in the queue.

Of course, we don’t deny that telling a good story can still create an IP. It’s just that the ROI and efficiency of that approach is getting lower and lower. We want to find a more efficient way to establish IP recognition and realize its commercial value.

We’re not stubbornly insisting on which path is right or wrong. There’s no need to get hung up on the old playbooks – there may be many different paths to reach the same destination.

Ultimately, the key to an IP’s lasting vitality is continuous, healthy investment into it. Think about why Star Wars is such a phenomenon. They’ve made like a dozen films in the franchise, and they keep using commercial means to continuously invest in, reawaken, and iterate on that IP. Mickey Mouse is the same way – it stays relevant because it’s constantly reintroduced and refreshed. Conversely, if you ever stop investing in an IP, it can quickly be forgotten.

So every company needs to figure out a method for sustained and healthy investment in their IP – to keep it alive and thriving.

Lastly, I want to say a couple of words about a term everyone’s been talking about lately: “generation gap” – in other words, differences between age groups. (Note: “generation gap” here refers to the disparities, disconnects, or even conflicts in values and choices between different generations.) People often discuss the generation gap in terms of technology. But I’ve found that brands also experience a kind of generation gap.

For instance, many of us today understand consumer marketing in a certain way. Take the oft-mentioned idea of “consumer upgrades” – it’s about people gradually realizing that besides working, they should also enjoy life. In other words, “live in the moment, enjoy life” – forget the bad things, go ahead and eat and drink and be merry. Under this mindset, the highest phrase we could think of was “enjoy your life,” and we built our products and advertising around that idea 【celebrate your life】.

One day, I saw an advertisement by a British brand. It advocated “celebrate your life.” That really struck me! I felt that the word “celebrate” is just brilliant. Celebrate entails more occasions than enjoy; its emotional intensity is higher, too.

This exemplifies what I mean by the generation gap. I realized that different people – from different generations or cultural backgrounds – have vastly different understandings of branding, of consumption, even of civilization. This insight pushed me to iterate our own marketing strategy.

So here, I hope that all of us can not only enjoy your life, but truly celebrate your life. Let’s not just enjoy living, but celebrate living to the fullest every day.

In early summer 2025, a throng of desperate employees crowded into Neta Auto’s sleek new Shanghai office and cornered founder and CEO Fang Yunzhou, demanding months of unpaid wages . It was a shocking scene – not long ago, Neta Auto had been hailed as a “dark horse” in China’s electric vehicle (EV) market, a little-known startup that briefly outpaced more famous rivals in sales through relentless ambition and savvy marketing. Now its headquarters were under siege by its own staff. The dramatic rise and precipitous fall of Neta Auto has captivated China’s tech and business circles, offering a cautionary tale of how meteoric success can swiftly unravel in the country’s cutthroat EV industry.

Founded in 2014 by Fang Yunzhou, Neta Auto rode China’s EV boom from obscurity to industry prominence in just a few years. This account recounts how Nezha rose from an underdog to a market leader with key funding and a low-cost strategy, then charts the company’s recent implosion – from employee protests and unpaid wages to factory shutdowns and bankruptcy rumors – on through an analysis of the multi-faceted causes behind its downfall. In the end, Nezha’s saga speaks volumes about the brutal “survival of the fittest” underway in China’s EV arena, where breakneck growth has given way to harsh consolidation.

An Underdog’s Meteoric Rise

Neta Auto’s ascent began with a focus on affordable electric cars for the masses. Its very first model, the Neta N01 launched in 2018, was a compact EV priced around ¥60,000–¥80,000 (~$9–12k) – far cheaper than most competitors . In 2020, Neta followed up with the Neta V, a small crossover starting below ¥60,000, which quickly gained traction in China’s lower-tier cities . This ultra-low pricing strategy – encapsulated in Neta’s slogan “building cars for the people” – helped the startup tap into a huge underserved market of budget-conscious buyers. By offering functional EVs at roughly half the price of models from Tesla or NIO, Neta rapidly expanded its sales volume.

Crucially, investors took notice of Neta’s early promise. The company (whose legal name is Hozon New Energy Automobile) raised funding from a roster of powerful backers, including local government funds and China’s battery giant CATL . A turning point came in 2021 when tech mogul Zhou Hongyi – founder of Internet security firm Qihoo 360 – led Neta’s Series D financing round with a ¥2 billion investment. That round raised a total of ¥4 billion, valuing Neta at around ¥25 billion (≈$3.8 billion) . Zhou’s involvement brought not just cash but also publicity: the outspoken billionaire declared he would use “internet thinking to disrupt traditional carmakers,” giving Neta a viral marketing boost . By the early 2020s, Neta had secured an estimated ¥22.8 billion (over $3.3 billion) across ten funding rounds – an extraordinary war chest that it poured into scaling up production, building retail stores, and R&D.

All that investment and expansion paid off – at least initially. By 2022, Neta Auto was on top of the world (or so it seemed). The company sold about 152,000 vehicles that year, more than any other Chinese EV startup, even surpassing higher-profile players like NIO, Xpeng, and Li Auto . This achievement earned Neta the title of annual sales champion among China’s “new forces” in automaking, cementing its status as the industry’s unexpected dark horse . Neta’s budget EVs – particularly the Neta V and Neta U models – were moving in huge volumes, helping the company double its sales in a short span. Fang Yunzhou’s vision of “EVs for everyone” seemed to be working, and the firm eyed an IPO to fuel further growth. Yet beneath the celebratory headlines, the seeds of trouble were already planted: those tens of thousands of cheap EVs were being sold at razor-thin or even negative profit margins, a fact that would eventually catch up with Neta.

Read more

Major Chinese car manufacturers – from EV leader BYD to legacy state-owned giants – have publicly pledged to pay their parts suppliers within 60 days of delivery . This industry-wide initiative comes on the heels of an intense price war in China’s auto market and mounting pressure from both regulators and suppliers. In early June 2025, at least 14 to 17 automakers including BYD, Chery, Geely, Great Wall Motor, SAIC, FAW, Dongfeng, GAC, Changan, as well as newer EV firms like Xpeng, Nio, Li Auto and others, announced plans to drastically shorten their supplier payment cycles to two months or less . Automakers say the move is aimed at fulfilling corporate social responsibility and ensuring the “high-quality development” of China’s automotive supply chain .

This wave of pledges was largely catalyzed by external pressures. Chinese authorities had issued new rules in March requiring big companies to settle most payments to suppliers within 60 days, effective June 1, 2025 . Suppliers, especially in upstream industries like steel, had also voiced outrage over being squeezed by long overdue payments. In fact, China’s Iron and Steel Association took the unusual step of publicly complaining that some carmakers demanded steel price cuts of over 10% while delaying payments by months, leaving steel suppliers with “little profit margin and mounting liquidity pressure” . Regulators responded: on May 31, the Ministry of Industry and IT warned that authorities would punish automakers engaging in unreasonable price cuts to grab market share . Clearly, Beijing signaled that brutal price wars and off-balance-sheet supplier debt would no longer be tolerated, setting the stage for this industry payment reform.

BYD’s Leadership and Motivation

BYD, China’s largest electric vehicle maker, has emerged as a prominent champion of the 60-day payment initiative. It was among the first to announce compliance – releasing its statement just after 1:00 a.m. on June 11 – vowing to unify its supplier payment terms to within 60 days . BYD’s proactive stance reflects both external expectations and its own strategic interests. The company has been in the spotlight recently after a rival executive ominously warned that an “Evergrande of the auto industry” could be looming – a veiled reference that many interpreted as pointing at BYD’s rapid expansion and heavy liabilities . BYD’s management fiercely refuted those allegations, publishing data to show its finances are sound. For instance, BYD’s debt ratio (~70%) and accounts payable days (around 127 days) were argued to be comparable to or better than peers like Geely (127 days) and significantly lower than some rivals such as Great Wall Motor (163 days) or SAIC (164 days) . In 2024, BYD boasted over RMB 777 billion in revenue and RMB 154.9 billion in cash reserves – implying it has the liquidity to support faster supplier payments.

By embracing the 60-day payment rule, BYD demonstrates the “leading enterprise” responsibility expected of an industry front-runner . Motivations abound: First, it aligns BYD with the new regulatory requirements, avoiding any perception of non-compliance. Second, it helps defuse criticism that BYD’s success comes at the expense of suppliers – a narrative the company is keen to dispel. BYD’s brand chief stated that claims of BYD being an Evergrande-like ticking time bomb are “frustrating and amusing”, underscoring that no Chinese mainstream automaker is an ‘Evergrande’ and that undermining China’s EV industry with such rumors is unacceptable . The 60-day pledge allows BYD to underscore its financial health and commitment to fair partnerships. Third, supporting suppliers now is in BYD’s long-term interest: as BYD’s production and exports surge, maintaining a stable, innovative supply chain is critical for sustained growth. In short, BYD’s early adoption of the initiative is both a public relations message – signaling goodwill and stability – and a strategic move to fortify its supply chain for the future.

Industry-Wide Response: From Chery to Great Wall

Crucially, BYD is not alone – the 60-day payment commitment has quickly become a united front across China’s automotive sector. In an unprecedented show of solidarity, virtually all major Chinese automakers made coordinated announcements over June 10–11, 2025, committing to the new supplier payment standard . This group includes state-owned giants (FAW Group, Dongfeng Motor, GAC Group, SAIC Motor), leading private carmakers (Geely Auto, Great Wall Motor, Chery Automobile), and EV startups (Xpeng, Nio, Li Auto, Leapmotor, Xiaomi Auto) among others . For example, GAC was the first to confirm the change on the evening of June 10, followed in quick succession by FAW, Dongfeng, Seres (Chongqing Sokon), Geely, Changan Automobile, and others, with BYD and Xpeng issuing statements in the early hours of June 11 . By June 11, at least 17 automakers had joined this “collective action”, standardizing their supplier payment terms to 60 days or less .

Notably, this concerted action was not entirely voluntary – it was strongly encouraged by industry bodies and regulators. China’s Association of Automobile Manufacturers (CAAM) had issued an industry initiative calling on OEMs to end “involution” (vicious, self-defeating competition) and to “stop occupying upstream funds to sustain development”, following complaints about egregiously long payment periods . Automakers explicitly framed their pledges as a response to government and CAAM guidance: for instance, BYD’s statement cited the need to “implement the directives of the state authorities to stabilize the supply chain and promote high-quality development”, and Changan and Geely invoked their duty as central state-owned enterprises to set an example . In effect, the 60-day payment rule has rapidly become an industry norm. As one analyst noted, with regulatory enforcement kicking in, this is “no longer optional, but a new baseline standard for doing business in the automotive sector” . Any automaker that drags its feet risks both government scrutiny and losing supplier trust or even access to critical supplier resources if they don’t match the new standard .

It’s worth noting that a few carmakers had not yet publicly committed at the time of the initial announcements , possibly smaller local manufacturers or those assessing the financial impact. But given the powerful bandwagon effect and the explicit legal mandate now in force, it is expected that holdouts will have little choice but to follow suit. The Regulation on Ensuring Timely Payments to Small and Medium-Sized Enterprises, which took effect June 1, 2025, legally requires all large enterprises (and government entities) to pay SMEs within 30 days of delivery by default, or within 60 days if a longer period is contractually agreed . Crucially, the regulation also bans the practice of forcing suppliers to accept promissory notes or other non-cash payment instruments that delay actual cash receipts – a loophole automakers frequently used in the past to stretch payment times. In line with this, several automakers like BAIC and SAIC specifically vowed not to use commercial paper for supplier payments going forward . The fact that leading OEMs are applying the 60-day rule not just to small suppliers but to all suppliers, regardless of size, as a voluntary higher standard , indicates the industry’s determination (or obligation) to reset its payment practices across the board.

Lifeline for Suppliers: Cash Flow Relief and Survival

For automotive suppliers – especially the thousands of small and medium-sized parts makers in China – this change couldn’t come soon enough. Extended payment terms have long been a pain point in the industry, effectively using suppliers as a “bank” for assemblers. Chinese automakers’ payables were notoriously slow: on average it took about 182 days for Chinese car companies to pay their suppliers (as of 2024), roughly double the payment cycle of international automakers (around 90 days) . In some extreme cases, payment cycles stretched beyond 240 days – meaning certain suppliers waited 8 months or more to get paid, essentially financing the automaker’s operations for up to a year . This practice has pushed many suppliers to the brink. It eroded suppliers’ profit margins (which are often single-digit in the auto parts business) and forced them to shoulder extra costs – for example, by borrowing money to cover their own expenses while awaiting payment, or by selling invoices to banks at a discount to get cash. According to the China Iron & Steel Association, some car OEMs were delaying payments and settling with corporate promissory notes instead of cash, shifting financing costs onto upstream suppliers and “significantly increasing financial pressure” on those suppliers . Suppliers often had no choice but to accept these delayed-payment notes, which they might only redeem or factor months later at a loss, further chipping away at their already thin profits .

The new 60-day payment commitment promises to radically improve this situation. Slashing average payment times from ~6 months to 2 months means suppliers will see cash much sooner – a nearly two-thirds reduction in their accounts receivable cycle . Analysts estimate that if fully implemented, this could unlock “hundreds of billions of yuan” in liquidity across China’s auto supply chain . That freed-up cash will relieve suppliers’ debt burdens and interest costs, potentially saving many smaller firms from insolvency. Official data shows that as of April 2025, large industrial companies in China were waiting longer for payments (70.3 days on average, 4 days more than a year prior) – a sign of worsening payment delays . In the auto sector it was even worse: one report noted 16 listed Chinese automakers took an average of 182 days to pay suppliers in the first 9 months of 2024, a full month longer than the prior year’s average, and roughly twice as long as global peers . The 60-day rule directly targets this chronic problem. Industry insiders told Yicai Global that although shortening payment terms will put some strain on carmakers’ cash flow, it is a necessary correction to an unhealthy trend . It will “firmly uphold a fair and orderly market environment” and support the high-quality development of the auto industry, they said .

From the perspective of a small supplier, the benefits are concrete and immediate. “For SME suppliers, having a clear 60-day payment deadline means far more certainty of cash recovery,” noted a Sina Finance commentary, “which alleviates cash flow pressure and lowers financing costs. It gives them the confidence and ability to invest in technology R&D and expand capacity, thereby boosting innovation across the supply chain.” . In other words, instead of worrying about survival, suppliers can plan for growth. One automotive parts maker told First Financial that if receivables truly convert to cash within 60 days of delivering goods, it will “greatly improve our cash flow”. In the past, even when nominal payment terms were 60 days, big automakers often paid with acceptance drafts that could only be cashed much later – for example, getting paid 8 months after delivery was common, unless the supplier was willing to accept about a 6% annualized discount to cash it in early . “Now, getting cash in hand in two months is certainly much better,” the supplier said, “since our profit margins aren’t high to begin with – the key is we won’t have to give up so much of our profit in exchange for prompt payment.” Many supplier firms are also hopeful that faster payment will allow them to reduce bank loans or invoice financing usage, cutting interest expenses. In capital-intensive sectors like auto components, cash is indeed king – and the difference between 2 months and 6+ months in payment time can be the difference between a healthy supplier and a bankrupt one. It is telling that right after these pledges, Chinese auto parts stocks jumped sharply, with investors betting that supplier balance sheets and survivability will improve** **.

Strengthening the Supply Chain: Efficiency, Transparency and Credit

Beyond immediate cash relief, industry experts say the 60-day payment reform could transform the dynamics of China’s automotive supply chain. Shorter payment cycles encourage a more balanced and collaborative relationship between automakers and suppliers, rather than the previous zero-sum game. As the South China Morning Post noted, when carmakers delay payments to fund price cuts, they are effectively creating a hidden, interest-free debt at suppliers’ expense . This not only hurts suppliers, it also undermines trust and long-term innovation – suppliers starved of cash are less able to invest in new technologies or maintain quality. A former executive vice-president of CAAM, Dong Yang, remarked that protracted payment delays had become a “typical tool of involution-style competition” in the auto industry, one that weakened the resilience and innovation capacity of the entire supply chain . Now, with timely payments, the supply chain’s collective health improves. Upstream firms will have more working capital to upgrade equipment, develop better components, and ensure product quality – benefiting automakers with more reliable and advanced supplies in the long run.

Moreover, several automakers have signaled they will use this opportunity to increase financial transparency and trust with suppliers. State-owned FAW Group and GAC, for example, said they will leverage digital tools to make the payment process more transparent, ensuring every invoice is paid on time and establishing a stronger trust foundation between OEMs and parts makers . This kind of digitized payment platform can reduce disputes over payment status and build a credit history for timely-paying manufacturers. Other companies, like BYD, have spoken about a “dual optimization of technology and management” alongside the 60-day payment rollout, aiming to move the supply chain relationship from a zero-sum bargaining mindset to one of collaborative innovation . In practical terms, this could mean closer coordination on inventory management, joint cost reduction efforts, and co-development of new solutions – since suppliers who are paid promptly may be more willing to share innovations exclusively with cooperative automakers.

Importantly, the new payment discipline is expected to reshape competitive behavior. When automakers can no longer lean on overdue payables as a cushion, they must manage cash flow more rigorously and compete on real efficiency rather than on unsustainable discounts. “Shortening payment times will impact automakers’ cash flow, limiting their ability to wage price wars,” noted Dong Yang, implying that firms will have to curtail the practice of using supplier credit to bankroll destructive price cuts . Executives echo this view: Changan Auto’s chairman Zhu Huarong recently urged the industry to “resolutely oppose immoral price wars”, and Seres CEO Zhang Xinghai warned that “low-price competition endangers quality standards – safety is the greatest luxury” . With a mandated 60-day pay cycle in place, carmakers can no longer prop up aggressive discounting campaigns by stringing along their suppliers. This should naturally dial down the frenzy of the price war that has raged since early 2023 . As one Chinese auto consultant observed, this concerted action by carmakers is a first step by authorities to regulate the market, and “automotive companies will eventually refrain from offering steep discounts as government agencies tighten oversight of their sales strategies.” The focus of competition, he suggests, will shift from purely price to areas like technology, quality, and brand – a much healthier scenario for the industry’s sustainable development.

Another anticipated outcome is the bolstering of a credit culture and payment credibility in the auto sector. In the past, many Chinese suppliers had to worry about their customers’ creditworthiness and might factor in long delays or default risk into their pricing. If the 60-day term becomes strictly upheld, suppliers can have greater confidence in getting paid promptly, effectively improving the credit profile of the entire industry. Some automakers have even framed the new practice as building a “community of shared future” with suppliers and shifting competition toward credit and management capability . In other words, those OEMs that manage their finances well and honor agreements will gain a reputational edge. Over time, one can envision an ecosystem where superior payment reputation becomes a competitive advantage in attracting the best suppliers. Manufacturers might begin to compete for supplier loyalty not by squeezing payment terms, but by offering better collaboration, which ultimately improves vehicle quality and innovation. This is analogous to the Japanese auto industry model that Chinese industry groups have pointed to – Japanese automakers are known for long-term, stable partnerships with suppliers, leaving them reasonable profit and fostering joint innovation . By committing to fair payment cycles, Chinese carmakers are taking a page from that playbook, which could raise the overall technical level and reliability of China’s automotive supply chain.

Outlook: Toward a New Standard and a Healthier Competition

Read more

On June 11, 2025, Chinese camera maker Insta360 made a splashy debut on Shanghai’s STAR Market, with its stock surging 285% on opening day and its market capitalization briefly topping ¥70 billion (over $10 billion) . Founder Liu Jingkang – known as JK Liu – proudly rang the opening bell while wielding an Insta360 X5 panoramic camera, marking him as the first 1990s-born founder to lead a company to China’s Nasdaq-style board . The blockbuster IPO is the culmination of a decade-long journey of entrepreneurial grit and continuous innovation. This narrative retraces Insta360’s path from a college project to a global leader in 360° cameras and action cams, its hard-fought road to an IPO, and what lies ahead for the Shenzhen-based pioneer.

Founding a 360° Vision: JK Liu’s Origin Story

Insta360’s story begins with a tech-obsessed student in Nanjing. Liu Jingkang (JK) was born in 1991 in Guangdong and developed a precocious talent for computers. As a Nanjing University undergrad, he became a campus legend for audacious geek projects – famously decoding the mobile number of Qihoo 360’s CEO by analyzing phone keypad tones . This stunt won Liu instant fame (and the nickname “standard哥”), and even job offers from tech luminaries. But Liu declined big-company opportunities; instead, his entrepreneurial spark was lit early. Backed by a 15万元 loan from his father and personal savings, he launched a startup while still in school .

Liu’s first venture in 2013 was a campus live-streaming platform called “名校直播,” providing multi-camera video streaming for top universities . The young founder and his small team pulled off hundreds of live events and even expanded into corporate streaming within a year . This hectic baptism in startups taught Liu to “grow through exploration”, but also exposed him to the next big thing: virtual reality. After casually trying a Google Cardboard VR headset and seeing an overseas team capture a 360° balloon video over Australia, Liu had an epiphany – the future was immersive imagery . He pivoted the company to focus on VR panoramic cameras, declaring that Insta360 was born to enable immersive storytelling .

In 2015, at just 24, JK Liu officially founded Shenzhen Arashi Vision Co., branding its products “Insta360影石” (YingShi). The vision was clear: leverage 360-degree imaging technology to help people “better record and share life” in every scenario . Early on, building a VR camera from scratch wasn’t easy – Liu lacked hardware supply chain experience and had only a tiny team. He relocated the startup from Nanjing to Shenzhen to tap into China’s best electronics ecosystem . Even then, their very first camera prototype failed to reach mass production due to technical hurdles. It was December 2015 when Insta360 finally unveiled its inaugural product: a professional VR camera dubbed the “4K Beta.” This chunky, enterprise-grade 360° rig wasn’t a GoPro clone for consumers but rather a “big square camera” designed for event videographers to rent for weddings and parties . It was a humble start, but it demonstrated Liu’s determination to push into uncharted territory of panoramic video.

That bold bet soon attracted believers. In 2014, as China’s “大众创业、万众创新” (mass entrepreneurship and innovation) boom took off, IDG Capital spotted news of the “genius coder from Nanjing University” and came knocking . Impressed by Liu’s technical chops and quick execution – if not yet by a polished business plan – IDG embraced a “invest in the person” philosophy . In March 2015, IDG led a $1 million angel round for Insta360, taking 20% equity. “He had a baby face and was quiet, but his thinking on tech and product was outstanding,” recalled IDG partner Niu Kuiguang of first meeting the 23-year-old founder . Other prominent VCs soon followed – Qiming Venture Partners joined in a Series B, and strategic investors like Xunlei (a Chinese tech firm) and retail giant Suning participated in 2016 financing rounds . By keeping his burn rate low and raising only modest capital (totaling ~¥500 million over 5 years), Liu maintained significant ownership and a laser-focus on product-market fit . Little did early backers know that their faith would be richly rewarded: IDG’s initial $0.65 million stake, held for a decade, swelled to a paper gain of over ¥8 billion (100× return) by IPO time .

From Panoramic Pioneer to Pocket Camera Innovator

The heart of Insta360’s success lies in relentless product innovation, evolving from a niche 360º camera maker into a multi-category smart imaging brand. After the 4K Beta pilot, 2016 proved a breakthrough year. Insta360 launched the Nano, a pocket-sized 360° camera that could work standalone or plug into an iPhone – instantly stitching 360 photos and videos on the go . The Nano addressed a key pain point: earlier 360 cams were clunky and made users wait to see their shots. With its plug-and-play ease, Insta360 Nano caught fire. It debuted to acclaim at global tech expos (CES and Germany’s IFA), and even Apple co-founder Steve Wozniak praised it as a “remarkable product.” Thanks to Nano’s popularity among vloggers and gadget reviewers, Insta360’s revenue spiked to ¥20 million per month that year . However, the 360° camera market was so new that after the initial buzz, sales dipped sharply once early adopters were saturated, underscoring that broader consumer education was still needed .

Flush with 2016’s buzz (and two new funding infusions from Xunlei and Suning that year ), Insta360 expanded its lineup. It introduced the Air (a 360 camera for Android phones) and ventured into professional gear with the 8K Insta360 Pro for VR filmmakers. But 2017 turned into what Liu later called “the toughest year” for the company . The initial 360 craze cooled, no blockbuster new product was ready, and the startup was supporting ~100 employees on dwindling cash . “Our valuation was stuck around ¥500 million and I had no confidence to even seek new funding,” Liu quipped in hindsight . He resisted the temptation to over-hype or burn money recklessly, insisting that in hardware, scaling headcount faster than revenue was a dangerous sign . Instead, the team doubled down on R&D, determined to create a device that could break out of the niche.

Redemption came in 2018 with the launch of the Insta360 ONE X. This sleek, palm-sized 360° action camera became a global hit, thanks to its inventive design and user-friendly features. The ONE X could capture immersive 5.7K video at 30fps with impressive stabilization – one demo famously showed the camera being tossed like a flying frisbee, yet still producing smooth footage . More importantly, Insta360 nailed the software side: the ONE X’s app automatically stitched and edited highlights from the 360 footage, making it effortless for users to share dynamic content to Instagram or TikTok . The impact was dramatic – ONE X catapulted Insta360 to No.1 worldwide in consumer 360 cameras (beating rivals like Samsung Gear and Ricoh Theta), and even to No.2 globally in the overall action camera market, just behind GoPro . In that year alone, Insta360’s revenue surged past ¥200 million with nearly ¥20 million in net profit, proving that the startup’s business model was viable . Remarkably, this ascent happened just 3 years after Insta360’s official founding – a speed that impressed investors and industry watchers alike . As IDG’s team realized by 2018, “the model was validated” and their young portfolio company had indeed built a “global first” in a new category with very limited capital .

Building on ONE X’s success, Insta360 kept up a brisk cadence of product releases. In 2019, it unveiled the Insta360 GO, an ultralight thumb-sized camera meant to clip onto clothing or hats for effortless, hands-free shooting. Weighing under 20 grams, the GO was marketed as a “tiny stabilized vlogging camera” that could be a game-changer for capturing everyday moments in ways a phone or bulky camera couldn’t . Around the same time, Insta360 introduced the ONE R, a modular action camera co-engineered with Leica that featured interchangeable 360° and wide-angle lens mods – directly challenging GoPro’s flagship cams with a fresh twist. The company also continued pushing the pro horizon with devices like the Insta360 Titan, an 11K VR camera for filmmakers.

In 2022, as remote work and live streaming spiked, Insta360 expanded beyond action cams into the desktop imaging realm. It launched the Insta360 Link, an AI-powered 4K webcam mounted on a miniature gimbal that can track the user’s movement. The Link garnered rave reviews – The Verge noted its best-in-class image quality and slick AI tracking software – helping Insta360 tap into the booming demand for high-end webcams. By now Insta360’s portfolio covered a “four-pronged” range of solutions: 360-degree cameras for consumers and professionals, standard action cameras for sports and adventure, wearable mini cameras (GO series) for casual life-logging, and even smart webcams for streaming and video calls . This breadth set Insta360 apart from single-category players. As founder Liu put it, “we make a 360 camera every year to keep our top-tier position” in immersive imaging , but the company was equally adept at applying its imaging know-how to adjacent categories. It is now commonplace to see Insta360’s devices used by everyone from YouTube vloggers and extreme sports enthusiasts to journalists, travelers, and remote workers – all part of Liu’s broader mission to “help people record and share their lives in better ways.”

Building a Sustainable Business Model

Unlike many hardware startups, Insta360 has managed to build a profitable business while scaling globally, in part by smartly balancing its revenue streams and keeping costs in check. The company’s core business is selling its imaging devices and accessories across consumer and professional markets . Each time Insta360 rolled out a hit product (from the Nano to ONE X to GO and beyond), it saw a corresponding leap in sales. By 2019, the company was already profitable, helped by the success of the ONE series – in fact, Insta360 boasted five-fold revenue growth from 2017 to 2019 and turned a profit while GoPro was still losing money . Liu has confirmed that since 2017 the firm has stayed in the black , a rarity among gadget makers. In 2019, Insta360’s annual revenues neared ¥600 million with over ¥56 million net income , indicating healthy margins for a company of its size.

Read more

From die-hard collectors flying across countries in search of rare figurines to K-pop superstar Lisa of Blackpink gushing “I go Pop Mart everywhere” , it’s clear that Pop Mart – a Chinese designer toy brand – has sparked a worldwide craze. At the heart of this phenomenon are Pop Mart’s whimsical blind-box collectibles and its roster of original characters, none more buzzworthy lately than Labubu, a furry “ugly-cute” creature capturing global hearts. This article explores: (1) Pop Mart’s rise from a Beijing boutique to China’s top “潮玩” (designer toy) empire, (2) the story behind Labubu’s creation and why it became a breakout star, and (3) how designer toy culture evolved from niche art into an international trend.

Pop Mart’s Rise and Business Model

Founded in 2010 by entrepreneur Wang Ning, Pop Mart has rapidly grown into China’s undisputed leader in designer toys . What began as a hip lifestyle boutique in Beijing soon tapped into a hidden goldmine: adult collectibles. In 2015, a tiny 8 cm Japanese figurine called Sonny Angel went viral, suddenly accounting for one-third of Pop Mart’s sales . Sensing the public’s appetite for cute collectibles, Wang pivoted the business from reselling trinkets to creating a platform for original toy IP (intellectual property). The next year, Pop Mart licensed Molly – a wide-eyed doll character by Hong Kong artist Kenny Wong – and launched its first blind box series (the Molly Zodiac in July 2016) . The blind boxes – mystery packs where buyers don’t know which figure they’ll get – were a smash hit, propelling Molly to icon status and validating Pop Mart’s new model.

Under the hood, Pop Mart transformed into a vertically integrated “IP engine.” Rather than merely retailing toys, the company began licensing or creating original characters, partnering with artists, and turning their creations into high-quality figurines . It designs and manufactures these collectibles, then sells them through its own vibrant channels – colorful stores and “Roboshop” vending machines across shopping malls, airports, and beyond . (Pop Mart is often said to be “50% an IP company and 50% a retailer,” seamlessly blending content creation with nationwide distribution .) By controlling both the characters and the sales channels, Pop Mart can discover a promising design, build buzz around it, and scale it up into a trend . This IP-centric strategy also enabled collaborations with global franchises (like Disney and Sanrio) alongside the company’s own homegrown characters .

Today, blind boxes remain Pop Mart’s core product – typically priced around $8–$15 each – fueling a treasure-hunt thrill among consumers . Fans eagerly line up outside stores and share unboxing videos online, chasing rare figures to complete their collections. The surprise element of blind boxes, combined with the artful designs and limited-edition “chase” figures, creates a lottery-like excitement . Many collectors feel a sense of community and nostalgia, trading duplicates with each other and displaying their hauls on social media . As one business columnist noted, Pop Mart isn’t just selling toys – it’s selling an experience and emotion to its young customers .

Pop Mart’s growth has been explosive. By 2020, the company’s IPO in Hong Kong sent its stock soaring, minting Wang Ning (born 1987) as one of China’s youngest billionaires . Even a pandemic hasn’t slowed its momentum – Pop Mart boasted over 50 million registered members by the end of 2024 , and its revenues more than doubled in 2024 alone to 13 billion yuan (≈$1.8 billion) . In China, Pop Mart has become synonymous with the “潮玩” (cháo wán) trend, meaning “trendy toys.” The company’s flagship characters like Molly, Skullpanda, Dimoo and others are now household names among Gen Z collectors. In September 2023, Pop Mart even opened “Pop Land,” a 40,000 m² theme park in Beijing dedicated to its toy characters – a testament to how these once-niche art toys have entered the mainstream.

Crucially, Pop Mart also set its sights beyond China’s borders. As early as 2018, it began expanding abroad, and that global push has accelerated in recent years. As of December 2024, Pop Mart operated over 530 stores worldwide (from Asia to Europe and North America) and some 2,490 Roboshops, with more than 130 stores and 190 machines outside mainland China . Locations range from a flagship store inside the Louvre Museum’s mall in Paris to a sprawling new shop in Los Angeles – planting Chinese designer toys in the heart of global shopping districts . International sales are surging (over $700 million in 2024, about 40% of total revenue) as overseas fans join the craze . The Pop Mart mobile app even hit #1 on the U.S. Apple App Store’s shopping chart at one point , underlining its overseas appeal. In short, within 15 years Pop Mart has evolved from a single Beijing storefront into one of the world’s most successful art toy brands – proving that the right mix of creativity and strategy can turn adorable plastic figurines into a cultural force.

Labubu: The Making of an “Ugly-Cute” Icon

If one character encapsulates Pop Mart’s international explosion, it’s Labubu. This impish little creature – often seen as a fuzzy, big-mouthed gremlin with a devilish grin – has become the must-have designer toy of the moment. Labubu was created by Kasing Lung, a renowned Hong Kong-born artist who spent much of his youth in Europe . Lung’s artwork (branded “The Monsters” series) draws heavily from Nordic fairy tales and European folklore, which fascinated him as a child growing up in the Netherlands . “Labubu was born from my fascination with Nordic mythology…creatures that live in the space between fantasy and reality,” Lung says. He wanted to craft a character embodying duality – “mischievous yet endearing, strange yet deeply human” . The result was Labubu: a naughty-but-lovable forest elf with pointy ears, wild eyes and a wide toothy smile – a design that is equal parts cute and creepy. Lung originally featured Labubu in storybooks and illustrations, but partnering with Pop Mart allowed him to bring Labubu from page to physical product, letting fans “collect, connect with, and make [the character] their own” .

Labubu’s art style stands out in the Pop Mart lineup. While many designer toys are unabashedly cute, Labubu leans into a more whimsical, slightly dark aesthetic often dubbed “ugly-cute.” Imagine a creature from Jim Henson’s Labyrinth or Maurice Sendak’s Where the Wild Things Are – but shrunken to keychain size and sold in pastel-colored blind boxes. This quirky vibe has enthralled young adult collectors. Lung notes that fans today love characters that feel playful, expressive, and a little rebellious – and Labubu hits that sweet spot, becoming a form of self-expression for its devotees . The Monsters universe includes other characters (like the rare pink Mokoko or the tailed leader Zimomo), but Labubu quickly emerged as the breakout star of the crew due to that extreme ugly-cute charm . In Pop Mart’s lore, Labubu is a “Nordic monkey-like forest elf” with a mischievous personality . Fans, however, have made Labubu their own: some imagine her (yes, Labubu is often referred to with she/her) as a cheeky gremlin sprite, others simply adore her oddball looks and fuzzy texture.

Since Pop Mart began releasing Labubu toys (first as vinyl figurines and now also as plush dolls), the character’s popularity has skyrocketed across continents. In China, Labubu was an early hit and remains a top-selling IP (often neck-and-neck with Pop Mart’s flagship Molly). But it’s overseas where Labubu truly became a phenomenon. According to Pop Mart’s 2024 sales, Labubu is the company’s bestselling character globally . Part of the boost came from celebrity fandom: Blackpink’s Lisa has flaunted her Labubu collection on Instagram, and Hollywood stars like Emma Roberts and Harvey Guillén have been spotted with Labubu charms . On TikTok, “#Labubu” unboxing and haul videos rack up hundreds of thousands of views . By 2025, reports of Labubu mania were making headlines worldwide. From New York to Bangkok and Dubai, fans are lining up for hours just to snag Labubu keychain dolls . In London, the frenzy got so intense that Pop Mart temporarily pulled Labubu from shelves “for safety reasons” after scuffles broke out in stores . In the Netherlands, the sole retail shop selling Labubu had to hire extra security to control the crowds . Chinese customs officials even cracked down on travelers trying to smuggle suitcases full of Labubus out of the country .

What makes Labubu so beloved? Collectors often say it’s the “ugly-cute” appeal – that paradoxical charm of something that’s grotesque yet adorable. Labubu’s snaggle-toothed grin and scruffy form set it apart from more saccharine mascots; as one fan put it, “she’s weird in just the right way.” The character also benefited from a rich backstory and artistry – being the brainchild of an established artist gives Labubu an artistic pedigree that resonates with adults who view the toys as art pieces. Moreover, Labubu became a flag-bearer of China’s creative culture abroad. State media even called Labubu “a benchmark for China’s pop culture” and proof that Chinese designer toys can captivate international audiences . In 2025, Pop Mart celebrated Labubu’s 10th anniversary with special edition releases, further fueling fan enthusiasm. All this has cemented Labubu as not just another toy, but a crossover pop icon bridging East and West. As Kasing Lung humbly remarked when seeing his sketch travel so far: “Seeing Labubu embraced by fans worldwide…is surreal and incredibly humbling…It pushes me to keep expanding the universe” .

Designer Toy Culture: From Niche to Global

Labubu’s wild success is part of a bigger designer toy movement that has swept through Asia and now the world. Often referred to as 潮玩 (cháo wán) in Chinese – literally “trendy toys” – designer toys are collectible art toys typically created by artists or designers, produced in limited runs, and marketed towards adults rather than kids. This subculture has its roots in the late 1990s, when visionary artists began turning their original characters into stylized vinyl figures. In Hong Kong, for

Read more

Has NIO Truly Passed Its Lowest Point? Recently, NIO’s founder and CEO, William Li (Li Bin), asserted that the company “believes that the lowest point has already passed in the first quarter of this year” , expressing confidence that NIO will return to an upward trajectory thereafter. This bold claim comes on the heels of a turbulent period for the Chinese EV maker. NIO faced steep losses and slumping sales in 2023, followed by a mix of record highs and sharp dips in vehicle deliveries through 2024 and early 2025. Given the intense competition in both China and international EV markets, Li’s optimism warrants scrutiny. Is NIO truly on the rebound, or is this confidence premature? We examine this through two lenses: NIO’s latest sales trends and its competitive position against rivals like Li Auto, Xpeng, BYD, and Tesla.

Recent Sales Performance: Rebound or Temporary Relief?

NIO’s delivery figures tell a story of partial recovery with lingering volatility. In 2023, the company’s sales hit a nadir in the second quarter – June 2023 saw only 6,155 vehicles delivered, marking NIO’s lowest monthly volume in years . This slump coincided with NIO’s transition to a new product lineup: the company launched six new or upgraded models in 2023 (switching all major models to its second-generation platform) . The rollout included the new ES6 and ES8 SUVs, the ET5T wagon, and the flagship ET9 sedan, among others. This ambitious overhaul strained NIO’s finances and temporarily hurt sales as older models were phased out before new models could ramp up. As a result, NIO ended 2023 with 160,000 vehicles sold and a massive net loss of RMB 21.1 billion (approx. $3 billion), a 45% wider loss than the prior year . Clearly, 2023 was a painful year, though arguably not as dire as the 2019 crisis when NIO needed a government bailout.

Encouragingly, sales rebounded in late 2023 and into 2024. With the new models gaining traction, NIO’s deliveries climbed again. The company delivered 221,970 vehicles in full-year 2024, a 38.7% jump from 2023 . Notably, NIO achieved its first-ever five-figure monthly sales in late 2024 – delivering 31,138 vehicles in Dec 2024, a record high (72.9% higher than Dec 2023) . This surge was powered in part by NIO’s new sub-brand “Onvo”, a more affordable, family-oriented EV line. Onvo’s first model (the L60 SUV launched in September 2024) contributed over 10,500 units in December , roughly one-third of NIO’s monthly total, while the premium NIO-branded models made up the rest. By year-end 2024, NIO’s main premium brand was growing modestly (+25.7% YoY in 2024), but the addition of Onvo’s 20,761 units for the year helped boost overall volume significantly .

However, the claim of being past the “lowest point” is not entirely confirmed by recent quarter-to-quarter trends. In fact, after the Q4 2024 peak, NIO’s deliveries dropped dramatically in Q1 2025, reflecting seasonal weakness and the uneven nature of its recovery. NIO delivered 42,094 vehicles in Q1 2025, which was 40% higher than the depressed Q1 2024 (30,053 units) but 42% lower than the bumper 72,689 units in Q4 2024 . This sequential plunge underscores that NIO’s momentum remains shaky. Much of the growth is year-end loaded – likely boosted by incentives and a backlog of new model orders – whereas the start of 2025 saw demand pull back again. Even Li Bin admitted internal issues kept NIO from meeting operational goals in 2022–2024 . The core NIO brand’s sales have stagnated: in March 2025, NIO’s premium models delivered just 10,219 units, down ~14% year-on-year (as March 2024 benefited from the then-new ET5 sedan). It was only thanks to 4,820 Onvo deliveries that NIO’s overall March 2025 volume showed a YoY increase . This suggests that NIO’s original high-end lineup is still struggling to grow in the face of competition, and the company is leaning on new segments (and lower-priced models) to bolster its numbers.

To NIO’s credit, the company has been aggressively expanding its offerings and improving margins from the worst levels of the downturn. After plunging into single digits (even negative territory) in mid-2023, vehicle gross margin recovered to 11.9% in Q4 2023 – a respectable figure among Chinese EV startups (many of which had low or negative margins). NIO achieved this without resorting to broad price cuts, preserving a premium pricing strategy even as a price war raged in China’s EV market in early 2023. Instead, NIO focused on cost control and efficiency: Li Bin implemented a 10% workforce reduction (3,000 layoffs) and shelved certain non-core projects, saving about RMB 2 billion in costs , while simultaneously hiring thousands of sales staff to boost deliveries . These moves helped stabilize NIO’s finances – the company’s cash on hand swelled to a robust **RMB 57.3 billion ($8 billion) by Q4 2023** after securing strategic investments (including over $1 billion from a Middle East fund). With this cash, NIO insists it can ride out the storm and continue heavy R&D spending (over RMB 10 billion per year) on new technology and models.

Looking ahead, NIO’s management is banking on a strong rebound from Q2 2025 onward, fueled by its refreshed product lineup. The company has issued an upbeat guidance for Q2 2025 deliveries at 72,000–75,000 vehicles, which would be an astonishing ~75% jump quarter-on-quarter and ~30% year-on-year . This implies NIO expects monthly sales to average 24k–25k in Q2 – back to record levels. The optimism comes from multiple new launches: deliveries of the flagship ET9 sedan (an electric executive sedan starting around ¥788,000) began in March 2025 , and NIO’s third brand (code-named “Firefly” for a compact EV aimed at Europe and lower-tier cities) is set to debut its first model in mid-2025 . Additionally, the Onvo sub-brand plans two new SUV models in 2025 to target Li Auto’s popular family SUVs . These expansions could significantly boost volume – but also carry execution risks. Ramping up multiple new models and brands simultaneously is costly and complex. It remains to be seen if consumer demand will match NIO’s production ambitions, especially as competition intensifies.

Competitive Landscape: Fierce Battles on All Fronts

NIO’s claim of having weathered the worst must be measured against the competitive environment, which in China’s EV market is nothing short of cutthroat. In recent years, several key rivals have surged ahead or eroded NIO’s early lead in various segments. We evaluate NIO’s position against major competitors on sales, product offerings, and technology – both in the domestic Chinese market and abroad.

Sales Volume and Market Share. In terms of sheer sales, NIO has fallen behind its closest Chinese EV startup peers and is dwarfed by the industry giants. Arch-rival Li Auto has firmly overtaken NIO in the premium segment. Li Auto delivered 500,508 vehicles in 2024, more than double NIO’s volume, after growing ~33% from 2023’s 376,000 units . Unlike NIO’s volatile sales pattern, Li Auto’s growth has been consistently robust, with monthly deliveries frequently in the 25k–30k range by late 2024. In fact, Li Auto’s December 2024 sales neared 60,000 (in a single month) as it pushed toward its half-million annual tally . Li Auto’s focus on extended-range electric SUVs (which cater to family buyers with range anxiety) has proven a winning formula. By contrast, NIO’s highest monthly result to date – ~31k in Dec 2024 – included its new mid-range sub-brand contributions. In the pure premium EV space, NIO’s volumes per model are still relatively modest. For example, its best-selling model historically was the ES6 SUV; the newly introduced second-generation ES6 helped revive sales in mid-2024, but each NIO model typically sells only a few thousand units per month, whereas Li Auto’s Li L7 and L8 SUVs routinely each top 10,000 units monthly soon after launch.

Another compatriot, Xpeng Motors, also staged a comeback in late 2024. Xpeng’s sales had languished in early 2023 (leading to a major strategic overhaul), but the launch of its G6 SUV and a technology cooperation with Volkswagen gave it new momentum. In 2024, Xpeng delivered 190,068 vehicles, a 34% YoY leap , nearly closing the gap with NIO’s total. By December 2024, Xpeng hit a record 36,900 deliveries in one month thanks to strong G6 demand and the rollout of its new P7i sports sedan. While NIO still exceeded Xpeng in annual volume for 2024, the resurgence of Xpeng demonstrates how dynamic the market is – and how NIO faces threats from all sides, including from earlier strugglers turning things around.

The incumbent giant BYD presents the most daunting challenge in terms of scale. BYD’s dominion over China’s new energy vehicle market grew even larger: it sold an astonishing 4.27 million NEVs in 2024, a 41% increase YoY . This includes over 1.76 million pure electric vehicles and 2.49 million plug-in hybrids, spanning price points from budget minicars to luxury models. While many of BYD’s sales are in lower segments that NIO doesn’t target, BYD is increasingly encroaching on the premium space. Its Denza sub-brand (a premium marque co-developed with Mercedes-Benz) sold ~126,000 high-end EVs in 2024 , on par with NIO’s volume. Denza’s SUV and MPV models (priced ¥300k–¥500k) have seen solid acceptance, proving that BYD can move upmarket. Moreover, BYD’s upcoming “Yangwang” ultra-luxury EVs (like the ¥1 million U8 off-road SUV) and its high-performance F-brand indicate an ambition to compete even at the top end. NIO’s core proposition as a premium EV maker with superior service is under pressure as BYD leverages its massive scale to deliver value and technology (e.g. BYD’s in-house Blade batteries and hybrid tech) at aggressive prices. BYD’s volume also gives it a cost advantage that NIO lacks, allowing it to weather price wars and still remain profitable.

Internationally, Tesla remains the benchmark and a fierce competitor in China’s premium EV segment. Tesla’s China sales hit a record 657,000 units in 2024 (up 8.8% from 2023) , making it the best-selling luxury EV brand in China by far. The Tesla Model Y crossover, in particular, has been a smash hit – it was not only the top-selling electric SUV, but one of the top-selling passenger cars in China overall. Priced from ~¥263,000 (after multiple price cuts) , the Model Y significantly undercuts NIO’s SUVs while offering strong performance and brand cachet. Tesla’s aggressive price reductions in early 2023 sparked a price war that forced many EV makers to respond; NIO resisted outright slashing prices, a decision which preserved its margins but likely constrained its sales growth as some price-sensitive consumers flocked to Tesla or BYD. As Tesla continues to expand production and cut costs (its Shanghai Gigafactory now churns out over 800k cars per year ), NIO faces an uphill battle to win over customers who might be weighing a NIO ET5/ES6 against a cheaper Tesla Model 3/Y with similar range and performance. In technology, Tesla also leads in areas like efficient powertrain and automated driving software (though NIO has its own NIO Pilot/NAD system, Tesla’s Autopilot and FSD are strong draws for tech-savvy buyers).

Beyond these major players, the competitive field is crowded. Other Chinese EV startups and tech giants add to the pressure. Zeekr (Geely’s premium EV brand) sold over 70,000 EVs in 2024 and is targeting 140,000 in 2025, with new models that overlap with NIO’s sedan and SUV segments. Huawei-backed AITO and Luxeed EVs are trying to lure high-end buyers with Huawei’s tech ecosystem. Even foreign luxury brands like BMW, Mercedes, and Audi are electrifying their fleets in China – and unlike a few years ago, they now have locally made EV models (e.g. BMW i3 sedan, Mercedes EQE SUV) that compete in NIO’s price range. In short, NIO is squeezed between domestic rivals who are scaling up fast and established luxury marques pivoting to EVs with their brand legacy.

The competition is not just about sales volumes; it’s also about technology and product strategy, areas where NIO must prove its edge to justify Li’s optimism.

• Product Mix and Innovation: NIO has prided itself on being a pioneer – it introduced features like battery swapping, a unique subscription-based battery service (BaaS), and a strong user community culture. These remain differentiators: NIO’s network of over 1,300 battery swap stations in China (as of late 2024) offers convenience that Tesla’s Superchargers or others’ charging networks don’t match in the same way. However, the merits of battery swapping are debated; it’s capital-intensive, and rivals have largely stuck to fast-charging and larger battery packs. NIO’s expanding lineup now covers sedans, SUVs, coupes, and soon MPVs and smaller cars via sub-brands – a broader portfolio than Li Auto (which so far sells only SUVs). This could attract more segments of customers, but also poses challenges in marketing and service. Notably, NIO’s ET5 sedan (launched late 2022) was aimed at the entry luxury segment and initially boosted sales, yet its momentum slowed as competitors (Tesla Model 3 refresh, Xpeng P7i, BYD Seal) wooed the same customers at lower prices. Meanwhile, Li Auto’s focused three-SUV lineup captured the family SUV niche decisively. NIO is responding by pushing into Li Auto’s turf – in 2025, the Onvo brand will launch a 6–7 seater SUV and a large 5-seater SUV to compete directly with Li Auto’s Li L9/L8/L7 series . This is essentially NIO conceding that it needs a more mass-market, value-oriented product line (Onvo) to win volume, instead of relying purely on the high-end niche. It’s a sensible strategy, but success is not guaranteed: Xpeng tried a lower-cost sub-brand (“Mona” platform) strategy as well, and it remains to be seen if NIO can manage the brand positioning without cannibalizing its premium image.

• Profitability and Spending: A key competitive metric is that Li Auto has become profitable, while NIO is still far from it. Li Auto enjoyed a healthy 20.5% gross margin in 2024 (even after some decline from 2023), thanks to relatively higher selling prices and efficient cost control on its range-extended EVs. NIO’s vehicle

Read more

In the global electric vehicle (EV) race, a striking contrast has emerged between Silicon Valley and China. Apple Inc. – which reportedly spent a decade and billions of dollars on a secretive car project without bringing a vehicle to market – now Beijing-based Xiaomi, best known as a smartphone maker, only announced its EV ambitions in 2021 yet managed to roll out its first electric car about three years later . This swift execution highlights China’s industrial drive and the advantages of its booming EV ecosystem. Chinese automakers already dominate the world’s largest EV market, with China alone accounting for 11 out of 17 million global electric car sales in 2024 . Xiaomi’s foray into autos exemplifies how China’s tech firms are expanding into large-scale manufacturing, and could signal new competition for established players in the global EV arena.

Lei Jun Addresses Apple’s EV Struggles and Xiaomi’s Approach

On June 4, Xiaomi’s founder and CEO Lei Jun took to social media to address a question on many minds: Why has Apple failed to produce an electric car after ten years of effort, while Xiaomi succeeded in a little over three? In a post on his official WeChat blog, Lei admitted he isn’t sure why Apple’s car project – often dubbed “Project Titan” – still hasn’t borne fruit. He emphasized that Apple remains “one of the greatest companies on the planet” and a model that Xiaomi has studied and benchmarked against throughout its 15-year history .

By contrast, when explaining Xiaomi’s own rapid success in building an EV from scratch, Lei Jun outlined three key factors :

1. All-Out Commitment and Founder-Led Focus. Xiaomi’s car project was led personally by Lei Jun himself, which he says ensured strong strategic focus and efficient resource coordination . Lei had often espoused a “stealth mode” approach to new ventures – “do it quietly; if it succeeds, then great, if it fails, act like it never happened” . But for Xiaomi’s automotive endeavor, the company took a very different tack: it held a high-profile launch event to declare to employees, partners, and the market that “we will go all out” . From the outset, Xiaomi committed a massive ¥65 billion (around $10 billion) investment into its EV business, giving the team the confidence and runway to focus fully on the long-term goal without short-term distractions . According to Lei, the entire group remained united over the past four years with aligned interests, forming a “very strong fighting force” to get the car initiative off the ground .

2. Respect for Industry Fundamentals and Benchmarking the Best. Lei Jun stressed the importance of reverence for the automotive industry – approaching car-making with humility and diligence. Xiaomi’s strategy follows the mantra “shouzheng chuqi” (守正出奇), which he described as sticking to the correct path while seeking breakthroughs . In practice, this means respecting industry norms and learning from established leaders in areas like mechanical design, intelligent systems, and electrification. “We have made it clear that we benchmark ourselves against Tesla and Porsche – these two companies represent the pinnacle in their respective fields,” Lei wrote. “The gap may be large at first, but as long as we dare to compare ourselves with them, we are on the road to victory” . By openly holding itself to the standards of the world’s best automakers, Xiaomi aimed high and stayed disciplined in its vehicle development.

3. Xiaomi’s Proven Business Model and Methodology. The quick success of Xiaomi Auto is fundamentally a triumph of the “Xiaomi model” and the firm’s operating methodology, according to Lei. Xiaomi had the confidence to start by launching only one car model, and that debut vehicle became a breakout hit in the market . (Xiaomi’s first EV, the SU7 sedan, indeed turned into a bestseller upon its release.) This outcome reinforced Xiaomi’s faith in its established playbook. Lei explained that the company applied the same Xiaomi methodology that drove its electronics business – including its “爆品” blockbuster product strategy, an integrated new retail model, and a high-end product development approach – to the auto sector, significantly improving the odds of success . In short, Xiaomi transplanted the formula that had worked in smartphones and gadgets to its car venture: focusing on one flagship product, delivering strong specs and value, leveraging online-to-offline retail integration, and aiming for an upscale image. This formula, he argued, translated effectively to making cars.

Strong Debut and New SUV YU7’s Outlook

Xiaomi’s YU7 electric SUV, the company’s second model, during its official unveiling. The strong performance of Xiaomi’s first car has set the stage for its next act – the YU7, a new electric SUV that the company unveiled in May 2025. Lei Jun addressed the question of whether the YU7 can replicate the success of the SU7 sedan, and early indicators appear promising. He revealed that in the three days following the YU7’s technology launch event, the number of prospective buyers who registered interest was about three times that of the SU7 during the same post-launch period . Notably, more than 60% of those would-be YU7 customers are first-time Xiaomi car intenders, and over 40% have never used any Xiaomi product before . This suggests the YU7 is attracting a broader audience beyond Xiaomi’s existing fan base, making the new model an even more significant “crossover” hit in terms of consumer reach.

Read more

China’s electric vehicle market has a new surprise leader. Leapmotor (零跑汽车), a relatively low-profile EV startup from Hangzhou, has suddenly surged ahead of more famous peers. In the past few months, Leapmotor has consistently outsold better-known brands like Nio, Xpeng, and Li Auto – marking an extraordinary turn in China’s EV boom. How did this underdog rise to the top? A mix of humble beginnings, affordable models, vertical integration, and a strategic global partnership have all powered Leapmotor’s leap into the limelight.

From Tech Roots to EV Ambitions

Leapmotor was founded in 2015 by Zhu Jiangming, an engineer who previously co-founded surveillance tech giant Dahua Technology . With support from Dahua, Zhu set out to apply his tech expertise to automobiles. The company name “零跑” literally means “zero run,” reflecting an aim to start from scratch and then leap ahead. By 2019, Leapmotor launched its first car – the S01, a small electric sports coupé – and sold its very first vehicles . The S01’s niche appeal meant it wasn’t a breakout hit, but it gave Leapmotor a foothold and valuable R&D experience in the EV arena.

Learning from that, Leapmotor pivoted to more practical, mass-market cars. In 2020 it rolled out the T03 city car, a compact electric hatchback priced from about RMB 66,000 (around $9,300) after subsidies . The T03’s affordability (comparable to a budget gasoline car) and decent 300+ km range made it one of China’s most accessible EVs. It quickly gained popularity, especially in smaller cities. By building a full-stack capability – from design to manufacturing – Leapmotor steadily expanded. The startup secured a production license, opened its own factory in Jinhua by 2021, and even developed some in-house auto tech like its own AI driving chip . In September 2022, Leapmotor went public in Hong Kong (following Nio, Xpeng, and Li Auto’s earlier IPOs) and raised HK$6.06 billion to fuel further growth . Though its IPO debut was rocky, Zhu remained focused on steadily improving products and winning market share .

Record-Breaking Sales Streak in 2025

Leapmotor’s patient approach started paying off spectacularly in 2024 and 2025. The company more than doubled its sales in 2024 to nearly 294,000 EVs , and set an ambitious target of 500,000–600,000 units for 2025 . Early 2025 results suggest they are on track. In March 2025, Leapmotor shocked industry watchers by becoming the best-selling Chinese EV startup, surpassing longtime leader Li Auto . It delivered 37,095 cars that month – a 154% year-on-year jump – edging out Li Auto’s 36,674 units . That was the first time this “new force” EV maker had ever led the pack.

Leapmotor didn’t stop there. In April, it kept the No.1 spot among China’s EV startups with 41,039 deliveries, up 172% YoY . Rival Xpeng also surged to 35,000 units, but Leapmotor still led by a healthy margin . Even Li Auto – known for its strong sales – fell to third among the startups with 33,939 April deliveries . By comparison, premium player Nio sold about 15,000–24,000 cars in those months , underscoring how dramatically Leapmotor’s volume had pulled ahead.

A Leapmotor showroom in Shenzhen displays the company’s EV lineup, including the T03 city car (left), C01 sedan (center), and C11 SUV (right). Leapmotor’s mix of affordable compact models and family-sized vehicles has helped it capture a broad swathe of China’s EV market.

Then in May 2025, Leapmotor hit a record monthly delivery: 45,067 vehicles, a whopping 148% higher than a year prior . This marked its third consecutive month as China’s top-selling EV startup. The feat is all the more impressive considering China’s fierce EV price war this year – established giants like BYD and Tesla have slashed prices, yet Leapmotor has managed to thrive and grow in volume. In fact, aggressive pricing and new model launches helped Leapmotor expand its market share while some bigger automakers felt the squeeze . By the end of May, Leapmotor had delivered over 173,000 cars in 2025 , putting it well on the way to its half-million annual sales goal.

EVs for Everyone: Leapmotor’s Model Lineup

A key factor behind Leapmotor’s rise is its product strategy – an accessible lineup that spans from tiny city cars to roomy SUVs, all at value-driven price points. Unlike some rivals that focus only on high-end models, Leapmotor has a vehicle for various needs, always aiming to deliver more for less. Its current lineup includes:

• T03 – A mini urban EV (launched 2020). This 3-door hatch is ultra-affordable (around $10k) yet offers over 300 km range. The T03 became a hit among budget-conscious buyers and young drivers in China’s cities.

• C11 – A mid-size SUV (launched late 2020). The C11 was Leapmotor’s first mass-market SUV, roughly akin to a Tesla Model Y or Nio ES6 in size. Notably, it’s offered in both pure electric and extended-range hybrid versions , giving buyers the choice of long EV range or a gasoline generator backup (a strategy similar to Li Auto’s).

• C01 – An executive sedan (launched 2022). Sleek and low-slung, the C01 is a fully electric 5-seat sedan about the size of a BMW 5 Series. It shares a platform with the C11 SUV and boasts modern tech like a triple-screen cockpit. An extended-range (EREV) variant of the C01 is also available for those worried about charging .

• C10 – A compact/mid-size SUV (first shown 2023, launching 2025). The C10 is built on Leapmotor’s new global EV platform (LEAP 3.0) and is designed with international markets in mind . It slots slightly below the C11 in size. The company unveiled fresh 2026 C10 models in May 2025, again offering both battery EV and EREV versions to appeal to a wide audience .

• C16 – A full-size three-row SUV (launched 2024). This is Leapmotor’s largest and most premium model to date – a 6-seater SUV aimed squarely at the family SUV segment dominated by Li Auto. Debuting at the 2024 Beijing auto show, the C16 was essentially an enlarged C10, featuring a longer wheelbase and six-seat layout to rival Li Auto’s extended-range SUVs . Like others in the C-series, it comes in both pure electric and hybrid form.

Together, these models cover segments from A00-class mini EVs up to high-end SUVs. Crucially, all are priced more affordably than equivalent models from Nio or Li Auto. For example, the C11 SUV starts around RMB 150,000 (~$21,000), roughly half the price of Nio’s mid-size SUV. The flagship C16 undercuts Li Auto’s 6-seat SUVs by offering a similar size and dual powertrain choice but at a lower price bracket . This value-for-money positioning has opened Leapmotor’s appeal to mainstream middle-class consumers, not just premium buyers. The strategy appears to be vindicated by the surging sales numbers.

The Vertical Integration Advantage

One of Leapmotor’s unique strengths is its commitment to vertical integration – doing more of the engineering and component production in-house rather than relying on suppliers. Zhu Jiangming has often highlighted this as the company’s secret sauce for both innovation and cost control. In fact, among Chinese EV makers, only industry giant BYD (which makes its own batteries, chips, etc.) is more vertically integrated than Leapmotor . Leapmotor designs and builds roughly 60% of its key components itself , including battery packs, electric motors, electronic control units, and even smaller items like LED lighting and bumpers . By comparison, most new EV startups outsource a much larger share of parts.

This in-house approach yields several benefits. First, it drives down costs – the company isn’t paying supplier mark-ups on critical parts. Leapmotor’s founder noted that reducing part costs by even 10% through vertical integration can translate to a 6% drop in the final vehicle cost . Those savings can be passed on as lower car prices (or help Leapmotor’s own margins). Second, it allows Leapmotor to innovate quickly and tailor its technology. For instance, as early as 2018 the firm co-developed its own automotive AI chip called “Lingxin 01” for autonomous driving applications – a rare move for a startup. It has since achieved full autonomy in core systems like battery management, intelligent cockpit software, and driver-assist algorithms. This vertical integration also means different Leapmotor models share many common parts and electronics, simplifying manufacturing. According to the company, its platform strategy yields an 88% parts-sharing rate among models, greatly streamlining the supply chain .

The end result is that Leapmotor can offer high-tech features at lower prices. Even its budget T03 mini-car came with a ~36 kWh battery and modern infotainment for under $10k – a testament to cost-efficient design. Meanwhile, the latest C16 uses a single advanced Qualcomm 8295 chip to power everything from ADAS to surround cameras, thanks to Leapmotor’s optimized software that squeezes maximum performance from that one chip . These kinds of efficiencies through vertical integration help Leapmotor stay competitive against rivals with bigger budgets. (Notably, Leapmotor’s R&D spending as a percentage of revenue has been around 7–10%, roughly half the level of Nio or Xpeng, yet it’s been able to keep up in tech .)

Global Ambitions with Stellantis

Leapmotor’s domestic success has now attracted heavyweight international attention. In late 2023, Stellantis – the Europe-based auto giant behind brands like Peugeot, Jeep, and Fiat – announced a major partnership with Leapmotor. Stellantis invested about €1.5 billion (US$1.6 billion) to acquire a 20% stake in Leapmotor , becoming its single largest shareholder. As part of the deal, the two companies formed a joint venture called Leapmotor International to take the Chinese EV brand global . Stellantis holds 51% of this JV, which has exclusive rights to manufacture and sell Leapmotor cars outside China .

For Stellantis – which has struggled in the EV transition – the tie-up offers access to Leapmotor’s cost-effective EV technology and a new pipeline of attractive models. For Leapmotor, it’s a fast-track to expand beyond China using Stellantis’ worldwide distribution network. Starting in late 2024, Leapmotor began rolling out its cars in Europe under this partnership . The plan was to introduce the T03 (branded as a “supermini” in Europe) and the C10 SUV first, targeting markets like France, Germany, the Netherlands, and more . By leveraging Stellantis’s dealerships and production facilities (one plan considered assembling the T03 in a Stellantis plant in Poland ), Leapmotor aims to offer European buyers a taste of its affordable EVs. Stellantis CEO Carlos Tavares even described the alliance as an “industry first” where a top global automaker joins forces with a new Chinese EV player on a fully global EV program .

The partnership isn’t limited to selling existing models overseas. The two companies have agreed to co-develop future vehicles and share technologies. One early project will see Leapmotor develop an EV for Hongqi (the luxury arm of China’s FAW Group) using Leapmotor’s platform, with production planned in 2026 . In Stellantis’s eyes, teaming with Leapmotor is also a strategic response to similar moves by rivals – for example, Volkswagen’s investment in Xpeng in 2023 to co-develop EVs. These collaborations signal that Chinese EV innovators like Leapmotor have technologies and products attractive far beyond China, and global automakers are willing to bet big on them rather than reinvent the wheel.

Outpacing Nio, Xpeng, and Li Auto

Leapmotor’s sudden rise naturally invites comparison to the “big three” EV startups that have dominated the narrative so far: Nio, Xpeng, and Li Auto. Those companies grabbed headlines in past years with rapid growth, Wall Street listings, and ambitious innovations – yet it is Leapmotor, often dubbed a dark horse, that now leads in deliveries. The contrast highlights differing strategies in China’s ultra-competitive EV market.

Read more