Yushu Technology IPO: A "Lame" Leader?

Source: Caijing Woji

In 2026, the U.S. did not end up waiting for Tesla’s Optimus to enter mass production, but China’s capital markets may welcome Unitree Robotics first.

On the evening of March 20, the Shanghai Stock Exchange officially accepted Unitree Robotics’ STAR Market IPO application, and before that, Unitree Robotics had already completed the SSE’s preliminary review, including responses to two rounds of inquiries.

According to the prospectus, Unitree’s shipment volume of humanoid robots in 2025 was 5,500 units (excluding wheeled robots), ranking first in the world.

Meanwhile, according to estimates by the industry research firm Omdia, the global shipment volume of humanoid robots in 2025 was about 13k units. That means Unitree alone captured nearly half of the global humanoid robot market.

You have to know that whether it’s most other domestic humanoid robot companies, or the overseas “pioneer” Boston Dynamics, or the “newcomer” Figure, or even Tesla, they are all still in the R&D and validation stage.

Why is Unitree so special—what exactly did it get right? When the entire industry is still trapped in the triple dilemma of “technological breakthroughs are difficult, commercialization lags, and losses persist,” what does Unitree’s listing mean?

In any case, Unitree’s listing is a major event for China’s embodied intelligence industry. Its significance lies in delivering a piece of business evidence that can be verified, and pointing humanoid robots toward something more concrete: robots are starting to be produced, sold, and delivered at scale—just like consumer goods.

Objectively speaking, Unitree’s financial figures are somewhat beyond what outsiders expected.

Why do I say that? Because in a not-yet-mature track like humanoid robots, Unitree has achieved “low investment + high growth + high profit”—a “contrary to common sense” performance report.

Looking at the details, the prospectus states that Unitree Technology, over the three years and nine months in the reporting period (from 2022 to the first three quarters of 2025), accumulated revenue of more than 1.8 billion yuan and net profit of 169 million yuan.

Don’t underestimate these numbers.

To break it down, although Unitree is now known to everyone, having appeared on CCTV Spring Festival Gala three times and repeatedly topped national performances and online trending searches, its market investment is actually not that high.

For example, in January–September 2025, Unitree’s advertising expenses were only 22.5728 million yuan—according to some netizens, this level of investment is even less than many internet companies’ single marketing campaigns.

What stands in sharp contrast with low investment is that Unitree’s revenue in recent years has surged rapidly.

From the pace of development, 2022–2023 were the startup period, with revenue of 123 million yuan and 159 million yuan respectively—small scale and consecutive losses for two years, with net losses of about 22 million yuan and 11 million yuan respectively;

In 2024, it entered an acceleration period: revenue rose to 392 million yuan, up nearly 146% year over year, and it achieved annual profitability for the first time. Full-year net profit was 95 million yuan, and after non-recurring items it attributable net profit was 78 million yuan;

2025 was the breakout period. In just the first three quarters, revenue already reached 13k yuan, net profit was about 105 million yuan, and after non-recurring items it attributable net profit rose to 431 million yuan.

What supported this performance growth was humanoid robots and high gross margins.

Unitree Technology’s business mainly consists of two segments: main business and other business. The main business is essentially “selling robots,” including humanoid robots, quadruped robots, and robot components.

From the revenue structure, behind the exponential growth is that robots are gradually “getting up”: quadruped robots’ revenue doubled for two consecutive years, but they were still surpassed by humanoid robots, which grew even faster.

Unitree’s humanoid robot business began in August 2023, scaled up in 2024, and saw explosive growth in 2025. In the first three quarters of 2025, humanoid robot sales revenue was 595 million yuan, more than five times the full-year figure of 2024.

More importantly, in addition to making robots, Unitree’s profitability is higher than that of its peers.

The prospectus shows that its consolidated gross margin was 56.98% in 2024 and 59.83% in the first nine months of 2025. By comparison, Uibitselect and Yuejiang Technology’s gross margins in 2024 were 28.65% and 46.56% respectively.

This means Unitree’s growth engine is not only humanoid robots—it also has exceptionally strong profitability.

That’s also why people say the prospectus hides a secret that makes you’s back go cold: Unitree wants to be the “BYD” in the embodied intelligence space.

In the past few years as China’s auto industry shifted from oil to electricity, BYD and new forces building cars have shown clear technical differences. With a vertical integration model, the former used extreme cost control, upstream and downstream supply chains, and rapid technological iteration to become the world’s sixth-largest automaker.

And today, we see the same shadow on Unitree.

In fact, Unitree has built fully self-developed capabilities across key components and the whole system products, including motors, reducers and joint modules, battery modules, circuit modules, structural components, and more.

At the same time, it doesn’t avoid outsourcing processes where technology maturity is high and market supply is sufficient, adopting a strategy of subcontracting/manufacturing partners to optimize resource allocation and improve cost-effectiveness.

In addition, Unitree’s “extreme cost control” doesn’t stop at the production end with careful spending. Even with “floods of traffic” like exposure on the Spring Festival Gala, the company has not blindly expanded production capacity. Instead, it precisely matches supply and demand through a “produce based on sales” model.

In other words, this combination of “self-controlled and self-reliant barriers + flexible strategies to control costs + precise production and sales to improve efficiency” allows Unitree to break out of the weird loop of “burning money to buy growth” in a frontier technology track. The essence of its high gross margin is the concentrated realization of three advantages: technological self-reliance, scale effects, and fine-grained operations.

To a certain extent, Unitree has been successful.

Open up Unitree’s shareholder list—Sequoia, Meituan, Shunwei, Deep Creative Investment… the “Hall of Fame” of China’s hard-tech investment circle is nearly all there. Multiple rounds of continuous capital injections from these top institutions have driven Unitree Technology’s valuation to climb rapidly.

According to public disclosures, Unitree plans to raise 1.17B yuan, issuing shares of no less than 10%, corresponding to an overall valuation of 42 billion yuan.

Unitree’s success proves one thing: whoever can turn the technology in the laboratory into products that users are willing to pay for will gain the market pricing power—and even the benefits of being first to list on the capital markets.

However, this track is far more “competitive” than people imagine.

More than 20 companies, such as Leju Robotics, Cloud Deep, Stand, U&I Intelligent (Youai Intelligent), Luoshi, XianGong Intelligent, AntoMo, Jiazhi Technology, Kanovo, JiuWu Intelligent, and others, have already clearly announced listing plans. Each has its own special strengths.

Unitree’s near future will definitely face problems like “the competition king” BYD. Against the backdrop of intensifying external competition, BYD’s path to growth driven by scale advantages is running into bottlenecks, and to seize opportunities in mid-to-high-end niche markets, intelligent capabilities are key.

In fact, over the past few years, BYD has been “catching up” on intelligence.

For a company in embodied intelligence, Unitree’s challenge also comes from intelligence—what you might call the robot’s “brain.”

Breaking it down, although Unitree is known for being “technology-driven,” its investment in the future is clearly insufficient.

The financial reports show that Unitree has only 480 employees in total. In terms of personnel composition, its R&D staff is 175 people, accounting for 36.5%, which is not high.

More importantly, in R&D spending, Unitree accumulated 150 million yuan from 2022 to 2024, and 90.2 million yuan in the first nine months of 2025. Compared with peers, Uibitselect’s R&D in 2024 was 478 million yuan, and Yuejiang was 71.79 million yuan.

Generally speaking, a robot’s capabilities can be broken down into the “brain” and the “cerebellum.” The cerebellum handles motion control—making the robot run, jump, do backflips, and maintain balance. In this dimension, Unitree has reached top level in the industry: backflips, side flips, and cluster martial arts performances—the technical strength is beyond doubt.

But the brain determines a robot’s ability to understand, interact, and make autonomous decisions. Without a mature brain, a robot can only execute pre-set instructions; it can’t truly understand the environment or autonomously plan tasks. This is the key to whether general-purpose robots can move out of the lab and into factories and homes.

From this perspective, the underlying logic supporting Unitree is essentially the result of “short-term traffic dividend” and “supply gap” acting together.

The prospectus also shows that from 2022 to 2024, the proportion of overseas revenue stayed above 55% for three consecutive years, meaning domestic revenue proportion was below 45%; in 2025, domestic revenue proportion was 60.80% and overseas revenue proportion was 39.20%.

2025 was the first year Unitree’s domestic revenue proportion exceeded its overseas proportion, but the absolute value of overseas revenue continued to grow.

Unitree attributes the change in its revenue structure to the Spring Festival Gala’s boost. After all, before the 2025 Spring Festival Gala, embodied intelligence in mainland China still belonged to “underwater” projects—its performance attributes were not yet developed, the attention economy hadn’t formed, and demand hadn’t even begun to emerge. Meanwhile, overseas research institutions and university labs have a greater need for the robot than local supply can provide, which gave Unitree an opportunity.

Put simply, its customers are mainly research institutions, commercial performance venues, and some humanoid robot companies purchasing units to test algorithms. These scenarios are certainly good, but none are large-scale and stable demand scenarios.

That is to say, Unitree’s success—and the overall prosperity of China’s humanoid robotics—is built on a base that is not solid.

This is clearly very risky. An increasing number of humanoid robot companies will inevitably face serious homogeneous competition in the near future, followed by reshuffling and liquidation.

So, the global humanoid robotics industry in 2026 is in a highly deceptive stage: on the surface, hardware prices are plunging sharply, but in reality the competitive barrier is increasing geometrically with the emergence of foundation models.

And this will be Unitree’s next major test.

In the humanoid robotics industry, the U.S. and China have taken different paths—

Whether it’s smart cars in the past, or AI today, or humanoid robots to come, China has the world’s most complete industrial chain categories. And precisely because of this, China’s humanoid robotics companies generally focus on the hardware body of robots.

Meanwhile, the North American market still maintains an advantage in the underlying AI architecture, foundation models, and compute infrastructure.

When we split open Unitree’s humanoid robot supply chain, we find that behind it stands a strong and inexpensive underlying supply chain built by advanced domestic manufacturing enterprises in China.

As the leading reducer company, Zhongdali De (002896.SZ) covers 63% of Unitree’s procurement volume of planetary reducers. In the domestic 3D vision field, the leading company Orbbec (688322.SH) holds 72% of Unitree’s procurement volume of 3D vision modules. The miniature motors “dark horse” Mingzhi Electric (603728.SH)’s hollow cup motors are Unitree’s exclusive supplier for that product category (accounting for over 15% of procurement value in the same category)…

Without exaggeration, Unitree has taken the hardware body of China’s humanoid robotics industry to an unprecedented height.

But it also must be admitted that this robot is “lame-legged.”

At the GTC this March, Wang Xingxing proposed that the industry still has a clear gap from truly crossing the “ChatGPT moment.” The biggest challenge is that no embodied intelligence model has yet emerged that genuinely has strong generalization ability and can reliably execute tasks in unfamiliar scenarios.

Compared with Tesla, which is deeply tied to Grok, Unitree may be farther away from such a “brain.” The Qwen-based self-developed UnifoLM foundation model is still in the R&D and testing stage; previously, Unitree relied more on third-party large language models to provide interaction capabilities.

On March 25, Tesla released the latest Optimus video through official channels, for the first time presenting in a concentrated way the robot’s core hardware designs and development environment, including the reducer gearbox and dexterous hands. It showed the concrete degree of engineering progress.

“We’ve entered the final completion stage of Optimus 3. It will undoubtedly become the most advanced robot in the world, and no products can rival it. In fact, I’ve never seen a robot demo whose performance can match Optimus 3.”

Musk then said that Optimus 3 is expected to start production this summer, and plans to achieve large-scale mass production in 2027, with an initial annual production capacity target of 1 million units. In the picture he描绘s—Optimus may ultimately surpass Tesla’s car business and create revenue as high as $1 trillion.

Facing the formidable competitor coming next, Unitree and the “Unitree group” behind it need to answer one question: how to build a closed-loop of “high-performance hardware + general-purpose foundation model + high-quality data flywheel,” and first form a closed-loop profit in specific sub-scenarios to support the subsequent leap into the general services domain.

From this IPO, Unitree has already started “catching up.”

The prospectus shows that Unitree plans to invest 48% of the total funds raised (about 2.02 billion yuan) into smart robot model R&D. This means Unitree is trying to shift from “leading in hardware” to “competence in both software and hardware.”

In this wave of the tech revolution, the end point of the race is not who lists first or who makes the product first, but who can transform the capital’s scorching heat into real productive power on the production floor—truly making humanoid robots stand firmly on the ground of factories and homes.

Unitree Technology has already submitted its IPO answer sheet, but the real test is just beginning. And these tests are even harder—and even more critical.

Massive information, precise interpretation—available in the Sina Finance app

责任编辑：杨红卜