Gasgoo Munich- Unitree Robotics's application for an initial public offering on the STAR Market won approval from the Shanghai Stock Exchange's listing committee on June 1.
In the race for embodied AI, this is undoubtedly a landmark moment.
Unitree has grabbed headlines over the past two years with viral moments like robot dances at the Spring Festival Gala and gold medals at the World Robot Games. Yet, it wasn't until the prospectus was released that the outside world got a complete picture of its actual business performance—specifically, how many robots it has sold and how much money it has made.
Overall, Unitree has carved out a clear lead in motion control and industrial-scale production. But the race is just heating up in the “brain” department—the critical component that determines whether a robot can truly comprehend the physical world.
Image Source: Unitree Robotics
From Quadrupeds to Humanoids
Unitree's story begins with quadruped robots.
Founded in 2016, the company focused its early years squarely on quadrupeds. During this period, Unitree rolled out a succession of products—including Laikago, AlienGo, and A1—gradually laying the technical foundation for its dominance in the field.
A turning point arrived in 2020, when Unitree began advancing two product lines in parallel: the consumer-grade Go series and the industrial-grade B series. With the subsequent launch of models like the Go1, B1, Go2, B2, and B2-W, Unitree grew into a top-tier player in the global quadruped market.
According to the prospectus, Unitree sold over 33,000 quadruped robots between 2023 and 2025, securing the top spot in global market share for multiple consecutive years.
Driving this success is a synergy between the Go and B series that shares technology and splits costs. Consumer-grade products use volume to amortize R&D and supply chain expenses, while industrial-grade models contribute higher margins through greater per-unit value.
In 2023, Unitree initiated a second strategic pivot.
That August, the company released its first in-house developed full-size humanoid robot, the H1, marking its official entry into the humanoid race.
Over the next two years, Unitree rolled out a series of new products through rapid technological iteration.
In May 2024, the medium-sized G1 hit the market with a starting price of 85,000 yuan, becoming Unitree's first humanoid robot to be delivered at scale. Then, in 2025, came the R1 and H2. The R1 drove the entry price for consumer humanoids down to 29,900 yuan, while the H2 focused on high dynamic performance and adaptability to industrial settings.
With a diverse product matrix and relatively approachable pricing, Unitree quickly gained traction in the humanoid market.
Image Source: Unitree Prospectus
By 2025, revenue from humanoid robots reached 868 million yuan, accounting for 51.78% of main business income and overtaking quadrupeds as the primary revenue source for the first time. Shipments of pure humanoid robots topped 5,500 units that year—ranking first globally, according to third-party data.
Underpinning this growth trajectory is Unitree's long-standing commitment to a full-stack, in-house development system.
The industry typically divides the functional architecture of general robots into three layers: “brain, cerebellum, and body.” The brain handles intelligent decision-making and environmental interaction, corresponding to cognitive intelligence. The cerebellum acts as the motion control center, translating abstract intentions into precise physical movements. The body serves as the physical vessel, comprising hardware like joint motors, reducers, and dexterous hands.
Across all three sectors, Unitree has pursued deep in-house development—covering everything from joint modules, motors, and reducers to LiDAR and motion control algorithms. This has not only significantly reduced reliance on external supply chains but also shortened development cycles for new product categories.
The most immediate result has been the full unleashing of product performance.
Image Source: 2026 CCTV Spring Festival Gala Live Broadcast
At the 2026 Spring Festival Gala, 25 Unitree robots performed "WuBOT," a fully autonomous cluster martial arts routine that set records for consecutive parkour moves and backflips. At the first World Robot Games, Unitree took home 11 medals—the most golds and total medals of any participating company. These achievements serve as side-by-side validation of the robustness and reliability of its humanoid robots in high-dynamic tasks.
Market recognition is also evident among investors and clients.
The prospectus reveals that entities affiliated with Meituan collectively hold about 9.65% of Unitree's shares, making it the largest external institutional shareholder. Sequoia China and Matrix Partners China hold 7.11% and 5.45%, respectively. Additionally, corporate investors including China Mobile, Tencent, Alibaba, Ant Group, and Geely Holding are listed among the shareholders.
In its most recent funding round in June 2025, Unitree's pre-money valuation stood at 12 billion yuan, with a post-money valuation of 12.7 billion yuan.
For this IPO, the prospectus filing shows Unitree Robotics plans to issue at least 40.44 million shares, raising 4.02 billion yuan. Based on a public float of at least 10%, the company's current overall valuation is approximately 42 billion yuan.
On the client front, Unitree's roster includes industry giants such as State Grid, China Southern Power Grid, PetroChina, Sinopec, Baowu Group, Amazon, and BASF, along with numerous prominent universities and research institutions.
Notably, overseas revenue consistently exceeded 40% throughout the reporting period, indicating strong brand recognition in international markets.
Despite Its Lead, Challenges Remain
There is no doubt that Unitree has secured a significant position in the embodied AI race. Yet, the risks lurking behind its IPO cannot be ignored.
The first hurdle is the shift in growth gears and mounting pressure on profitability.
Image Source: Unitree Prospectus
In the first quarter of 2026, Unitree's revenue climbed 68.49% year-on-year—a sharp deceleration from the 332.64% surge recorded in the same period of 2025. The pressure on profitability was even more severe: net profit excluding non-recurring items plummeted 52.55%, shrinking to roughly 40.25 million yuan from 84.84 million yuan a year earlier.
Unitree attributed the decline directly to rapidly expanding expenses. R&D costs jumped by about 38.33 million yuan in the first quarter, while sales expenses surged due to spending on brand promotion efforts like the Spring Festival Gala appearance.
Although Unitree forecasts that the year-on-year decline in adjusted net profit will narrow to between 21.97% and 6.43% in the first half of the year, the prospectus offers no clear prediction on whether the combination of slowing revenue growth and rigidly expanding expenses will persist over the longer term.
In fact, Unitree acknowledges in the prospectus that as its revenue base grows, industry hype cools, and competition intensifies, there is a risk that “operating performance may experience significant fluctuations or even a marked decline” in the future.
The second hurdle is the “brain” gap that needs filling.
Unitree's core advantage currently lies in motion control—the “cerebellum” level. However, the key to determining whether a general-purpose robot can handle complex tasks in open environments is the “brain”—the embodied large model.
The prospectus indicates Unitree has made moves in this direction. In September 2025, it open-sourced its in-house general WMA model, UnifoLM-WMA-0. This was followed by the release of the general VLA model, UnifoLM-VLA-0, in January 2026. An industrial-grade embodied large model has already been deployed on a pilot basis in its own factories, capable of autonomously completing tasks such as joint motor assembly.
Yet, compared to its dominance in motion control, Unitree's R&D spending on large models has been “relatively conservative.” The prospectus admits that “during the reporting period, self-developed general embodied large models have not yet been applied at scale to robot products.”
With global technical routes for embodied AI yet to converge and rivals like Tesla accelerating their deployments, the speed at which Unitree closes this gap will directly influence its long-term competitive standing.
Image Source: Unitree Prospectus
Clearly, Unitree is aware of this. According to the prospectus, the IPO proceeds will primarily fund four major projects: intelligent robot model R&D, robot body R&D, development of new intelligent robot products, and the construction of a manufacturing base. Nearly half of the funds are earmarked for tackling core technologies in embodied AI models.
This means Unitree is preparing to ramp up investment to bridge the “brain” gap. However, technological breakthroughs are inherently uncertain. Whether the company can clear this hurdle before the industry landscape solidifies remains a test of its R&D capabilities and financial reserves.
The third hurdle is intensifying competition and pressure on pricing power.
Looking globally at the humanoid robot sector, Tesla has announced the start of small-scale trial production for the Optimus Gen-3, while numerous domestic automakers and consumer electronics firms have officially entered the fray. The deep pockets, supply chains, and manufacturing experience of these cross-sector players pose a potential competitive threat to Unitree in terms of product pricing, market share, and profit margins.
That pressure is already visible in pricing and gross margins.
In 2023, the gross margin for Unitree's humanoid robots stood at 87.67%; by 2025, it had fallen to 63.18%. While still high for the hardware industry, the annual downward trend suggests the sector's “high-premium window” may be closing earlier than expected.
Echoing this trend, Unitree actively cut prices in 2025: the base model G1 starts at 85,000 yuan, while the R1 Air was lowered further to 29,900 yuan.
Image Source: Unitree Prospectus
Overall, the average selling price of Unitree's humanoid robots has plunged from 593,400 yuan per unit in 2023 to 166,400 yuan in 2025. Similarly, its robot dogs have fallen from 38,600 yuan to 30,300 yuan.
While this price compression is partly an aggressive strategy to capture market share, it also confirms a shift in pricing power due to industry competition. As more players with scale advantages enter the arena, cost control will become the core battleground. For Unitree, a first-mover in this space, the ongoing challenge will be maintaining reasonable profit margins while driving volume through price cuts—balancing the need for market expansion with the demands of profitability.
Another detail easily overlooked is the patent portfolio.
As of January 31, 2026, Unitree held 262 registered patents domestically and abroad, but only 20 of these were domestic invention patents. For a company whose core competitiveness relies on technological barriers, that number is not robust.
Unitree acknowledges in the prospectus that the limited patent count “is not conducive to fully safeguarding the legitimate rights and interests of core technologies,” and creates a “potential legal risk” that it will be difficult to effectively prevent and stop technical infringement, imitation, and copying through intellectual property barriers.
Moreover, application scenarios for embodied robots remain relatively limited at this stage, and future market demand carries significant uncertainty.
Unitree bluntly states in the prospectus that real-world application scenarios are numerous and complex, often involving edge cases or unconventional situations. Handling these exceptional circumstances requires high-performance general robots to possess extreme autonomy and adaptability—something that remains difficult to achieve with current technology levels.
By comparison, current commercial applications for embodied robots are largely concentrated in quadruped inspection, scientific research using quadrupeds or humanoids, and consumer entertainment. In broader industrial applications, high-performance general robots have yet to form fully mature commercial use cases. Meanwhile, in consumer applications, high prices and limited functionality continue to test the viability of mass commercialization.
Accordingly, the largest revenue source for Unitree's humanoid robots remains research and education. Conversely, industrial applications—which hold broader market potential—are still largely in the phases of development, training, and validation.
Conclusion
Unitree Robotics's IPO approval marks a critical step for the embodied AI sector as it moves from concept validation to capital market pricing.
Yet for Unitree itself, going public is just a milestone, not the finish line.
Looking ahead, the key questions remain: Can Unitree extend its first-mover advantage in motion control to embodied large models? How will it balance R&D spending with profit expectations as growth shifts gears? And how will it maintain pricing power in an increasingly crowded competitive landscape? The answers to these questions will determine whether Unitree becomes a briefly shining star or a technology company that truly endures across cycles.
