In 2009, Wang Xingxing, after reading several books on neural networks and artificial intelligence in the library of Zhejiang Sci-Tech University, independently built a bipedal humanoid robot. At that time, few could have predicted that this once extremely introverted and self-doubting student with uneven academic strengths would, over a decade later, shine on the Spring Festival Gala stage and seek to enter the STAR Market with an initial valuation of at least 42 billion yuan.
However, others have observed the other side of the coin. As its momentum to list on the A-share market surged, Unitree Robotics' latest disclosed prospectus (the version for the listing review committee meeting) shows significant changes compared to the version disclosed in March, particularly in the risk disclosure section. The revisions include five major modifications, such as adding a "prompt on year-on-year decline in net profit" and placing "slowing growth and operating performance volatility" at the top of all special risk warning clauses. Additionally, the company has adjusted its fundraising projects. Many market participants have also speculated about regulatory signals, though some interviewees have dismissed this as "overinterpretation." All of this occurred just before the listing review meeting scheduled for June 1st.
The Pause Button on High-Speed Growth If one only looks at the financial reports from the previous two years, Unitree Robotics almost fulfilled all of A-share market's imagination for the "first humanoid robotics stock." Data from the earlier prospectus (the filing version) shows that from 2023 to 2025, Unitree Robotics' operating revenue soared from 159 million yuan to 16.99 billion yuan, with a compound annual growth rate of 226.78%. The company's net profit after deducting non-recurring items reversed from a loss of 18.0191 million yuan to a profit of 591 million yuan. Its gross profit margin for the main business increased for three consecutive years, rising from 44.22% to 60.13%.
While its peer, Ubtech Robotics, saw its operating revenue exceed 20 billion yuan in 2025, its losses were equally staggering—a net loss of approximately 790 million yuan, with cumulative losses over the past four years exceeding 4.2 billion yuan. Other embodied intelligence companies that have applied for listing or are already listed in Hong Kong are even further behind. In 2025, the operating revenues of Dobot, DeepRobotics, and Leju Robotics were only 492 million yuan, 337 million yuan, and 258 million yuan, respectively, with net profits attributable to the parent company of -20.57 million yuan, 28.684 million yuan, and -69.7794 million yuan, respectively.
During the reporting period, Unitree Robotics' cumulative sales of quadruped robots exceeded 33,000 units over three years, with its global market share consistently ranking among the top. Its humanoid robot shipments (pure humanoid, excluding wheeled dual-arm robots) exceeded 5,500 units in 2025, ranking first globally with a 32.4% share, significantly surpassing its peers.
However, in the prospectus for the listing review meeting two months later, Unitree Robotics' performance suddenly decelerated. In the latest prospectus's "Major Matters Disclosure," the company specifically added a "prompt on year-on-year decline in net profit," noting that in the first quarter of 2026, it achieved operating revenue of 423 million yuan, a year-on-year increase of 68.49%, while net profit after deducting non-recurring items was 40.2536 million yuan, a year-on-year decrease of 52.55%.
Unitree Robotics attributed the slowdown in growth to "the significant increase in the 2025 revenue base, the gradual moderation of industry enthusiasm, and the increasingly fierce market competition." The decline in profit was due to "a substantial increase in period expenses such as R&D and sales expenses." According to the company, in the first quarter, it continued to increase technological investment and expand its team in areas such as robot body and structure R&D, embodied intelligence large models, and motion control algorithms, driving a sharp year-on-year increase of 38.328 million yuan in R&D expenses for the period. Simultaneously, intensive brand promotion through platforms like this year's CCTV Spring Festival Gala also pushed up sales expenses for the period.
Looking ahead to the first half of 2026, Unitree Robotics expects revenue growth to further slow to 35.62%–45.41%. Net profit after deducting non-recurring items is projected to be between 236 million yuan and 283 million yuan, a decrease of 21.97% to 6.43% compared to the same period last year. Behind the decline in performance lies the reality of a rapidly changing competitive landscape in the humanoid robotics sector. As market competition intensifies, the industry leader, Unitree Robotics, is undergoing a dramatic shift from a "high-growth myth" to "growth rate adjustment and strategic reshaping."
The Evolution of the "Risk Lexicon" Equally noteworthy as the performance update is the reshaping of Unitree Robotics' risk disclosure framework. Comparing the prospectus filed in March this year with the latest version for the listing review meeting in May, Unitree Robotics has made significant adjustments to the description and prioritization of risks, with its special risk warnings increasing from six to seven items.
Among these, the primary risk has changed hands. "The risk of slowing growth and operating performance volatility" has replaced the initially broad risks such as "failure to meet expectations in technological breakthroughs and product innovation" and "failure to meet expectations in large-scale commercial applications downstream." Combined with the company's first-quarter financial data, this statement indicates that the company is directly confronting the reality of slowing growth.
The second risk warning is also particularly telling. The prospectus changed "the risk of intensified industry competition and disorderly or improper competition" to "the risk of intensified industry competition and weakening leading advantages." The shift from "disorderly or improper competition" to "weakening leading advantages," while only a subtle textual change, subtly conveys a fundamental shift in the company's perspective. The former primarily explained potential "disorderly or improper" industry trends from an external competition angle, while the latter begins to examine internally, acknowledging that the company's moat may be shaken or even eroded.
Unitree Robotics mentioned that as global tech companies accelerate their efforts, startups emerge widely, and cross-industry entrants actively deploy, market competition in the embodied intelligence industry is further intensifying. In its latest prospectus, the company specifically "named" the potential threat from Tesla Motors, candidly stating: "Tesla Motors, as an international tech company with capabilities in mass production and supply chain integration, AI technology resources, and advantages in deploying its own factories, has announced that its humanoid robot, Optimus Gen-3, has begun small-batch trial production. Future commercial mass production will place it in direct competition with industry players, including our company."
Previously, Tesla Motors announced that its third-generation humanoid robot is expected to debut in mid-2026, with production starting as early as late July to August at its Fremont factory in California. The former Model S/X production lines have been fully converted into dedicated Optimus production lines, with a designed annual capacity of 1 million units.
Additionally, Unitree Robotics' description of core technology risks has also changed, shifting from "the risk of key technology leakage and loss of core personnel" to "the risk of key technology protection and loss of R&D talent." The prospectus shows that, previously, based on considerations for protecting technical information confidentiality, the company applied for a relatively small number of patents. As of January 31, 2026, it had obtained a total of 262 registered patents domestically and internationally, including 20 domestic invention patents. From a legal perspective, a smaller number of patents is not conducive to the company fully safeguarding its legitimate rights and interests in core technologies.
Shifting from "Body" to "Brain" Notably, in terms of R&D, Unitree Robotics made an even more surprising and bold admission—the prospectus for the listing review meeting added "the risk of R&D investment direction and effectiveness falling short of expectations." Unitree Robotics frankly stated that its previous R&D investments focused on the body and "cerebellum" (motion control and limb coordination) areas, with less investment in the "brain" (embodied large models), and it had not conducted large-scale real-world data collection or factory deployment training.
Simultaneously, it pointed out that since 2024, the company has gradually increased R&D investment in embodied large models, i.e., the brain. Although phased R&D achievements have been made, the proportion of R&D investment during the reporting period was relatively small. In recent years, many peer companies have moved ahead of Unitree Robotics in this field, accelerating deployments in data collection and training to promote the faster application of robots.
For example, Zhiyuan Robotics has deployed data collection centers in Shanghai and Chengdu; Paxini Perception Technology has built four super data collection factories; Luming Robotics has established three standardized data collection sites; JD.com announced it will build the world's largest embodied intelligent data collection center; and Leju Robotics officially released its Data Collection Training Field 2.0 in April this year.
Analyst Gao Chang from the Gaogong Robot Industry Research Institute believes that industry competition has now shifted from competing on "bodies" to competing on "brains," and the fuel for the "brain" is data. Companies need to build a data flywheel of "scenario—data—model—body," and the speed at which this flywheel operates determines the speed of performance improvement in embodied intelligence.
Wang Tianmiao, Honorary Director of the Beihang University Robotics Institute, also predicted that one of the core trends in the second half of embodied intelligence over the next 3–5 years will be "the gradual convergence of embodied large models with multimodal data fusion and edge-side chip embedding into carriers, moving toward large-scale practical application." Wang Xingxing is clearly aware of this as well. In an interview in February this year, he直言, "(The biggest problem in the current development of embodied intelligence) is that the capabilities of the AI models themselves are still insufficient, with inadequate generalization and universality. Robots can achieve nearly 100% operational success rates in fixed-scenario training, but with slight changes in the scenario, success rates drop significantly. This is also the core bottleneck facing the industry."
Facing this challenge, Unitree Robotics has begun to accelerate catching up in the "brain" race. In the first quarter of this year, Unitree Robotics released the "UnifoLM-WBT-Dataset" full-body teleoperation real-world dataset. In January 2026, it also open-sourced and released the general VLA model "UnifoLM-VLA-0" and completed embodied large model deployment testing in its own factory.
Simultaneously, Unitree Robotics made minor adjustments to the use of raised funds in the prospectus for the listing review meeting. Among these, the investments for the "Headquarters Base Construction Project" and the "Global Technical Service and Sales Network Construction Project" were adjusted to be funded with the company's own capital. All raised funds will be directed toward four core R&D and manufacturing projects, with the "Intelligent Robot Model R&D Project" being the largest single item, planning to raise approximately 2 billion yuan. This also means that Unitree Robotics hopes to concentrate all raised ammunition on the front lines of technological攻坚 after the IPO.
Entering the New Track of "Ecosystem Builder" In fact, whether it is the performance deceleration due to increased investment, the admission of shortcomings through revised risk warnings, or the strategic adjustment to攻坚 the "brain," a clear narrative thread can be traced from the details of Unitree Robotics' series of prospectus changes: it is undergoing a difficult transformation from a "hardware manufacturer" to an "industrial ecosystem builder."
Unitree Robotics stated that the company aims to build an ecosystem around its products, open-sourcing the entire process algorithms for data collection, model training, real-world deployment, and motion control based on its products, to attract global developers to participate in algorithm optimization and general robot R&D. However, this transformation is inevitably accompanied by阵痛 on the profit side and a sharp increase in前期投入. Behind the halved net profit after deducting non-recurring items in the first quarter lies this company's proactive让步 and strategic investment during the industry paradigm shift.
Although humanoid robot shipments have surged in recent years, and institutions generally expect them to maintain relatively high growth rates, the competitive landscape of embodied intelligence is also moving from the "everyone showcasing their strengths" of the past two years to a watershed of survival of the fittest. In 2025, the unit price of humanoid robot整机 dropped from hundreds of thousands of yuan to within tens of thousands of yuan or even below ten thousand yuan. Since the beginning of this year, voices suggesting that the AI industry, including embodied intelligence, contains a "bubble" have grown louder.
A research report from Kaiyuan Securities indirectly confirms the changes in the industry landscape. The report指出 that starting from the second quarter of 2026, the humanoid robot industry is accelerating toward mass production and capitalization. Tesla Motors' Optimus V3 has begun mass production, Figure 03 has completed超长 testing, and four domestic manufacturers, Unitree, DeepRobotics, Leju, and Dobot, are密集推进 IPOs during the same period. The industry has entered a stage of hardware落地, with ecology and capital advancing in parallel.
Wang Feili, China Industrial Industry Analyst at UBS Securities, recently stated that even if many manufacturers reach tens of thousands of units this year, humanoid robots may not truly enter a commercialization拐点, because current industry orders are still dominated by "validation purchases" rather than expansionary purchases based on productivity needs. Unitree Robotics used a three-year revenue compound growth rate of 226% to tell a "miracle story" of a Chinese硬科技 company sprinting. However, industry规律一再证明 that the core competitiveness of robots does not lie in short-term shipment races, but in who can率先跨越 the critical threshold from the laboratory to the factory floor, from科研展示 to production line operations.
Comments