By Xie Zefeng, Edited by Yang Xuran
A rather fragmented phenomenon has persisted in the robotics sector for quite some time:
On one hand, there are commercial robot companies that have been established for over a decade, shipped hundreds of thousands of units, and achieved a closed-loop business model, generating substantial annual revenue and even profits. Yet, their valuations have long stagnated at around several billion to over ten billion CNY;
On the other hand, humanoid robot or component startups established only two or three years ago, with little to no revenue and still in the early stages of development, can secure valuations of over ten billion or even twenty billion CNY.
Commercial robot companies with strong commercialization are generally given low valuations, while startups in the early stages of commercialization can experience rapid ascents in the capital market in a short time. Facing shrinking valuations or market capitalizations, many commercial robot companies have started labeling themselves with the hot tag of "embodied intelligence" to attract attention.
Just as the halo of the "Four AI Dragons" from the AI 1.0 era fades, the valuations of Zhipu AI and Moonshot AI have skyrocketed like rockets. The Chinese robotics industry also seems to have emerged a valuation gap between "two eras." But can the reality in the robotics field really be entirely mapped using past capital stories?
This capital chasm, wide to the point of exaggeration, still spans the current capital market and remains difficult to bridge.
This article is an in-depth value piece from the content team of Juchao WAVE. Welcome to follow us across multiple platforms.
The Chasm
Behind the same "robot" label lie completely different survival realities.
On September 9, UDI Robot (03231.HK) was listed in Hong Kong, claiming to be the "first stock for all-scenario commercial service robots." Founded by Lu Ying, former CEO of UTStarcom, this robot company is already 13 years old. In an era where new contenders emerge frequently, this age already belongs to the "previous wave."
The company's main products are delivery and cleaning robots. If you stay in a hotel, you might sometimes encounter robots delivering items in the corridors or elevators, which are very likely produced by UDI.
It has cumulatively sold 114,000 robots, with a revenue of 318 million CNY last year. In the first half of this year, its adjusted loss was less than 80 million CNY, and the gross margin reached 11%. Before the IPO, its valuation in the last round was only 3.84 billion CNY, and its market capitalization since the listing is only 15 billion HKD.
Similarly, Geek+, which has ranked first globally in autonomous mobile robots (AMR) for seven consecutive years, achieved adjusted profitability in 2025. In the first half of this year, its gross margin was 36%, with overseas revenue accounting for over 75%, and a customer repurchase rate of 80%. However, despite being the champion of this track, its P/S ratio (2026 forecast) is less than 4 times. As its stock price continues to fall, its total market capitalization is now less than 8.5 billion HKD.
In addition, 10-year-old Pudu, the "leader" in global commercial robots, has a cumulative shipment of over 130,000 units. In 2025, its revenue doubled, and its EBITDA is close to turning positive, with a current valuation of around 10 billion CNY. Its cumulative financing since establishment is only 2 billion CNY, which is merely equivalent to the financing scale of Spirit AI's Series A round.
Other peers face the same dilemma. Keenon, established earlier, has shipped over 100,000 units but is valued at only about 1 billion USD. Gaussian Robotics is a pioneer in commercial cleaning robots, valued at the ten-billion CNY level. After Yunji's stock price plummeted by 70%, it is left with only 5.7 billion HKD.
Almost in the same time window, Spirit AI, which bets on the robot brain, has raised about 6 billion CNY since its establishment. This year, it absorbed 4.5 billion CNY in less than four months, elevating its post-investment valuation to the 20 billion CNY level. Established just over two years ago, it took only 26 months to reach a 20 billion CNY valuation. Such performance is not even an isolated case in the capital market.
Galbot, from its establishment in 2023 to March this year, has raised a cumulative 7.5 billion CNY, with a valuation exceeding 20 billion CNY. In early March, the company absorbed 2.5 billion CNY in one go, raising the bar for the largest single financing amount in the robotics industry that it had previously set. The company has become the highest-valued unlisted embodied enterprise in China at present, and its commercialization path is also in the early stages.
Relying on "real-scenario embodied data" and world foundation model capabilities, Astribot has been established for less than three years. In February and April this year, it raised a total of 3 billion CNY, and its valuation doubled in two months to 20 billion CNY, becoming a star enterprise highly favored in China's primary market.
In the first half of this year, the primary market for embodied intelligence was extremely hot. A total of 93 billion CNY in funding was poured into this track, a year-on-year surge of 5 times, completing a total of 322 financing rounds.
Some investors have publicly reminded embodied intelligence founders, "Don't fuss over valuations, get the money as soon as possible." As this track continues to boom, the valuation divergence between it and the commercial robot field is highly likely to persist in the long term.
Divergence
Data shows that, except for Unitree, the vast majority of embodied enterprises are still in a state of loss. From January to October 2025, the total annual revenue of the first-tier embodied enterprises was less than 10 billion CNY, while industry financing in the same period exceeded 50 billion CNY.
The valuation difference in the capital market for different types of robot companies is essentially the pricing gap between "traditional industrial consumer goods" and the "ultimate form of AI."
The reason why the valuations of commercial robot companies such as Geek+, UDI, and Kaileshi Technology are far lower than those of embodied intelligence companies like Unitree and Galbot is that the market views the former as a type of "equipment" and the latter as the physical entry point for Artificial General Intelligence (AGI).
Robot companies are generally in the stage of strategic losses, so the market mostly adopts P/S as a reference standard. For example, after Unitree's listing, its P/S ratio exceeded 260 times at its peak. Even after the market capitalization fell back, it was still 60 times (estimated based on the 2026 revenue forecast).
FANUC, one of the "Big Four" in industrial robots, has long maintained a stable Price-to-Sales (P/S) ratio of around 10 times. This is already the highest valuation level given by the global capital market to the industrial robot field.
In contrast, Geek+'s forecasted P/S ratio for 2026 is only 3-4 times, Kaileshi Technology is around 7 times, and the floor cleaning robot Ecovacs is only 1.4 times. All are lower than FANUC, and far lower than those players in the embodied intelligence track.
In terms of product form, the capital market understands commercial robots as "working tools," and the imagination space is severely compressed.
Commercial robot companies are generally concentrated in a certain type of scenario. For example, Geek+ focuses on logistics and warehousing, Gaussian Robotics focuses on the cleaning track, and UDI focuses on hotel robots. These scenarios face growth ceilings and scenario migration difficulties, so capital pricing is constrained in many places.
It is precisely for this reason that an extremely abnormal phenomenon has emerged in the secondary market: commercial robot companies have more mature business models, are closer to the market, have continuously rising revenue, and continuously narrowing losses, yet their valuations keep dropping.
Commercial robot companies are caught in a dilemma. If they explain commercialization too thoroughly, they will be classified into the valuation framework of "traditional manufacturing." If they do not talk about commercialization and closely follow the hot topic of embodied intelligence, they can obtain higher valuations and new financing.
A slightly absurd scene has emerged. Delivery, cleaning, and warehousing robot companies with good commercialization use more space in their press releases to tell stories about embodied intelligence. On the booths and in the press releases of these companies, most of them also display various humanoid robots.
Contrasting Fortunes
The contrasting fortunes in the robotics track are somewhat similar to the huge differences between the AI 1.0 and AI 2.0 eras.
The "Four AI Dragons," once surrounded by a halo, now show their decline, while AI large model enterprises shine brightly. Despite the shrinkage in market capitalization, the market capitalization of one Zhipu AI is still equal to that of six SenseTime or 28 CloudWalk.
Capital preferences always change with the occurrence of industrial transformation. Looking back through the rearview mirror of history, the primary market's fondness for robot enterprises has undergone a transition from commercial robots to humanoid robots, and then to smart brains, world models, and component enterprises.
For example, Keenon, a representative of commercial robots, completed the largest single financing in the track at that time in September 2021, with SoftBank, CICC Jiazi, and Saudi Aramco injecting 200 million USD into it. It was not until April 2025 that it received 500 million CNY in financial support from Taizhou state-owned capital. That is to say, there was no financing income for more than three years during this period.
The reason for starting a new round of financing is mainly due to the release of its first humanoid embodied service robot XMAN-R1, as well as the bipedal humanoid embodied service robot XMAN-F1.
Pudu Robotics, another benchmark enterprise in the industry, experienced a massive layoff in 2022 when ChatGPT emerged. The number of employees plummeted from a peak of 3,000 to about 500. CEO Zhang Tao emphasized in an all-staff letter: "For the company to survive, it essentially relies on operations to stay alive, not on financing."
Since then, its operational efficiency has significantly improved. In April this year, it once again obtained nearly 1 billion CNY in financing, with its valuation breaking through 10 billion CNY. These funds are mainly used for the R&D of embodied intelligence, forming a technical architecture of "one brain, multiple forms." It is conceivable that if the concept of embodied intelligence had not been integrated to a greater extent, even an industry-leading company like Pudu would probably be defined as a "shell" company.
A negative example might be Gaussian Robotics, which has had no movement since its financing in 2023, and its listing has also made no substantive progress.
On the other hand, embodied intelligence is also being spontaneously "cooled down" by the market because the concept is too hot.
Unitree Technology, whose fundamentals are not bad, saw its market capitalization shrink by over 270 billion CNY after listing. Mech-Mind had a 3,835 times oversubscription but encountered a first-day break in issue price. The "premium" given to embodied intelligence in the primary market was not "paid for" in the secondary market.
At present, more than 50 robot and embodied intelligence enterprises are crowded outside the door of the Hong Kong Stock Exchange. Facing market capitalization fluctuations and commercialization difficulties, Li Chao, Deputy Director of the Policy Research Office of the National Development and Reform Commission, emphasized: "The robot industry must adapt to local conditions and prevent blind following and rushing in all at once." Obviously, this represents the attitude of the regulatory authorities.
Subsequently, reports indicated that regulatory authorities have issued informal "window guidance" to raise the IPO approval threshold for humanoid robots. Enterprises planning to go public need to prove their revenue-generating capabilities and be on a track of narrowing losses, or truly possess hardcore technological innovation; otherwise, the listing channel will be narrowed.
Zhu Xiaohu, who advocates "deviating from consensus by 15 degrees," began to "exit humanoid robot companies in batches" in 2025. He believes that the commercialization potential of embodied intelligence or humanoid robot enterprises is still not visible at present.
In just half a year, the robot capital market has experienced a peculiar transformation. At the beginning of the year, capital afraid of missing out rushed in crazily. Now, under high focus, the focus of investors has shifted from technological leadership to commercial implementation.
To this day, commercial robot enterprises are eager to label themselves with "embodied intelligence" to survive the valley of death; while embodied intelligence enterprises have begun to be keen on demonstrating commercialization capabilities, or potential commercialization capabilities.
Capital preferences and industrial leaps are being staged in a somewhat blackly humorous way.
Final Words
"One day in embodied intelligence, one year in the human world."
From "unclear vision" to "cannot miss out" and then to "calm cooling down," from capital encouragement to policy support, and then to spontaneous market cooling, embodied intelligence has experienced complex changes in just over a year that might take years to appear in the history of venture capital, industrial history, and financial history.
From traditional service robots to embodied intelligence robots, and then to components such as world models and dexterous hands, the robot industry continues to emerge new investment opportunities amid the iteration of AI technology. This is an opportunity of the times.
Even Liang Wenfeng, the founder of DeepSeek, who is perhaps the entrepreneur with the most profound understanding of artificial intelligence, has begun to pay more attention to the future development of embodied intelligence and has started to select relevant partners on a small scale.
From these perspectives, the "investment and financing storm" of embodied intelligence is not a thorough bubble, but rather requires more breakthroughs in technology, commercialization, and other aspects to further move towards maturity.
Capital preferences are constantly changing, but the fundamental logic of business will not change. While being optimistic about the continuous development of embodied intelligence technology and actively placing bets, the market should also give more recognition and attention to commercial robot enterprises.