Shortly after the company went public, the founder of Mech-Mind blew the lid off the industry's long-hidden insider secrets, explicitly naming the "Four Little Dragons" and Galaxy Universal.
In 2026, the embodied intelligence industry is seeing a dramatic rise in popularity.
In the primary market, massive hot money has flooded into the embodied intelligence industry. According to data from IT Juzi, the financing amount of the embodied intelligence track in the first half of this year has reached nearly 935 billion yuan, while the total financing amount of the entire industry in 2024 was just over 100 billion yuan. With a large amount of investment pouring in, a large number of start-ups have emerged in the industry, among which 8 unicorns have a valuation of over 200 billion yuan.
However, in the secondary market, the embodied intelligence concept presents a completely different picture.
Unitree Robotics (688836.SH) is one of the most well-known targets in the current embodied intelligence industry. On the first day of its listing in August this year, the share price of Unitree Robotics hit a peak of 1,100 yuan, but then the share price continued to fall. By September 10, the share price had fallen below the 500 yuan mark, and the total market value had evaporated by about 2.4 trillion yuan from the peak. Similarly, UBTECH (9880.HK), the "first stock of humanoid robots" in the Hong Kong stock market, has seen its latest share price halved from its highest point within the year.
The track presents a "two extremes of ice and fire" in terms of popularity. An insider in the embodied intelligence industry publicly reflected on the current situation of the industry, which has aroused widespread market attention.
Recently, a screenshot of the Moments post by Shao Tianlan, founder of Mech-Mind Robotics (9615.HK), has been circulating on social platforms. In the screenshot, the founder who graduated from Tsinghua University said frankly that there are many "gathering-type" embodied intelligence companies now, including some well-known, high-valuation companies located in Beijing and Shanghai that have even appeared on the Spring Festival Gala. These companies use "data collection centers" and related party transactions with local governments, investors, and suppliers to generate unsustainable revenue.
Shao Tianlan pointed out frankly that this approach is not only an attempt to get listed quickly, but also "digging a hole" for themselves. In the comment section of the Moments post, Shao Tianlan even directly named the well-known embodied intelligence company Galaxy Universal Robotics.
Recently, discussions about the actual commercialization of embodied intelligence companies have emerged in an endless stream. One of the reasons why many embodied intelligence enterprises have suffered a "Waterloo" in their share prices in the secondary market is that the market has doubts about their future commercial prospects.
Taking Unitree Robotics as an example, in the first three quarters of 2025, the total revenue of Unitree Robotics from humanoid robots reached 595 million yuan, of which 73.6% came from the scientific research and education sector, that is, for scientific research, secondary development, etc.; 17.39% came from the commercial consumption sector, that is, for performances, exhibitions and other purposes; only 9.01% of the revenue came from industry application fields such as enterprise tour guidance, intelligent inspection, and intelligent manufacturing.
The commercialization status of other embodied intelligence companies is also relatively similar. Although some products have appeared in scenarios such as logistics and industry, their proportion in the total revenue is not high.
It is worth noting that Shao Tianlan's discussion on the current situation of the embodied intelligence industry came at a time when the news that "the regulatory authority has raised the IPO threshold for humanoid robot enterprises" is circulating. The news shows that the approval threshold for IPOs of humanoid robot enterprises will be raised, and enterprises need to prove that they can generate recurring revenue, are gradually reducing losses, or have achieved real innovation. However, this news has not been confirmed yet.
Mech-Mind Robotics founded by Shao Tianlan is one of the embodied intelligence companies listed recently. On September 1, Mech-Mind Robotics was listed on the Hong Kong Stock Exchange, becoming the "first stock of embodied intelligence vision, brain and hand". The company's issue price was HK$101.7, but its share price fell below the issue price on the first day of listing. After that, the company's share price fell to a minimum of HK$75.4 per share, down more than a quarter from the issue price.
This week, the share price of Mech-Mind Robotics has rebounded. As of the close on September 10, the company's share price was HK$89.45 per share, with a total market value of about HK$11.2 billion.
The Reality and Illusion of Data Collection Revenue
Long before this Moments post, Shao Tianlan had already raised objections to the practice of embodied intelligence companies obtaining revenue through the data collection center channel.
In an interview in early September, Shao Tianlan said that some embodied intelligence companies generate revenue through untrue or unsustainable means. For example, local governments invest in the construction of data collection centers, purchase equipment from embodied intelligence companies, and the companies then buy back the data to form book revenue.
Shao Tianlan emphasized that Mech-Mind Robotics has no revenue from data collection centers, and the company's revenue mainly comes from leading manufacturing and logistics customers.
In contrast, many embodied intelligence companies have cooperative relationships with data collection centers. According to a report by the third-party data platform Interact Analysis, as of the end of April this year, at least 90 humanoid robot data collection centers in China have been put into use or are under planning and construction.
Among them, at least 30 data collection centers use humanoid robots from Agibot. Leju Smart, Unitree Robotics and UBTECH have also participated in the construction of more than 10 humanoid robot data collection centers. Humanoid robots from manufacturers such as Galaxy Universal, Realman, Starsea Diagram, and Fourier Intelligence have also become embodied intelligent data collection terminals in multiple data collection centers.
Although the way of obtaining revenue through data collection centers is controversial, it is undeniable that data collection centers are very necessary for the training of the "brain" of embodied intelligence. Different from large language models, embodied intelligence large models do not have massive texts as training data. To promote the training of embodied intelligence large models, it is necessary to use methods such as real machine data collection and simulation data synthesis to generate training data.
The data collection center is exactly an important source of data collection. Similarly, in the era of rapid development of the autonomous driving track, autonomous driving data collection centers are also one of the sources for collecting data related to autonomous driving.
Yao Maoqing, CEO of Mifeng Technology under Agibot, previously stated that "before embodied intelligence is truly commercialized on a large scale, data as infrastructure will generate commercial returns earlier than end-point applications."
Of course, compared with autonomous driving, embodied intelligence products, especially humanoid robot products, are applied in more scenarios and require more abundant types of data. The data collected in the laboratory is too "clean", and such data is difficult to match the actual working environment of embodied intelligence products.
Some embodied intelligence companies have begun to adjust the data collection mode, choosing to deploy robots in real working environments, or directly adopt the "no-ontology data" mode for data collection. In this context, whether embodied intelligence enterprises can continue to generate incremental revenue through the data collection center channel in the future remains to be answered by the market.
The Question of Commercialization
Both Shao Tianlan, founder of Mech-Mind Robotics, and Wang He, founder of Galaxy Universal, are graduates of Tsinghua University.
Public information shows that Mech-Mind Robotics was founded by Shao Tianlan, Fu Ao, and Ding Youshuang, three alumni of Tsinghua University. Shao Tianlan, 37 years old, graduated from the School of Software of Tsinghua University, and later obtained a master's degree from the University of Munich. As early as 2016, Shao Tianlan returned to China and founded Mech-Mind Robotics, which has been ten years up to now.
The founders of Galaxy Universal are Wang He and Yao Tengzhou. Wang He, 34 years old, graduated from the Department of Electronic Engineering of Tsinghua University, and later received his doctorate from Stanford University in the United States. He is currently an assistant professor and doctoral supervisor at the Center for Frontiers of Computing, Peking University. In May 2023, Wang He co-founded Galaxy Universal with industrial robot expert Yao Tengzhou. Wang He is mainly responsible for software algorithm related work.
Although they are both in the embodied intelligence industry, the commercialization models of Mech-Mind Robotics and Galaxy Universal are not completely the same.
Mech-Mind Robotics does not operate complete embodied intelligent machines, but provides robot vision, dexterous hands, large models and other related software and hardware products. The company's customers mainly include automobile manufacturers, industrial manufacturers, logistics and warehousing operators, as well as robot manufacturers, etc.
In recent years, the revenue scale of Mech-Mind Robotics has continued to increase, and losses have narrowed overall. From 2023 to 2025, the revenue of Mech-Mind Robotics was 181 million yuan, 269 million yuan, and 389 million yuan respectively; the adjusted net loss was 334 million yuan, 214 million yuan, and 109 million yuan respectively. In the first quarter of this year, the company's revenue was 106 million yuan, a year-on-year increase of 73.09%; the adjusted net loss was 33 million yuan, narrowing by 14.57% year-on-year.
In contrast, the R&D direction of Galaxy Universal is more focused on large embodied intelligence models. According to the company's official website, its self-developed "Galaxy Star Brain" is an end-to-end large embodied model that can realize the integration of "brain - cerebellum - neural control". In addition, Galaxy Universal also has a number of complete embodied intelligent machine products such as the wheeled robot Galbot G1/S1 and the humanoid robot Galbot ET1.
Although some products of Galaxy Universal are used in data collection centers, the company is also expanding application scenarios such as industry and home. Leading large manufacturing enterprises such as CATL, Bosch from Germany, Toyota Motor, BAIC Group, and SAIC Motor are downstream customers of the company. In addition, the company's humanoid robot Galbot ET1 was also launched for commercial and home scenarios this month. According to the company's data, within 24 hours after Galbot ET1 opened for reservation, the number of reservations exceeded 200 units.
Galaxy Universal will also embark on its own capital market journey. Recently, there is news that the company has secretly submitted its listing application to the Hong Kong Stock Exchange.
In his Moments post, Shao Tianlan publicly asked, "Who dares to support such a company to get listed?" But for enterprises in this track, landing on the capital market requires multiple inspections by regulatory authorities, stock exchanges and investors. Whether Galaxy Universal can land on the capital market in the future and win the favor of investors remains to be answered by the regulatory authorities and the market together.
This article is from the WeChat official account "Damo Finance" (ID: damofinance), written by Damo Finance, and authorized for release by 36Kr.