When robots start to "tell the truth", the intelligent agents adjust their narrative logic.
Recently, an unexpected turmoil has hit the embodied intelligence track.
Shao Tianlan, CEO of Mech-Mind who just led the company to complete its listing, publicly criticized on WeChat Moments a "gathering-type" embodied intelligence entrepreneurship prevailing in the current industry. The routine is to generate unsustainable revenue through various related-party transactions (such as "data collection centers" and "leasing companies"), and hype by releasing "big news" to achieve the goal of a quick impact on listing.
Subsequently, Shao Tianlan named and questioned Galaxy Universal in the comment section.
In response, Galaxy Universal stated that it "will not engage in verbal disputes and focus on product implementation", and chose to report to the police for other malicious rumors circulating online...
But that's not all. According to online news from media such as Yicai, leading embodied intelligence enterprises QX Intelligence, StarSeas Map, and StarDynamics Era may face a key major assessment this month, and recently a peer reportedly reported them to the "examiner", with the content of the report not yet made public.
On September 13, QX Intelligence responded that the online information is not true; the media contact person of StarSeas Map said that they are not aware of the relevant situation for the time being; StarDynamics Era said that they will not make an official response for the moment.
Although the news has not been further verified, as far as the current situation is concerned, apart from financing, the major assessment for enterprises like QX Intelligence may be the long-rumored IPO listing.
Coincidentally, under the background that regulators clearly emphasized last month that the robotics industry should "focus on real demands", a piece of news spread in the investment banking circle in recent days: the China Securities Regulatory Commission plans to tighten the IPO of humanoid robot enterprises through window guidance, and put forward requirements on recurring revenue, real innovation capability and other aspects...
Thus, from Shao Tianlan's "outburst" to the outflow of the report rumor, and then to the tightening of the regulatory side.
Under a series of impacts, not only many star embodied intelligence enterprises have begun to be demystified, but even the trend of the entire industry seems to show a sense that the "sky is about to change".
Although in the long run, this change is not a bad thing, and it is even more conducive to the robotics industry to reduce excess weight and enhance core competitiveness.
But for companies such as Agibot and StarSeas Map that are preparing for listing, the feeling may be completely different...
When robots start to tell the truth, is it harder for embodied intelligence enterprises to go public?
In fact, the related-party transactions exposed by Shao Tianlan, CEO of Mech-Mind, are nothing new.
As early as March this year, Liu Yang, co-founder of Force Infinite, said bluntly: In 2025, a large number of so-called "commercial orders" in the industry are essentially PR-oriented display procurement and data collection cooperation, rather than productivity substitution in the real sense.
At the same time, on the market side, in the past two years, we have also seen many news about UBTECH, Unitree and other enterprises winning bids for local humanoid robot data collection center projects.
For example, UBTECH won consecutive bids for projects such as the humanoid robot data collection and testing center in Fangchenggang, Guangxi last November, with a total amount of 264 million yuan, and the humanoid robot data collection and training center project in Jiujiang, Jiangxi, with an amount of 143 million yuan;
Another example is Agibot. In July 2025, Agibot Robot won the bid for the data collection and exhibition robot procurement project of Zhuhai Embodied Intelligence Application Innovation Center, with an amount of about 12.736 million yuan.
According to Fortune China disclosure, the purchaser of this project is Zhuhai Zhihui Yuanqi Technology Co., Ltd., a joint venture established by Agibot and the local government — Agibot holds 30% of the shares, and the local holding company with 40% shares is a shareholder of Agibot, and the two sides have deep equity connections.
In September 2025, Agibot Robot quoted 31.0161 million yuan and pre-winning the bid for the robot procurement project of Hubei Humanoid Robot Innovation Center.
Four months later, in January 2026, Agibot signed an agreement with the center in turn to purchase thousands of hours of training data...
Thus, a transaction chain emerged: local parties cooperate to set up a data collection center, then buy robots from the enterprise, and the enterprise spends money to buy data from the data collection center. Although the whole process does not go beyond the circle of stakeholders, enterprises like Agibot have realized a closed loop of revenue, and the contracts and delivery of each link are real.
Of course, even so, the existence of data collection centers has great industrial value.
After all, like large models, data is also the driving fuel for embodied intelligence. How much force the robot uses for grasping, what friction feedback the surface of the object has, this kind of information can only be generated in real operations and must be obtained through special collection.
Therefore, the gold content of this related-party transaction model may be debatable, but it cannot be regarded as false revenue.
The only and biggest problem is whether anyone in the whole chain pays money because "the robot can substantially do the work for me"...
In that case, Shao Tianlan just repeated the current situation of the embodied intelligence industry, but why did it still cause such a big impact?
The key lies in — the timing.
First, it hits the sensitive node of the sentiment contrast between the secondary market and the primary market of the embodied intelligence industry.
Why did the CEO of Mech-Mind "fire" from the perspective of the secondary market at this time?
Looking back, Mech-Mind was listed on September 1 with an issue price of HK$101.7, and the public offering part recorded an oversubscription of about 3835 times, setting a new record.
However, it broke on the first day of listing. The closing price on September 15 was HK$81 per share, a retracement of about 20.35% from the issue price.
Further back, Unitree Technology was listed on the Sci-Tech Innovation Board on August 19, and its opening price once rushed to 1100 yuan, but the stock price retraced by more than 50% from the high point in less than a month...
It can be seen that the "embodied intelligence premium" given by the primary market in the past has generally not been realized in the secondary market, even for strong players like Unitree Technology.
This is actually the most concerned issue in the current embodied intelligence investment market.
So at this time, Shao Tianlan only needs to open a new direction, and the market sentiment will spread quickly:
The valuation system of the whole industry has been deviated by the bubble premium created by the "gatherers".
It is not that Unitree and Mech-Mind are not worth the price, but someone "inflated" the expectation. So his first sentence is: The secondary market is not a "venture capital market", and should not be primaryized...
Thus, once the capital market forms a consensus, is it possible to change the valuation anchor? On one side is the revenue that can stand the test of the market, and on the other side is the volume expanded by related-party transactions. The two naturally cannot be valued on the same dimension.
Of course, for the industry, this is also the beginning of bringing order out of chaos.
If everyone relies on data collection centers across the country to expand revenue instead of implementing commercialization through real technological innovation, the entire embodied intelligence industry may move towards the outcome where bad money drives out good.
After all, by relying on storytelling and related-party transactions, others can achieve multiple rounds of financing and double their valuation in a year. If you honestly do R&D, the financing rhythm may fall behind, and the talent and supply chain cannot keep up. At this time, you will either fall behind or have to take the same path...
Second, it hits the special period when embodied intelligence enterprises are gathering to go public.
As of August this year, at least 28 domestic embodied intelligence robot enterprises are promoting or planning to go public, including 7 that have been officially accepted by the stock exchange, 5 that have entered the counseling filing or submission stage, and 16 that are in the pre-IPO stage.
For this reason, after the continuous decline of Unitree's listing and the public questioning of gathering-type embodied intelligence enterprises, the overall situation will inevitably further promote the tightening of the regulatory authorities, thus raising the listing access threshold.
This change will not only put pressure on second-tier players such as Galaxy Universal, but also pose a considerable challenge to Agibot, which has the gold signboard of "the largest shipment volume of humanoid robots".
Although Agibot's shipment volume is higher than that of Unitree Technology and its revenue maintains high growth, after that, the capital market may also pay attention to the gold content of shipment volume, as well as the revenue structure and actual industrial value.
Of course, in theory, this is also a value opportunity for the "latecomers" of embodied intelligence to surpass Unitree Technology, but the key depends on whether enterprises like Agibot can seize it...
Embodied intelligence needs to change its narrative logic
In fact, looking at the reality, the focus of the entire embodied intelligence industry has changed now:
From focusing on dance performances in the past, it has actively shifted to competing for industrial implementation.
For example, according to data from Tianyancha and the prospectus, in the first three quarters of 2025, among the revenue of Unitree's humanoid robots, scientific research and education accounted for about 73.6%, commercial consumption accounted for 17.39%, and the industry application that was really expected only accounted for about 9.01%.
Now, Unitree has also begun to increase its efforts in implementing applications in the industrial field. Especially recently, Unitree Robot Benben has joined Li Auto, and its available positions include entering vehicles, entering stores, and entering factories...
For example, UBTECH Walker S1 is now also undertaking quality inspection and assembly tasks in automobile factories such as BYD.
In addition, even the recent World Robot Games have added practical projects such as unpacking, industrial loading and unloading, and packing and warehousing. Among them, Agibot Robot won 18 gold medals, 16 silver medals and 12 bronze medals, ranking first in both lists.
These are all good trends, but unfortunately there is still a certain distance from full implementation.
Previously, Agibot Robot held a live broadcast in the factory to verify whether the robot can "work stably, in batches and sustainably".
As a result, an accident occurred in the middle: a robot as the communication master node sent a request to the production line belt, but the belt failed to respond in time, resulting in the previous plate not being transported away. When the robot was discharging according to the predetermined procedure, it encountered insufficient space, and finally a material stacking abnormality occurred.
However, hard-core technology cannot be achieved overnight. Technological progress is a curve, and the entry of robots into factories also requires accumulation and polishing.
Therefore, after the window paper of the gathering-type embodied intelligence industry is pierced, the top priority for enterprises like Agibot who are waiting for listing is to change a set of narrative logic to the investment market.
After all, with the existence of related-party transactions such as data collection centers, the benefits of simply talking about the first shipment volume and revenue scale will continue to decline in the future. Coupled with the sharp retracement of Unitree after listing, the primary market may no longer hype story premiums.
Therefore, enterprises like Agibot need to take a pragmatic PR/marketing route next.
First, in terms of narrative focus, talk less about the nominal first place, and talk more about what you have actually done.
Just like the first place in the gold medal list that Agibot won at the Games before, many outsiders only thought that Agibot robots run faster and jump higher, and hardly know what Agibot has achieved in technology and what it can do.
Moreover, since Agibot robots are so versatile, which market-oriented customers have they been sold to?
These are the focus of publicity for enterprises like Agibot next — even if the scale of cooperative revenue is limited, it can prove that Agibot robots have real industrial value.
In this regard, the capital market will most likely not be stingy with support.
The adjacent AI market is the most direct example. In 2025, the revenue of Zhipu AI and Minimax were 724.3 million yuan and 555.5 million yuan respectively, and the corresponding attributable net profits were -4.698 billion yuan and -13.16 billion yuan.
Even so, the market value of the former once exceeded the trillion mark, and the latter, even after continuous retracement, now has a market value of more than 80 billion yuan.
The core point is that companies like Zhipu AI disclose ARR revenue as a reference almost every month. Although it is not large, they let the market truly see the future path of AI commercialization...
Second, in terms of marketing, the trust endorsement of product technology should also change its anchor in time.
Mainly under Shao Tianlan's "outburst", the local winning bids of the embodied intelligence industry and the performance credibility of brand companies may be questioned by investors in the next step.
Therefore, at this time, enterprises like Agibot need to get close to market-oriented third-party customers, and reset investment confidence through factory entry cooperation with CATL, BYD, JD Logistics and other enterprises.
And release data such as customer repurchase situation and cooperation progress in time.
All of these will eventually become a key part of opening the imagination of future listing valuation for embodied intelligence enterprises...
In short, the imagination of the embodied intelligence industry is unlimited, but the pace needs to be slower and steadier.
Just like the new energy vehicles that were once questioned at the beginning, but now have achieved subversion of fuel vehicles, the bubble in the robot market has receded, which is not a cold winter, but a critical moment for real technology, real products and real scenarios to surface.
We believe that when robots are truly implemented in thousands of industries, all the calm today will become the footnote of tomorrow's outbreak.
We look forward to Agibot, Unitree and Mech-Mind achieving their goals at an early date!
Statement: The information cited in this article is all derived from the statutory disclosure documents of the company and public materials. The author does not make any guarantee for its completeness and timeliness. Stock market investment is risky, please be cautious. The content of the article is only a comment and does not constitute any investment advice. If you choose to participate in the investment, please make your own judgment and bear your own risks.