The 22-billion-yuan star robotics project remains unsold even when offered at a 50% discount.
Robot pre-IPO shares that were frantically snapped up for purchase last year are now hard to sell even at a 50% discount this year.
It is heard that some investors are offloading their robot pre-IPO shares recently, with the price cut to around 50% of the original level.
According to media reports, an investor invested in a world model startup at a post-money valuation of RMB 3 billion earlier this year. By October, the company's public valuation had risen to RMB 22 billion. He tried to sell his shares at a valuation of RMB 12 billion, which is half the price the company itself claimed, but no one has been willing to take over so far.
A little over a year ago, the same assets had to be scrambled for.
In August 2025, fund shares holding stakes in leading humanoid robot companies directly released by sellers appeared on social media message boards, with the price calculated based on a valuation of about RMB 39 billion. During the same period, shares of Agibot Robot were also put on the market at a valuation of about RMB 25 billion.
Looking back, buyers seeking shares of Agibot and Starbot Era would open with investment quotas of tens of millions of yuan. At the Private Equity Fund Liquidity Summit held in Shanghai in January this year, attendees were still talking about pre-IPO shares of popular companies such as humanoid robots and commercial aerospace, which drove the entire S-trading market to boom.
In less than a year, buying spree has turned into fire sale.
In addition to the above situation, differentiation has already emerged in the embodied intelligence circle in October. StarDust Intelligence announced in the morning of the 13th that it had completed a RMB 1 billion financing at a valuation of RMB 15 billion, with Donghai Investment, an investment firm under the Shenzhen Investment Holdings system, leading the investment, and existing shareholders also continued to increase their stakes. Six days later, on October 19, the first creditors' meeting of INHE Robot will be held at the Shenzhen Intermediate People's Court. Creditors need to count a batch of unsold equipment from failed auctions, a string of frozen equity of subsidiaries, as well as payment for goods and wages in arrears for many years. While some new capital enters the market, some players exit. The pre-IPO share market, caught in between, is the first place where real prices appear.
Has the bubble burst? Liang Dongjian, Partner of China Core Capital, believes that the liquidation has begun, but the tide has not fully receded. Instead, the water level has been redistributed: the head players are still riding the rising tide, while the trailing players have already run aground.
Financing can use new funds to support valuation, but pre-IPO shares can only be sold when someone is willing to make an offer. A year ago, a demonstration video on the Spring Festival Gala stage could make a room full of investors queue up for investment. Now, investors have begun to calculate when and at what price the shares in their hands can be converted into cash.
Pre-IPO shares sold at 50% discount, still no buyers
The turning point can almost be pinpointed to a specific day. In early September, U.S. tech media The Information cited people familiar with the matter as saying that regulators have given informal "window guidance" to some investment banks and investment institutions: humanoid robot companies that want to go public must first prove that they have recurring revenue, narrowing losses, or truly important technological innovations.
After the news was released, CLS confirmed with a number of investment bank practitioners, and got more detailed replies: revenue must be sustainable, real and replicable, cannot rely on subsidies, related party transactions, one-off income or "data collection centers" for support; losses must continue to shrink, and there must be large-scale shipments and stable customers.
Some brokerages also said that they have not received the notice of window guidance, and what they have done is more to urge sponsor institutions to strengthen project review. The statements are different, but the views in the market are consistent.
At present, quite a number of robot companies are queuing up for listing. According to incomplete statistics, more than 20 core domestic humanoid robot enterprises are undergoing IPO counseling, declaration or issuance, and 46 robot-related enterprises are queuing up for listing on the Hong Kong Stock Exchange.
As listing becomes more and more distant, early investors are the first to lose their composure. Money in the primary market generally exits through IPO. If the exit period is extended, the valuation will decline, and the company will fall behind in the competition. All three situations will happen, so some people choose to cash in their returns before risks emerge. Most of these sellers entered the market last year or early this year, when the valuation was not too outrageous. For example, the investor mentioned earlier bought in at RMB 3 billion and sold at RMB 12 billion. Even at a 50% discount, the book profit is as high as 4 times.
The difficulty lies in the lack of buyers. In July, about 34 companies in the domestic embodied intelligent robot track were raising funds, with the disclosed amount of about RMB 120 billion, down about 30% from RMB 17.8 billion of the previous month. The share price of Unitree has been falling since its listing, and more people have chosen to wait and see. Some investors said that institutions are no longer optimistic about embodied intelligence projects, and no project has passed the review in the past month. A FA put it more bluntly: they have not signed new embodied projects for a long time. The landing promises made by enterprises at the beginning have not been fulfilled so far, and only the rankings and awards are updated in each financing round.
Once the pre-IPO share price is put on the market, it will in turn put pressure on the company. There are already pre-IPO shares with a valuation of RMB 12 billion on sale in the market, and it is difficult for any investor to increase capital at a valuation of RMB 22 billion.
A Shenzhen-based embodied intelligence company with a valuation of RMB 4 billion originally planned to raise a new round of financing at a valuation of RMB 8 billion. In the end, the valuation did not drop on the surface, but new shares were sold together with a part of pre-IPO shares, and the actual valuation was pushed down to about RMB 6 billion. Some investors also revealed that in order to restrict the outflow of pre-IPO shares, many embodied intelligence companies now simply do not want to promote share reform.
The 50% discount figure is not necessarily a random guess. Reuters cited an estimate from a person close to investors that if the revenue related to data collection centers is excluded, the valuation of some robot companies may drop by 60% to 70%. The ruler in the mind of pre-IPO share buyers probably measures this part of the valuation bubble.
The ones who are really in an awkward situation are the group of people who entered the market at a high price around the middle of this year. They bought in at the time when the valuation doubled the fastest: ZhiSquare took 126 days to go from RMB 10 billion to RMB 20 billion, and Qianxun Intelligence only took 33 days. Many investors reminded that many embodied companies still have unspent financing from before, and the real financing pressure may be concentrated after November.
The same card table, two different outcomes
Pre-IPO shares are a mirror, reflecting two completely different fates in the same track.
The story of INHE has been covered in previous reports by China Venture. In March 2021, it signed an agreement with Shunde to acquire 500 mu of land in Beijiao Robot Town, with a total investment of RMB 10 billion, aiming to realize "robots manufacturing robots". This was the first RMB 10 billion-level investment promotion project in Shunde that year, and the negotiation and signing took less than a month. Its founder Shen Gang once served as Director and Executive Deputy General Manager of Shanghai FANUC, and helped Country Garden establish Botai. Panda Capital almost led the RMB 70 million angel round and the nearly RMB 400 million Series A round entirely because of his resume. Meituan Strategic Investment, ByteDance Strategic Investment, and GSR Ventures also joined in successively. Less than ten months after its establishment, INHE was valued at USD 500 million, and its controlling shareholder is He Jianfeng's Efeng Group.
An impressive resume cannot bring in orders. Li Lun, Founding Partner of Panda Capital, publicly stated in June this year that without stable cash flow, the company paid its core executives an annual salary of more than RMB 3 million, and spent RMB 28 million on office decoration, but Efeng never responded to the matter. At present, the legal representative of INHE is restricted from high consumption, its official website cannot be accessed, and it is said that its total revenue last year was only about RMB 4.3 million.
At the end of the summer of the same year, Shanghai's Dark Blue Robot and Suzhou's Orange Dynamics were also on the verge of bankruptcy. Dark Blue Robot is a subsidiary of AI unicorn Dark Blue Technology, which once claimed to have received an order for 25,000 disinfection robots from a Fortune 500 company. By September, the bankruptcy administrator had confirmed 104 employees and employee claims of about RMB 23.475 million. Orange Dynamics made swimming pool cleaning robots, and was invested by Shunwei Capital, Clear Water Bay Fund, Zhixing No.1, and Ecovacs, with its products sold to Europe and the United States. Earlier, CloudMinds, which raised a total of about RMB 54 billion in financing and had a peak valuation of RMB 20 billion, was accepted by the court for bankruptcy liquidation in October 2025.
The situation on the other side of the card table is completely different. Unitree's operating revenue in the first half of this year was RMB 1.152 billion, with a net profit of RMB 274 million, and it successfully went public in August. Agibot's revenue grew from RMB 300,000 in 2023 to RMB 1.05 billion in 2025. Galaxy General Intelligence has received about RMB 7 billion in investment since its establishment three years ago, with a valuation of over RMB 20 billion. Hundreds of its robots are deployed in CATL's battery factories, and CATL is not only the lead investor of its Series B round, but also its largest customer.
If we arrange the failed companies in chronological order of establishment: CloudMinds was founded in 2015, Dark Blue Robot in 2018, INHE in 2020, and Orange Dynamics in 2021. Galaxy General Intelligence and Agibot, which are standing in the center of the card table, were both established after 2023. The time in between is the winter when ChatGPT was born. The companies being liquidated this time are mainly the generation of service robot companies that existed before the emergence of large models. They told stories of scene expansion, and most of their orders were demonstration and intention ones. Now they are paying off the debts owed in the last cycle.
Capital is concentrated in the hands of a few players. According to data from IT Juzi, the total financing of the embodied intelligence track in the first half of this year exceeded RMB 934 billion, a year-on-year increase of 350%. The top 20 enterprises raised a total of RMB 55 billion, accounting for 59% of the entire market, 17 of which are located in Beijing, Shanghai and Shenzhen.
Averages can also be misleading. After excluding transactions of more than RMB 1 billion, the average value of a single RMB-denominated financing this year is only RMB 212 million, slightly lower than RMB 222 million last year. Liang Dongjian feels that the transaction volume of mid-tier projects this year has dropped by nearly half compared with last year. What are being liquidated are mainly some mid-to-low tier humanoid robot whole machine projects, pure conceptual brain projects and data collection projects, while the "water seller" links such as dexterous hands, six-dimensional force sensors, integrated joints, and end-side chips are heating up.
The questions raised by investors have also changed. Lai Jie, founder of StarDust Intelligence, recalled that during the financing round in 2024, many investors said they would not invest in projects without end-to-end capabilities. In 2025, they said they would not invest in projects that do not develop VLA. This year, more and more people have come to ask about the progress of commercialization. Hearing this question, he thought to himself that investors are finally starting to pay attention to the right things.
There is no unified standard for calculating commercialization, and different companies have different evaluation rules. He Chen, General Manager of AVIC Chuangshi, said in an interview with Rongzhong Finance that the shipment volume piled up by policies, pilots and subsidies is all illusory. Only when the cost of robots completing standardized tasks is far lower than that of human labor, and customers actively repurchase out of operational needs, can this industry truly cross the commercialization threshold. Li Kewei, Chief Marketing Officer of Qiantianzu, focuses on the payback period. Since hardware is expensive and technology iteration is fast, customers are afraid of depreciation after buying robots, so Qiantianzu simply leases robots out, equipped with operation and maintenance services and insurance. Its valuation reached RMB 7 billion only half a year after its launch, with more than 4,000 schedulable robots on the platform, and part of its production capacity is extra units manufactured by ontology manufacturers to boost shipment volume.
Listing is not an escape pod
Pre-IPO share discount is not a new thing in the primary market. The money brought in by a new round of financing usually comes with a full set of protection clauses including repurchase, preferred liquidation, and anti-dilution. Most of these rights cannot be transferred when pre-IPO shares are resold. In addition, due to poor liquidity, it is normal for buyers to press down the price a little. When the market is good, this discount will not be too large, and there are even people queuing up to buy pre-IPO shares of popular projects.
In the first half of this year, it was common for embodied intelligence companies to launch two financing rounds at the same time with insufficient shares for distribution. The marginal buyer who bids the highest and is most afraid of missing out determines the valuation of the whole company. Some investors estimate that among the about 50 robot companies that have entered the listing process, only about 15 to 20 can finally get listed, and no less than half of them will have a listing price lower than their last round of primary market valuation. What pre-IPO share sellers are actually calculating is a time account: for the same 4x return, getting it in two years and getting it in five years make a big difference to the internal rate of return of the fund.
RMB funds generally have a "5+2" duration. Funds that entered the market last year were originally scheduled to start distributing profits to LPs in a few years. Once the listing schedule is pushed back, no matter how good the book return is, it may not be able to last until the exit day. In the past two years, the most frequently asked question from LPs is DPI. GPs would rather earn a little less now than get part of the money into their pockets first. From this perspective, the 50% discount does not necessarily cut down the value of the company, but the waiting time and the uncertainty during this period.
Not everyone can sell their shares. In this round of embodied intelligence financing, state-owned capital and industrial capital account for a large proportion. Among large transactions of more than RMB 100 million in the first half of this year, state-owned institutional participation rate is as high as 42%. The transfer of state-owned equity needs to go through procedures such as evaluation and on-site trading, which is very cumbersome, so it is difficult to sell at a discount for early cashing out. Industrial capital invests in the market for supply chains and scenarios, and is not in a hurry to exit.
What can flexibly enter and exit the pre-IPO share market are mainly market-oriented VCs. The selling pressure is concentrated on these players, which brings changes: the financial investors who first bet on this track exit first, leaving the most patient capital that cares least about price. When these companies are revalued later, the logic of national strategic allocation will carry more weight, while the logic of financial return will carry less weight.
Then who will take over the shares? Theoretically, S funds are natural buyers. Earlier this year, they were still chasing popular robot shares. But the pricing of S funds ultimately depends on exit. Only when the listing path is clear can they dare to give a price. If the path is unclear, they can only use a deeper discount to get a safety cushion. Many S funds with local state-owned background have emerged in the past two years, with low capital cost, and may take over part of the shares, but the price they are willing to offer is most likely not satisfactory to sellers.
The discount will eventually be transmitted back to the company. The gap between primary market valuation and pre-IPO share price will not exist forever. Either the next round of financing will lower the valuation honestly, or the revenue will support the valuation. The former hurts the company's reputation, while the latter takes time, and the cash on the books of many companies may not be enough to wait that long.
On October 19, INHE's creditors will gather at the Shenzhen Intermediate People's Court to count the remaining assets. StarDust Intelligence is still planning how to spend the RMB 1 billion that has just arrived. The investor who wants to sell his pre-IPO shares at a valuation of RMB 12 billion is probably still waiting for a callback. The three things go in different directions, but