Sorry, we no longer invest in embodied intelligence.

Author | QIU Xiaofen

Editor | YUAN Silai

In the embodied intelligence track, the somewhat reckless investor has started to hold back his moves.

Few people would question his status. He is on the list of early investors of a listed robotics company, and also the behind-the-scenes backer of another leading embodied intelligence enterprise. For many years, he has kept a cautious distance from the public eye, but in the primary market, he is renowned for his sharp and aggressive investment style.

"His judgments are very bold, he skips all the messy due diligence work, directly decides to invest in you, and the capital arrives in the account within a week," an investor familiar with him told 36Kr.

Now, this man who rarely hesitates has suddenly turned cold. He would tell some founders, "This project needs to be presented to the board of directors for review", and then there will be no further substantive feedback.

For entrepreneurs in the embodied intelligence sector, everything suddenly began to fall apart in the autumn of 2026. From the already chilly Beijing to the still midsummer-like Shenzhen, the entire embodied intelligence market is immersed in the same cold wave.

Three leading institutions that were once extremely enthusiastic about embodied intelligence have tightened their purse strings. An investor from one of the institutions made it clear in a meeting: "tighten investments in embodied intelligence" and "make judgments on embodied projects with prudence", while another institution has barely brought any embodied projects to the investment committee for review in the past one and a half months.

The capital market is realistic and ruthless. An embodied intelligence entrepreneur in Shenzhen was still pushing forward his financing normally before September, but when he approached FA again after September, he was told that they no longer take on embodied projects — he became an unwelcome party in less than half a month.

He can hardly blame the FA. A leading FA told 36Kr that the number of embodied projects he received recently "has decreased by half". This is not necessarily a bad thing, because embodied projects are simply unsellable now. "Projects that could be sold out in one month in the past are all stuck now, and it is estimated that it will take no less than three to six months to get things done." Eventually, they have to raise the threshold for accepting new projects.

Embodied intelligence has turned from the "highest priority project" in FA's eyes to an unsalable product in a very short period of time. Sometimes, entrepreneurs will even be questioned by them: "You missed the best timing when the market was hottest a few months ago. Can I assume that you don't have a good enough market sense?"

In fact, a FA said privately that he mostly feels sorry for the situation. He mentioned "timing" over and over again. "If that company had caught the right timing, it could have closed the deal very quickly at a very good price," he told 36Kr. The failure of that company is only because everyone's "decision-making cost is too high now".

No one can point out exactly which domino tile started the fall first. But the timeline is clear: many industry insiders told 36Kr that the turning point came in September this year.

According to statistics from industry service platform NE Times, the financing amount in the embodied intelligence sector in September dropped by 40% compared with August.

Image from Visual China

Many events broke out intensively in just a few weeks.

As the weathervane of the industry, Unitree's market value once halved after its listing. Then the CEO of a listed embodied intelligence company publicly pointed out on WeChat Moments that some companies had falsified their revenue. Rumors that "IPOs of multiple embodied companies have been suspended" began to spread privately. Since then, several leading companies "immediately got their operations out of shape and became extremely nervous".

The craze for aggressive expansion was cut off so suddenly that it seemed almost surreal.

A few months ago, an embodied intelligence founder proudly told 36Kr about their valuation growth curve: in February this year, the company's valuation was only 200 million yuan; by July, the valuation had topped over 3 billion yuan, surging 15 times in just 5 months. Some companies fought fiercely in public relations battles to be the first to break the 20 billion yuan valuation mark, while other leading companies attacked each other behind the scenes to compete for listing slots.

When capital retreats from the industry, all these struggles even seem absurd. Over the past three years, the embodied intelligence track has spawned hundreds of startups, and produced many star enterprises with valuations of tens of billions of yuan. Now, most of them are in a precarious, unstable state.

An investor told 36Kr that some of his peers have stopped paying attention to embodied intelligence, while some optimistic ones are holding cash and waiting — they all believe that the story is far from over.

He made a judgment that we are currently in the late stage of this cycle of the embodied intelligence sector. The industry is bound to go through a period of shock before it can enter a relatively stable stage. He predicted that this point may not come until the second half of 2027, or even the first half of 2028.

"Everyone is waiting for the second half of the game, waiting together."

Building Unicorns

The hustle and bustle of the embodied industry is inherently full of strong speculative atmosphere. Our research found that many of the so-called star projects with dazzling valuations are mainly driven by behind-the-scenes operators.

In the past few years, a "brand new" startup financing model has been popular in the industry, which some investors call "gathering-game entrepreneurship".

The way this model works is that certain "organizers" with resources first target people with specific profiles, such as genius technical scientists, senior executives from leading tech companies with rich industry experience, and then persuade them to quit their jobs and become founders. These "organizers" are usually investment institutions that may not have outstanding brand awareness, but hold abundant local resources and capital networks.

When these people decide to start a business, the organizers will come up with a complete set of plans, with only one purpose: to push up the valuation and pass the hot potato from hand to hand.

First of all, they will help the founders package themselves and build momentum. They will highlight labels such as big company background, technical genius, and successful entrepreneurial experience, and then push the project into the vision of the primary market.

These people are all masters at taking advantage of human nature. A FA told 36Kr that for a period of time in the past, they only gave investors two days to make investment decisions. They would tell investors: "You must give a decision by x o'clock tomorrow evening." As for more project materials and founder interviews, they are not available. "If you don't invest, the next person is waiting to invest."

If the other party is still hesitating, they will push further: "Because we have a good relationship, I'll tell you that institution X wants to invest, and institution Y also wants to invest. If you don't take it, I'll give the opportunity to someone else."

Under this kind of extreme negotiation, some inexperienced investors are easy to lose their rationality. "The whole market is full of FOMO (fear of missing out) sentiment."

Compared with creating a tense atmosphere, what is more important is to arrange the financing rhythm in advance. The aforementioned investor said that some "organizers" will introduce leading US dollar funds in the first few rounds, and try to push these institutions to continue to increase their investments in subsequent financing rounds. The continuous increase in investment from leading institutions has become an important bargaining chip to attract other investors to join the game.

In addition to providing advice on investment, the "organizers" also provide a lot of practical strategies for the company, such as how to get orders, how to "create" revenue, and so on.

In this carefully designed layout, the company's valuation is pushed up step by step according to the preset path. The aforementioned investor said that for similar projects, the valuation of the first round usually starts at around 100 million US dollars, the second round generally doubles to 1.5 billion yuan, and the third round ranges from 2.5 billion to 4 billion yuan. Moreover, some particularly marketable projects will also experience over-subscription, " sometimes the valuation of the third round jumps to 5 billion yuan or more immediately".

This is a precisely running machine. Investment institutions do not necessarily need to wait for a robotics company to truly complete commercialization, and they can obtain book returns through the valuation appreciation brought by subsequent financing. Therefore, in this valuation game, the technical teams, investment institutions and FAs have a tacit understanding to push the snowball to grow bigger and bigger until the Pre-IPO round.

However, for this mechanism to operate continuously, there is an important prerequisite: the company can be successfully listed. When the listing path is cut off and new capital no longer actively enters the market, the originally rising valuation chain will start to loosen from the bottom.

The Collapse of Old Investment Logic

Some keen investors have actually been deliberately controlling the investment rhythm since the first half of the year. "When you are in the game, you definitely know that the investment craze is irrational and the moves are out of shape, so when the investment reaches its peak, it is bound to decline", an investor analyzed to 36Kr.

But they cannot predict exactly when the bubble will burst and what will cause it.

Image from Visual China

In fact, many industry insiders told 36Kr that this round of cooling is not caused by a single event alone, but the result of the resonance of multiple factors from the primary market, secondary market and academic circle.

First of all, after three or four years of intensive investment, leading institutions have basically completed their layout in the embodied intelligence sector.

By 2026, capital has almost covered all the bottlenecks in the entire industrial chain of embodied intelligence. In 2024, hot money poured into hardware ontologies; in 2025, it shifted to robot brains and core components (dexterous hands, sensors, tactile sensors, bionic faces, etc.), and this year it flows to embodied data and world models. Almost every field has spawned a number of companies with tens of billions of yuan in valuation.

An industry that has not achieved commercialization at all obviously cannot accommodate so many unicorns, it's just that everyone found a buyer at the right time. An investment professional even said bluntly that the investment logic is as simple as "I only need to judge the feasibility of your listing, and I can invest as long as I can make money".

Under this transaction logic, the higher the valuation the secondary market gives to robotics companies, the more reason the primary market has to pay a higher price for unlisted startups. An investor said bluntly: "The higher the valuation of companies like Unitree rises, the better it is for the financing of unlisted companies."

Now, the secondary market cannot provide the returns they expected. Unitree, which went public in August, has seen its stock price fall all the way from the high point of 1100 yuan on the first day of listing, with the maximum drawdown exceeding 60% (the stock price has recently rebounded to 720 yuan); Mech-Mind, which went public in September, fell below its issue price; Geek+ once broke below the issue price, and Dobot's stock price also dropped significantly from its historical high.

These are the lucky ones that successfully landed on the market. What makes investors even more unacceptable is that the companies in their portfolio may not get the listing approval, and they cannot predict the direction of relevant policies.

In addition to the blocked exit path in the capital market, the latest progress in technology is also shaking some investors' previous judgments.

Image from Visual China

Over the past two years, a number of startups focusing on technologies such as VLA and world models have emerged in the embodied intelligence sector, trying to build general-purpose brains for robots. But in September this year, OpenAI released GPT-6 Astra — although it is not a model specially trained for robots, it has caused shocks in the embodied intelligence industry.

For example, Wang Xingxing from Unitree mentioned in a recent speech that after their actual test, GPT-6 Astra can make the robot complete painting with zero training, only through prompts. Surprisingly, the robot can also independently complete supporting tasks such as replacing pigments and cleaning brushes. "General large models may penetrate physical AI in the future," he judged.

In China, 36Kr learned that domestic foundational model vendors including DeepSeek have recently been in close contact with embodied intelligence enterprises, exploring opportunities to enter this field. ByteDance is also recruiting engineers and leaders for its embodied model business.

Obviously, large model vendors are the behemoths that are enough to swallow the embodied brain startups. An investor raised a question to us in the interview: if the model capabilities that an embodied startup spends nearly 10 billion yuan to train can be realized by general foundational models in the future, are the technical barriers established by these companies still worth a valuation of tens of billions of yuan?

Survive the Winter

The right to speak has reversed. "In the past, project parties often thought 'I am worth this price', but now investment institutions think 'You are not worth this amount of money'."

The cold wave has made many embodied companies change their attitudes one after another.

The founder of an embodied brain company told 36Kr that they have changed their financing strategy recently.

Previously, they preferred to gather several well-known investment institutions, with a leading institution taking the lead to complete a large-scale financing round. Now, they have turned to the "small steps, fast runs" strategy: as long as any investment institution is willing to enter, they can start a new round of financing.

"We must race against time to stock up on food for the winter," this founder told 36Kr.

Some high-valued leading embodied companies have begun to shrink their business lines.

36Kr exclusively learned that a leading embodied intelligence company with a valuation of more than 20 billion yuan is currently promoting the adjustment of its organizational structure, reintegrating the previously spun-off subsidiaries, and streamlining its R&D architecture. Some embodied intelligence companies with valuations of more than 10 billion yuan have also seen situations such as layoffs and offer withdrawals.

Investors are no longer willing to pay for empty, unrealistic visions easily.

An investor told 36Kr that now they have put forward several clear requirements for embodied intelligence projects — the company must have clear commercialization scenarios, real market demand, and it is best to have members in the founding team who have industrial resources and can promote product implementation.

"It is difficult for us to fund pure lab narratives now, and we can only offer relatively low valuations," said this investor.

Some previously marginalized projects have become one of the few remaining options now. Many investors told 36Kr that some funds have recently begun to flow to some projects that were not "favored" in the past.

These companies may not have the most dazzling technical background, nor are they committed to building general-purpose robots, and their stories are not attractive enough, but they have found clear customer demands in specific industries and scenarios, and obtained real orders, including some upstream component, joint or equipment manufacturers.

"Everyone can finally return to the normal investment judgment logic," the aforementioned investor