The founder of embodied intelligence came to me for an interview.

Have you noticed that Wang Xingxing of Unitree has been making more public appearances recently? He was seen at the Bund Summit, the Wuzhen Internet Conference, and the Hongqiao International Economic Forum, and there have also been many media interviews with him.

Just half a year ago, through an investor, I indirectly helped a local government invite him to an important event, but was told that "at most, he would send a vice president to attend." Of course, at that time, the "Six Rising Stars in Hangzhou" emerged, and Unitree, as the hottest company in the field of humanoid robots, had its concept stocks booming in the A-share market for half a year. Coupled with its amazing performance at the Spring Festival Gala, it was normal for it to be in the spotlight.

Now, Unitree remains a star company. Wang Xingxing's frequent statements may be related to the IPO process being on track. However, in the fields of humanoid robots and embodied intelligence, new companies, products, collaborations, and financings are constantly emerging. Variables still seem to be everywhere, but I have an illusion that despite the noise, the "industry hierarchy" has solidified. Otherwise, why are the leading companies still struggling so hard?

In particular, an investor told me that the phased pattern in the field of embodied intelligence has been formed. Zhipu, Unitree, Galaxy Universal, Xinghaitu, and Xingdong Jiyuan are undoubtedly in the first echelon. She told me that they are currently the "Five Tigers in Embodied Intelligence," following the example of the "Six Rising Stars." From this, it seems that we can conclude that at least in the field of embodied intelligence itself, the opportunities for new entrants are becoming fewer.

Investors may not be qualified to make this judgment, but I think it is at least an interesting topic. If you look at the new industries and tracks that emerged in the context of major historical trends such as real estate, the Internet, and mobile Internet, there have always been various players trying to make judgments about the first half, the middle, the second half, the end, and even the finale of these industries at different stages. Even if the reasons given at that time seemed conclusive, there were often unexpected variables.

Let me briefly introduce the analysis made by this investor from the rational, technical, capital, industrial, and some emotional aspects of embodied intelligence:

First of all, the "Five Tigers" have the highest valuations and revenues, are the closest to IPO, and all have relatively stable application scenarios.

Compared with the explosive period at the end of 2023, there has been no fundamental change in the technological foundation. The two technological routes of simulation learning and reinforcement learning are gradually converging, and the differences are becoming smaller. Each of the Five Tigers has its own advantages in technological reserves.

The threshold for entrepreneurship is also getting higher and higher, which can be seen from the constantly refreshed records of seed - round and angel - round financings.

Institutions especially value the background of embodied intelligence companies. Background determines resources. In addition, institutions that invested from the beginning or changed their minds later have almost completed their investments. Even if they want to invest further, they will probably choose the leading companies.

Moreover, even for those from the same academic background, the treatment and the amount of funds that can be raised for entrepreneurship now are much worse than those two years ago.

In addition, some startup companies in the field of embodied intelligence have closed down, such as Dakuoke Technology, Yixing Robot, Aldebaran, K - Scale... Even some entrepreneurs who have raised funds in several rounds can no longer hold on and have started looking for jobs.

Of course, whether the pattern in the fields of humanoid robots and embodied intelligence has solidified, who deserves to be called a "tiger," how many "tigers" there are in total, whether there are still opportunities for latecomers, and how big these opportunities are. China Venture Capital has interviewed several investors and entrepreneurs and obtained some opinions, which will be discussed in detail later. It is indeed time to have a phased discussion about the embodied intelligence industry.

For example, the public is almost aesthetically fatigued by the dazzling performances of robots, but Xiaopeng Robot still stimulates public opinion in a way similar to the "uncanny valley." Another example is that Wang Xingxing said at the Wuzhen Conference that the field of embodied intelligence has been like a dream in the past year, and science - fiction scenarios will accelerate their entry into reality in the next few years. The field of embodied intelligence AI is "relatively easy" compared to nuclear fusion and Mars exploration, and there will be more surprises in the next year or two. On the other hand, he believes that the biggest challenge in the current robot field is the entire AI model of embodied intelligence. "The model structure and data scale are not sufficient," and there is no unified consensus in many hardware and software directions.

Overseas, embodied intelligence is of course one of the focuses in the extension of the AI industry. Figure AI has no revenue at all, but its valuation is as high as nearly $40 billion. The biggest support for Musk's trillion - dollar compensation plan is also humanoid robots. "The future of Tesla is a technology company centered on AI and robots," said the world's richest man. Even if it's just a promise for now, you have to take it seriously.

A recent report from Morgan Stanley also shows that Apple's recent recruitment, patent applications, and supplier collaborations in the field of robotics indicate that it is preparing for large - scale production in the future. The report also reveals that Apple plans to launch an electric desktop assistant robot in 2027. The report predicts that Apple may enter the humanoid robot market in 2040 and form a new business segment with an annual revenue of $133 billion.

Wow, the scale is quite large, but it's pushed all the way to 2040.

You may not doubt that robots will one day be able to completely replace humans in work, but no one can be sure that the current embodied intelligence is the path to the endgame. Just like AI is revolutionizing human society, but the top players in the global AI field are still arguing with each other.

Recently, Yann LeCun, Fei - Fei Li, Jensen Huang, Geoffrey Hinton, Bill Dally, and Yoshua Bengio jointly won the Queen Elizabeth Prize for Engineering. I won't introduce these six individuals one by one. You can think of them as the "Six Dragons in AI," the global authorities. Regarding whether truly human - like AI will emerge in the future, some of the "Six Dragons" focus on solving practical problems step by step, some think that machines can never become human, some think it's too early to discuss this issue, believing that the current paradigm of large - language models is not a fundamental breakthrough, and some think that this day will come much faster than expected, and humans should be vigilant and make preparations.

The origin of the transformation is AI. The problem is as it is. Gathering these entrepreneurs and scientists together is not for getting a unified answer. The same is true for the derived field of embodied intelligence. For example, Rodney Brooks, the theoretical founder of embodied intelligence, recently wrote an article questioning the industry bubble: current humanoid robots cannot master fine - operation skills, and billions of dollars in investments may go to waste.

Therefore, even at this phased moment, it is difficult to draw a definite ultimate conclusion. But at least, we can catch some clues about the future. You can see that Fei - Fei Li thinks that language intelligence is not enough and that the future lies in spatial intelligence. Yann LeCun won an AI award, but it doesn't prevent him from denying the current paradigm of large - language models, saying that "something very important is still missing."

I can't find what's missing in Meta. So, what stage has the field of embodied intelligence in China reached? What judgments can be made? What is nonsense? What is there, and what is lacking? Can the early - starters rest easy? Do the late - comers really have no chance at all?

Detailed Analysis of the "Five Tigers"

If we have to divide the echelons, let's first look at the valuations and revenues. Generally speaking, regarding the "Five Tigers" judgment given by the investor at the beginning, Unitree, Zhipu, and Galaxy Universal are definitely in the leading position, while Xinghaitu and Xingdong Jiyuan are somewhat questionable.

Unitree's valuation has reached 12 billion yuan after the Spring Festival. Some investors say that its valuation will exceed 100 billion yuan if it goes public on the Science and Technology Innovation Board. Zhipu has just completed a transaction, and its valuation has exceeded 15 billion yuan. Recently, there was even a "rumor" that its valuation would reach HK$40 - 50 billion if it goes public in Hong Kong. Galaxy Universal is asking for a valuation of 10 billion yuan and is negotiating the latest round of financing with Middle - Eastern investors.

The second echelon consists of companies with valuations of around 5 billion yuan, including Xinghaitu, Xingdong Jiyuan, and Independent Variable Robot, etc. These companies are also mostly in the stage of negotiating the next - round valuation. For example, Independent Variable Robot is asking for a valuation of $1 billion in this round, and Xinghaitu's valuation is around $700 - 800 million.

Regarding valuations, we can make a comparison between Chinese and American embodied intelligence companies. Excluding listed companies like Tesla, among startup companies, Figure AI's latest valuation has reached nearly $40 billion, and others like PI and Skild AI are all around $5 billion. In contrast, the valuations of Chinese companies are at most around $2 billion when converted into dollars.

Obviously, either American investors are over - pricing, or Chinese embodied intelligence companies are greatly undervalued. New Horizons Investment Management, a UK - based company mainly engaged in cross - border investment and services, helps Chinese founders "go global" on the one hand and foreign investors "invest in China" on the other hand. Xia Yuchen, the founder of New Horizons, told me that the secondary market mostly allocates to the Chinese market through index funds, but in the primary market, especially European and Middle - Eastern capital, has a stronger demand than before to invest in Chinese startup company assets.

He believes that on the one hand, the liquidity of foreign funds entering China is still restricted, and on the other hand, Chinese enterprises are accelerating the process of building brands overseas. "Only when products gain recognition can value be formed, and currently, the process of gaining recognition is underway."

Jensen Huang visited China for the first time this year to attend the China International Supply Chain Expo and promoted robots from Galaxy Universal, Beijing Humanoid Robot Innovation Center, Accelerated Evolution, and Zhipingfang at the exhibition. Foreign research institutions have made large - scale purchases of mechanical dogs, robotic arms, etc. from Unitree, Xingdong Jiyuan, and Deep Robotics. Galaxy Universal has established a joint venture with Robert Bosch GmbH.

Liang Juncheng, a partner at New Horizons, also told me that foreign investment in Chinese assets goes through two logics: asset allocation and value discovery. Currently, some foreign investors are following the former logic, and there will be a process of value discovery in the future.

Next, let's look at revenues. Unitree has the highest revenue, reaching 1 billion yuan. Deep Robotics comes second. Zhipu has also signed several large orders with China Mobile, Fuling Precision Industry, and Longcheer Technology this year, and its revenue is at least hundreds of millions of yuan. For companies with valuations of around 5 billion yuan, such as Xinghaitu and Xingdong Jiyuan, their revenue targets are mostly around 100 million yuan, and the probability of achieving these targets is relatively high.

After all, times have changed. Even for cutting - edge technologies, blindly burning money is not practical today. However, Independent Variable Robot is an exception. Investors seem to have great patience for its commercialization. Generally speaking, these embodied intelligence companies, which have been established for only two or three years, are mostly exploring applicable scenarios and commercialization paths. For example, Galaxy Universal's revenue mainly comes from its cooperation with Meituan Pharmacy and CATL's lighthouse factory.

"The core logic of choosing Meituan Pharmacy as the core scenario is as follows: first, customers are willing to pay a high price to try out its products; second, customers can tolerate a certain sacrifice in efficiency to let the products operate in practice to accumulate data; third, the pharmacy scenario itself has a relatively high tolerance for robot errors. If the products are directly put into factories or the logistics industry, a single robot problem will disrupt the rhythm of the entire production line or logistics chain." An investor explained the reason why Galaxy Universal first tested its products in pharmacies.

Of course, Meituan and CATL are both important industrial investors of Galaxy Universal. They must support it, and the industrial parties themselves can also explore cutting - edge businesses through this.

Embodied Intelligence and "Family Background"

From an academic perspective, entrepreneurship in embodied intelligence pays special attention to academic origin. UC Berkeley and Stanford have become the sources for VCs to compete for talents in this wave of embodied intelligence entrepreneurship.

UC Berkeley's strong disciplines are robotics, vision, and reinforcement learning. Among them, Pieter Abbeel is a pioneer in the field of applying reinforcement learning to robot control. Two individuals who returned from Pieter's laboratory at Berkeley are Gao Yang, the CTO of Qianxun Intelligence, and Wu Yi, the founder of Frontier Technology. Chen Jianyu, the founder of Xingdong Jiyuan, and Xu Huazhe, the co - founder of Xinghaitu, also studied at Berkeley. Stanford's strength lies in image processing. An example of a returnee is Wang He, the founder of Galaxy Universal, who is a doctor from the Department of Electrical Engineering at Stanford University and studied under Leonidas J. Guibas, a member of the National Academy of Sciences, National Academy of Engineering, and National Academy of Medicine in the United States.

Since the technological routes have not yet converged, it's hard to say which company currently has an obvious technological advantage. However, companies with valuations above 5 billion yuan have initially shown different characteristics. Although they are all telling the same big story, the differences are reflected in details, including the understanding of technological routes, views on data, planning for applicable scenarios and commercialization, and the rhythm of where to focus their efforts at present.

For example, Zhipu impresses people with its strong resources and is quite skilled in capital operations such as reverse takeovers. It has four scientific research teams and three business lines internally. This horse - racing mechanism has also led to mixed reviews from the outside world. "You should regard Deng Taihua as an entrepreneur from the Huawei ecosystem. What he is doing is like Huawei in a niche field." An FA provided me with a perspective to understand this company.

An investor told me that Hillhouse Capital initially invested in this company because it valued Deng Taihua himself. An MD specifically monitored vice - presidents who left Huawei. "Investors are investing in Deng Taihua. When he wanted to enter the robotics field, he recruited core talents like Zhihui Jun and nurtured a group of scientists in the Huawei way. The core of Huawei's management is sales and finance. Sales control the revenue, and finance controls costs and salaries, so it can continuously expand into new businesses, from communications to mobile phones and then to automobiles. Deng Taihua is also managing Zhipu in this way." The aforementioned FA commented.

To some extent, the success or failure of this company will be a re - verification of Huawei's methodology in the field of cutting - edge technologies. At least for now, top - tier capital has invested, and Zhipu has received funds from both Sequoia Capital and Hillhouse Capital.

Xingdong Jiyuan's advantage lies in its combination of hardware and software. "One of the reasons why Figure in the United States has a high valuation is that from the first day of its establishment, Figure clearly stated that it would take the path of combining hardware and software. Those who are truly deeply involved in the robotics industry know that the operation of hardware must rely on software support, and the iteration of software also needs continuous adaptation with hardware." An investor who is optimistic about this company said.

Xingdong Jiyuan developed its dexterous hand independently at an early stage, and its core revenue comes from selling dexterous hands. This path is similar to Unitree's early practice of selling quadruped dogs to research institutions. "The most important thing is to be able to produce products first, enter universities, and start selling quickly to establish a position." Early commercialization is not only for accumulating data. It is said that Chen Jianyu's idea is that if there is no commercialization, it is impossible to set KPIs for the team, and once there is commercialization, customers' requirements become KPIs.

Chen Jianyu, who graduated from Tsinghua University and UC Berkeley, is the core of this company. He has been engaged in scientific research in universities and has no experience in running a company. This background is regarded as a double - edged sword by investors. The advantage is that the company's personnel structure is relatively simple, and it will not waver on major technological routes, and there is less internal strife between different technological directions. The disadvantage is that he is not good at promoting himself and the company. Some investors think that Xingdong Jiyuan is like a "well - rounded player," with comprehensive capabilities but lacking obvious highlights.

Now let's talk about Independent Variable Robot.