Interview with investors of Starry Epoch and Zhuji Dynamics: The embodied technology track won't be a "winner-takes-all" scenario.

Text by | Fu Chong

Edited by | Su Jianxun

When some voices began to sing down embodied intelligence and remarks like "unclear landing scenarios" and "the bubble is about to burst" were rife, Wang Guangxi expressed his judgment through patient investment.

Wang Guangxi is the vice president of Lenovo Group and the managing partner of Lenovo Capital and Incubation Group. As early as 2011, he invested in Megvii Technology, which provides industrial scenario solutions for automated robots. After the upsurge of embodied intelligence, he also bet on Xingdong Jiyuan and Zhujie Dynamics.

On July 21st, at the 2025 Lenovo Capital and Incubation Group CVC Venture Capital Week, we had a chat with Wang Guangxi about his views on embodied intelligence.

In his opinion, there isn't much difference between the so - called two sides in the dispute regarding the long - term technological direction of intelligent robots. The non - consensus mainly focuses on how to operate to maximize investment returns.

He speaks at a moderate pace and in a restrained tone, but he maintains a technological optimism about the development of embodied intelligence. Based on the experience provided by the Gartner Hype Cycle, Wang Guangxi judges that there will probably be a downward period in the embodied intelligence industry. But more importantly, it's about how to invest in enterprises with the potential to cross the cycle.

The current embodied intelligence industry is in a stage of flourishing development, which is also evident at the 2025 Lenovo Capital and Incubation Group CVC Venture Capital Week. From humanoid robots to composite - form robots, from remote - controlled surgical arms to hair - transplanting robots, etc., projects and products based on different technologies and scenarios were gathered on - site.

The hair - transplanting robot developed by Bangce Medical, an investee of Lenovo Capital and Incubation Group. Photo: Provided by the interviewee

However, even though the views are not convergent, there are still two clear - cut types of embodied intelligence companies invested by Lenovo Capital and Incubation Group.

One type consists of existing companies that had been engaged in robot - related businesses before the concept of embodied intelligence emerged, such as Elite Robotics and Yunji Technology. These companies have a basic customer base, and after adding AI capabilities, they can realize value in the original scenarios more quickly.

The other type is the new native companies in the AI era that challenge the boundaries, such as Pegasus Robotics, Kuawei Intelligence, Xingdong Jiyuan, and Zhujie Dynamics. They often have strong scientific research teams and the ability to dynamically adjust technologies. The risk of aiming high is greater, but the upper limit may also be higher.

Regarding the various non - consensuses and future predictions in the embodied intelligence industry, Wang Guangxi shared his views with "Intelligent Emergence". The following content is from the exclusive interview and was organized by the author:

Wang Guangxi, vice president of Lenovo Group and managing partner of Lenovo Capital and Incubation Group. Photo: Provided by the interviewee

Near - future incremental scenarios: Education and industrial closed - loop tasks

Intelligent Emergence: Is your investment style in embodied intelligence more relaxed or stricter?

Wang Guangxi: It depends on the stage and the prediction of technology. We started to pay attention to embodied intelligence in 2022, but we didn't make an investment after observing for half a year. At that time, we were quite cautious, and the point of hesitation was whether the generalization ability could be broken through.

It wasn't until April and May 2023 that we participated in a top academic conference in the field of embodied intelligence overseas and met many experts.

Although the entrepreneurship and financing in the field of embodied intelligence in the United States hadn't taken off at that time, after understanding the technological path, we believed that the Transformer framework had a very high possibility of improving the generalization ability of the model in the physical world. So we made consecutive investments in several companies in the second half of 2023.

Intelligent Emergence: In your opinion, in the near future of two or three years, in which scenarios will embodied intelligence mainly present opportunities? And what scale will it probably reach?

Wang Guangxi: The education and scientific research tracks, which have been proven by quadruped or biped robots, still have great potential for scale expansion. New technologies such as large models or world models will continue to drive research related to hardware and embodied intelligence.

Currently, relying on the education scenario, the market scale of quadruped robots has reached about several billion yuan. In the future, the scale of this field will be several times that of today, reaching the level of tens of billions of yuan.

Intelligent Emergence: Is the industrial field a good scenario for the implementation of embodied intelligence?

Wang Guangxi: As mentioned before, the education and scientific research tracks will reach the level of tens of billions of yuan, but in many people's eyes, the scale is still not large enough. Embodied intelligence will enter vertical industries in the future, and the industrial field will definitely be one of them. However, it's still hard to say how quickly the industrial scenario can reach a large scale.

Among the companies we've invested in, some have entered specific scenarios in the industrial, logistics, commercial, and business fields and provided products and solutions. However, it's still too early to say what percentage of the industrial field embodied intelligence can cover in the end, as this requires a very high level of stability.

However, in the scenarios and links that have already been entered, embodied intelligence and composite robots can reduce costs and increase efficiency for projects, and achieve a one - to two - fold or two - to three - fold increase in sales volume on the existing basis. This is a great opportunity.

Intelligent Emergence: So which industrial scenarios may be verified?

Wang Guangxi: There will be opportunities in fields where automation hasn't been fully achieved in the past. For example, in the past, AGVs, AMRs, and unmanned forklifts could complete transportation autonomously, but still required human assistance in loading, unloading, and connecting.

Embodied intelligence can eliminate the connection links and independently complete closed - loop work from movement, loading, and picking to transporting the finished products to the destination and packaging. There are still great opportunities to connect the end - to - end in the work that has already been covered by embodied intelligence.

At the same time, in the past, it was difficult for robots to understand customers' intentions through dialogue, resulting in problems in task initiation, scheduling, distribution, and the connection of various links.

Now, robots with LLM (Large Language Model) capabilities can understand and reason about users' demands more clearly after a dialogue and generate solutions. Especially after being equipped with arms, they can achieve unmanned pick - up and delivery.

That is to say, finding closed - loop scenarios in the industrial field will increase the value of intelligent robots. Although it can only replace a small percentage of the workforce, it will promote the scale of embodied intelligent robots in the industrial scenario to increase several times.

For example, several of our investee companies in the unmanned forklift field are doing well, but the penetration rate of unmanned forklifts in the forklift industry is less than 5%. If it can reach more than 10%, there will be a three - to four - fold increase.

Yunji Technology, an investee of Lenovo Capital and Incubation Group, is an existing company that had been engaged in robot - related businesses before the concept of embodied intelligence emerged. Photo: Provided by the interviewee

A flourishing track, not a winner - takes - all situation

Intelligent Emergence: Currently, there are many non - consensuses in the field of embodied intelligence. What's your view?

Wang Guangxi: It's currently in a stage of flourishing development. The technology and the prediction of the future evolution direction have not converged yet and are still in a very dynamic stage with many variables.

Even if some views may form a consensus, it doesn't solve much. For example, the acquisition of data is a bottleneck, which is an objective fact with a consensus, but everyone has different ways to solve this problem. Another example is that many people think VLA is a consensus, but the VLA models trained by different teams also vary greatly.

Having a consensus doesn't mean that the same level of ability, progress, and standard can be achieved in technological implementation.

Intelligent Emergence: So what kind of teams do you tend to invest in?

Wang Guangxi: Currently, the iteration speed of scientific research methods is very fast, several times faster than the innovation in previous eras. In such a rhythm, once a certain method is verified or converges to a certain extent, as a startup company, regardless of whether it's the same as the method it originally used, it must have the ability to catch up quickly.

At this stage, it's impossible to say that a certain technological path is definitely the best and most reasonable. It must be very dynamic. So the specific direction is not the most important.

This also explains why we've invested in some teams with a scientific - research background. This group of native scientists in the AI field, although excelling in different paths, have the ability to understand and absorb each other's technological strengths. Their strong academic and industrialization capabilities enable them to quickly keep up with the most cutting - edge changes.

Intelligent Emergence: From the perspective of team composition, which type of team is better?

Wang Guangxi: The profiles of companies emerging in different rounds are different. Currently, there are several types of entrepreneurs in the field of embodied intelligence, including young professor teams, teams with an autopilot background, and teams from large companies like DJI.

The genes and successful strengths of these people are different. I've always believed that when investing in early - stage technology companies, we must find those with long enough strengths.

Of course, they also need to have the ability to make up for their weaknesses, such as through co - founders or partners, or having strong learning ability and an open attitude, and being willing to cooperate with investors to make up for their weaknesses. Entrepreneurs with the above characteristics are what I think are better.

Back to the field of embodied intelligence, it can be seen that the strengths of different teams are different, but they are all valuable. For example, some have strong AI algorithm capabilities; some have manufacturing experience and are good at ontology research and development; or they have experience in the autopilot field, accumulating entrepreneurial experience and network resources that are of some reference value to the large embodied intelligence track.

So, this track is not a winner - takes - all situation like the Internet in the past. Teams leading in various aspects may all have their own opportunities. For such a diverse field, it's not advisable to bet on only one company or one type of enterprise. The probability of hitting the one - in - a - million winner is very low.

Intelligent Emergence: Currently, not every embodied intelligence company is good at building the "brain". Do you think this part needs to be "made up"?

Wang Guangxi: Everyone needs to make up for their weaknesses. Especially, existing companies need to consider how to better empower with AI. For example, one of our investee companies used to grow very fast in manufacturing core components, and the robot also showed very strong ontology capabilities, but its strength was not in the "brain" before.

However, different teams have different judgments on when and how to make up for this.

On this issue, some people think that they should add the "brain" ability when the technology reaches a certain level of maturity; others insist on staying at the forefront of the development of the "brain".

Moreover, the so - called "making up the brain" is also related to the direction and scenario of different products. Our investee company, Vita Dynamics, is developing companion robots, which may not require hand - operated operations and thus do not need particularly strong "brain" capabilities.

I think we should first do a good job in autonomous navigation, autonomous action, and autonomous interaction. As for how to make up for the generalized "brain" in different scenarios, there will be different strategies.

In Wang Guangxi's opinion, the field of embodied intelligence will be flourishing, and different forms of robots will emerge. Photo: Provided by the interviewee

Companies that find their own value can cross the "bubble" cycle

Intelligent Emergence: In the field of embodied intelligence, what do you think are the important milestone events or phenomena this year?

Wang Guangxi: Maybe there will be some IPOs this year or next year. Going public is a good milestone. As a secondary - market company, an enterprise can indeed gain many additional competitive advantages.

Among our investee companies, besides the new AI - native companies focusing on humanoid robots, there are also existing companies that had been engaged in robot - related businesses before the concept of embodied intelligence emerged. After adding AI capabilities, the existing companies may be able to enter the capital market faster than pure humanoid robot companies by developing composite - form robots.

Overall, in the next one or two years, whether it's new or old companies, some will reach a new starting point in the field of embodied intelligence through IPOs, which will have a certain impact on the industrial pattern.

Intelligent Emergence: In the field of embodied intelligence, you've invested in existing companies and are now also investing in incremental companies (new embodied intelligence companies). I believe your benchmark (evaluation criteria for projects) must be very high. How do you evaluate whether to invest in a company now?

Wang Guangxi: In this track, we already have more than 30 investee companies, and we've made arrangements for different types.

When making further investments, we'll consider two situations. First, we'll continue to pay attention to the more than 30 companies we've already invested in. The logic is simple. If the companies we originally favored are actively seeking financing, we'll also consider participating.

But as old shareholders, we'll judge the real progress of these companies. For example, we'll compare our prediction of the real progress of the company with the market's prediction of its development. If we think the real progress is faster and exceeds the market average expectation, and the valuation is at a reasonable level, we may increase our investment.

In addition, we'll be more cautious when it comes to investing in new companies.

Although the embodied intelligence industry is still in its early stage, don't underestimate the impact of the financing and market conditions on companies at different stages. To put it bluntly, if a new company starts much later, it will be at a disadvantage in the competition for funds, market resources, industrial endorsements, and talents.

So, if a new company in the field of humanoid embodied intelligence doesn't show a more cutting - edge and feasible technological tendency, it will bear a great risk of catching up. We'll be more cautious in our evaluation.

However, I think the market