While the humanoid robot sector is crowded with full-sized models of 1.6 meters and 1.7 meters, focusing on competing for "cool-looking" specifications, Zhao Weichen, co-founder of Acceleration Evolution, stated that robots of 1.6 meters, 1.7 meters, or even taller, although they look impressive, cannot enter households because they are unsafe.

This clear judgment has led Acceleration Evolution to firmly adhere to the direction of "making smaller" since its inception: from the 30-kilogram Booster T1 to the 95-centimeter Booster K1, which weighs less than 20 kilograms and can be lifted with one hand. Even though "the technical threshold for making smaller robots is higher," safety has always been the top priority. This "anti-involution" choice is creating a key breakthrough for the practical application of humanoid robots.

Recently, Acceleration Evolution officially revealed that all its demonstration robots at the CES 2026 exhibition held from January 6 to 9 were completely sold out by pre-order, achieving a remarkable sales per square meter ratio for the booth. Furthermore, in December 2025, Acceleration Evolution achieved its first positive monthly operating cash flow since its founding. Not only that, the company, which sold 45 humanoid robots in 2024, reached a sales volume of 1,034 units in 2025. This means that in just one year, Acceleration Evolution achieved a rapid growth of nearly 22 times.

In Zhao Weichen's view, the fundamental reason why Acceleration Evolution can quickly achieve a break-even point is that it has found a sustainable path combining education and competitions. The successive rounds of financing in June, July, November, and December 2025 confirm the capital market's recognition of its unique development path. In an exclusive dialogue, Zhao Weichen detailed how Acceleration Evolution can stand out from the intense competition in the humanoid robot sector, also offering a glimpse into a new blueprint for the future development of humanoid robots.

The following is a summary of the interview dialogue for your reference.

"Robots need to be introduced from a young age."

How does Acceleration Evolution think about its commercialization path?

Regarding commercialization, if we look at the普及 path of computers. When computers like Apple I and Apple II were born in the 1970s, there wasn't even an application ecosystem or the internet. Under those circumstances, commercialization was still possible, but it had to proceed along two dimensions. First, tech enthusiasts, who pursue firsthand information on cutting-edge technology. They will buy and experience, and engage in secondary development, regardless of what you've created, as long as it's cutting-edge and represents future technology.

The other category is school education. Educational settings essentially lead future普及 trends. Taking the United States as an example, many schools purchased Apple's early computers and IBM computers back then, offering computer education courses in schools. It started with typing, gradually moved to learning how to use Word, PowerPoint, how to send emails, and later to learning programming. This is a complete system. Since we define robots as personal robots, they are analogous precisely to personal computers or mobile phones. So, if positioned as personal computers, they will become a very important interactive terminal for us in the future.

Therefore, I believe the core of commercialization is similar to that of computers. Our first step is to切入 the education scene, using this educational setting to build student awareness. Then, based on this educational scene, these students will become future developers. Just like many programmers today, they weren't born with the skills; they first learned programming tools in school and now have become core application developers in our internet and mobile phone ecosystems. The robot field will also have this wave of individuals. This group itself also needs to learn these capabilities in school to become our core developers in the future.

Another direction for commercialization, which aligns perfectly with education, is competitions. We have been working in the field of robot soccer for many, many years. Tsinghua University actually had a几十人 soccer team twenty years ago. Our chief scientist, Professor Zhao Mingguo, was working in this area back then. Although some of the more traditional algorithms at the time were not like today's reinforcement learning or imitation learning algorithms, they could already participate in some robot soccer competitions. Under those circumstances, we had already accumulated a great deal of experience in such events.

Furthermore, a German university team used our smaller Booster K1 robots and also won the championship in the small-size league of the Robot World Cup. These phenomena prove that during the entire process of robot commercialization, these top universities have already realized that China's technological development is very rapid. After comparing all the hardware and software platform solutions worldwide, they ultimately chose products from Acceleration Evolution. We are gradually leveraging these two scenarios—competitions and education—to cultivate a large number of early developers, allowing them to develop applications and participate in competitions based on our software and hardware platforms. Both students and teachers need a scenario to validate their abilities.

How is the market size estimated from these two commercialization aspects?

China has approximately 3,000 universities, including institutions of higher education and vocational colleges, which represents a huge market volume. I looked at the普及 path of computer labs. In 1985, there were about 100 computer labs nationwide. By 1995, there were nearly 100,000 labs. By 2005, it might have reached a scale of 500,000 labs, including off-campus venues like internet cafes. So we believe this is a massive market. Currently, the Ministry of Education is actually promoting AI education, including education related to robots. This is a very good opportunity for Chinese children to learn how to use the next generation of hardware terminals, including robot programming skills, earlier than American children. All these things need to start from a young age. So we think this is a huge market. Second, it also has enormous social value. Third, we very much want Chinese robots, Chinese products, to seize this window of opportunity at a global level and go worldwide. I think this is an embodiment of Chinese strength.

(Including Europe, Southeast Asia, and South America, which are actually our existing customer groups. Can we rapidly scale this up based on the Robot World Cup event, quickly establishing close cooperative relationships with global universities and K-12 institutions through word-of-mouth? There is tremendous value in having children around the world learn about our Chinese technological products from an early age. This is a direction we will focus on promoting in the next stage.)

"Robots of 1.7 meters or even taller look cool, but they can't enter households."

What aspects does Acceleration Evolution focus on most for its humanoid robots?

Stability, safety, and interactive capability are what we focus on most. So-called safety involves many dimensions; for example, being lightweight is the foundation of safety—the robot must be sufficiently portable. Second is stability; it must maintain stable movement in different environmental scenarios. Third is interactive capability; these are the three major modules we are currently working on.

The first is a whole-body motion large model, which enables the robot to maintain stable movement in various scenarios. The second is a navigation large model, which allows it to move autonomously in various environments. One difference from autonomous driving is that it has many components. For instance, if the body passes through, but the hand is raised, the hand might get stuck. So the embodied navigation large model must consider the status of all the robot's organs and components during displacement. The third is an interaction model, which enables the robot to proactively initiate interactions with users in different scenarios.

Moreover, unlike many companies that make full-size 1.6-meter, 1.7-meter robots, we have been consistently focused on making smaller robots since our founding. Although the technical threshold for making smaller robots is higher, we judge that robots of 1.6 meters, 1.7 meters, or even taller, although they look cool, cannot enter households because they are unsafe. You might have seen some videos before where full-size robots knocked people over during robot sports events. In our view, this is a safety hazard that prevents them from entering households, let alone school education scenarios, because safety is always the primary factor in educational settings. Our 1.2-meter robot is called Booster T1, weighing about 30 kilograms, which is still a bit heavy. So we further launched Booster K1, which is 95 centimeters tall and weighs less than 20 kilograms, basically easy to lift with one hand.

(Acceleration Evolution Booster K1 Robot)

Faced with increasingly fierce competition, how does Acceleration Evolution plan to break through?

First, we are currently the only company making robots positioned at the 95-centimeter size. As I just mentioned, we believe this form factor can become a viable form for the early落地 of personal robots.

Second, we are the only robot company with intelligent AI reserves. This is also evident from many of our real-world scenarios, including 3v3 and 5v5 soccer matches, and our robot AI formations, etc. These are real scenarios. The fact that everyone currently sees these scenarios demonstrates that they possess the requisite AI capability reserves and智能化储备. Offering robot courses in schools definitely requires teaching at the AI and intelligence level, not just staying at the hardware level. So we believe we have a strong leading and first-mover advantage in this area.

Then, regarding our core software层面, I think this is something many companies overlook. Why was Apple successful? Not just because it had good hardware. On the contrary, I think there might not be a fundamental huge difference in hardware between Apple and other phones. It was the iOS ecosystem that succeeded because it had the iOS operating system and Xcode, a developer-friendly development tool. Therefore, we have a software development team with members from ByteDance. Now, 20% to 30% of our team members are very focused on developing the underlying operating system and development tools for robot developers, lowering the entire development threshold from originally requiring a master's or doctoral degree to now potentially a high school level. We can likely lower it further to middle school or even elementary school level. They can use some very simple programming languages, or even natural language, like the no-code/low-code programming we see now, to develop many applications for robot capabilities. I think regarding this matter, currently only we are investing very聚焦ly in this area.

"In the next ten years, the robot industry will definitely be a protracted battle."

Acceleration Evolution completed multiple financing rounds in 2025, but many robot companies burn through money quickly. What is your view on this?

Actually, regarding burning money, Acceleration Evolution doesn't burn money that heavily. We can quickly achieve break-even in 2026. This is actually based on our current commercialization progress. From the perspective of overall financing, we are more focused on maintaining a strong ability to generate internal造血. On one hand, through commercialization; on the other hand, through financing. The next ten years will definitely be a protracted battle for the robot industry. The development in this field is also a long-term struggle; it will likely take ten or even twenty years to achieve what I believe is a 50%普及率. Under this premise, having a relatively good capital reserve is still necessary.

Some industry practitioners say that humanoid robot companies need at least 5 billion RMB on their books to succeed. For Acceleration Evolution, is funding a门槛?

Any startup needs to focus. I don't believe any independent startup can handle every aspect of the humanoid robot赛道, which has a very long and deep industry chain, by themselves. This is a strategy that doesn't align with technological development and startup logic. Especially from a startup's perspective, it's impossible to do everything. If you really intend to do everything, then I judge that you would need massive cash reserves; perhaps just for the 'brain' aspect, you might need reserves on the order of 5 billion.

What should future AI robots be like?

We believe the next generation of AI robots must involve AI finding people. The previous generation involved people finding AI. For example, whether it's mobile phones or computers, these products are immobile; people have to find the AI, you have to go to them to interact or have them help with intelligent assistant tasks. But the next generation of AI must be about AI finding people. Why do we make bipedal humanoid robots? The core reason is that they must possess all-terrain mobility. They need to be able to climb stairs, go up and down slopes, walk on grass and sand, possessing such全场景 mobility. Only then can AI find people be realized, and we can truly treat it as an intelligent assistant or a家庭玩伴. It will follow me and help us do many things.

Agent is precisely what we consider the most important application in the ecosystem. For a long time, robots might not be able to be bought home and immediately help people accomplish all tasks. So for a long time, there will be a need for embodied Agents, which are analogous to current AI Agents.

AI Agents are based on a foundation large model, solving hallucination problems in vertical scenarios, and completing tasks through the Agent. It's the same in the embodied intelligence field. We now have underlying embodied models, whether for embodied motion control, navigation and interaction, or so-called embodied large models. Each vertical actually requires our developers to develop corresponding embodied Agents to meet the needs of users in that vertical. Only then can personal robots truly be applied in various scenarios, whether in life, home, entertainment, or work.