HomeArticle

What did Lei Jun, Wang Xingxing and Li Yanhong talk about at APEC?

芯智能头条2026-11-18 11:46
At the APEC Summit, three entrepreneurs held heated discussions on AI implementation and the robotics industry.

Lei Jun disclosed that Xiaomi's embodied intelligent robots can already complete 8 hours of work in factories; Wang Xingxing believes that the generalization capability of robot models is approaching the critical breakthrough point; Li Yanhong reminded that if AI applications fail to drive revenue and profit growth for a long time, the industry may face the risk of bubble.

On November 17, the 2026 APEC CEO Summit opened at the Qianhai International Convention Center in Shenzhen. Lei Jun, Wang Xingxing and Li Yanhong participated in discussions on topics including the digital economy and the AI innovation ecosystem respectively, talking about robot deployment, large model applications, enterprise organizational restructuring and international cooperation.

Lei Jun: Robots Can Already Complete 8-hour Factory Operations, Further Exploration of Industrial Agents Will Continue

Lei Jun, Founder, Chairman and CEO of Xiaomi Group, first talked about the gap between AI development and commercial implementation.

He believes that over the past four years, the rapid development of large models has changed many people's lives and work, but the scenarios that can be truly implemented are still limited. In this cycle, relatively few companies can form a closed commercial loop and achieve profitability.

The next step that Lei Jun values is the further integration of AI with terminal devices and actual application scenarios. In his view, after AI enters the physical world, it will generate huge industrial opportunities, and it is worthwhile to re-examine various industries with new technologies and ideas.

For AI-powered phones and AI-powered vehicles, he believes that no sufficiently revolutionary and landmark products have emerged in the past few years, but there are opportunities to launch important products in the coming years.

On the manufacturing side, Lei Jun introduced that Xiaomi has used AI to optimize all links of its factories, and is further thinking about the possibility of industrial agents: enabling factories to produce independently, identify problems and iterate continuously.

Robots are the direction he emphasized. Lei Jun said that Xiaomi's embodied intelligent robots have entered factories for "internship", and can complete 8 hours of work, during which they may recharge autonomously twice. He judged that the imagination space of this industry may exceed that of the automotive industry.

Talking about Xiaomi's own AI layout, he mentioned basic large models such as MiMo, hoping to improve user experience with the help of these technologies. Xiaomi is still promoting its business around the "full ecosystem of people, vehicles and homes", while maintaining an open ecosystem. According to Lei Jun's introduction at the meeting, Xiaomi's IoT Platform has accessed 1.16 billion devices and has a large number of partners.

On the issue of Asia-Pacific cooperation, Lei Jun emphasized that all economies are not only important locations for industrial chains and supply chains, but also have huge consumer markets, and differences and complementarities bring space for cooperation. He believes that mutual understanding and trust are the important foundation for promoting cooperation.

Wang Xingxing: Generalization Capability Is the Most Challenging Issue, Manufacturing, Service and Security Need to Be Addressed After Breakthrough

Wang Xingxing, Founder of Unitree Robotics, described 2026 as "the most exciting year".

He believes that the AI and robot achievements seen by the public only show part of the progress. Underlying R&D is still advancing rapidly, and many technological changes have not yet been fully reflected in the products visible to the outside world. He suggested spending more time using AI every day and actively adapting to technological changes.

For robots, what Wang Xingxing cares most about is whether they can truly undertake work. He believes that the rise in industry popularity is related to the fact that language models have demonstrated value in life and office scenarios, and people also expect robots to complete practical tasks with the help of AI.

At present, the most challenging technical problem is the generalization capability of models. He hopes that one robot using the same model can quickly start working after entering different families and different working environments, without retraining every time the scenario changes. Combined with global technological progress this year, he believes that the industry has seen the critical breakthrough point, "it may come very soon".

Wang Xingxing further raised a question: if major breakthroughs are made in robot AI models next year, what do enterprises need to prepare for next?

His answer involves manufacturing, reliability and after-sales service: the current annual output of tens of thousands of units is very different from the possible demand of millions or even tens of millions of units in the future in terms of production and service difficulty. To expand the scale, it is necessary to solve the problems of product stability and post-delivery service at the same time. He also mentioned that the computing power demand of robots in the future may far exceed most people's expectations, and continuous investment is also required for security issues.

Talking about international cooperation, Wang Xingxing believes that all countries and regions have their own advantages and disadvantages in manufacturing, cutting-edge models, computing power and talents. China's manufacturing industry has advantages, but there are still areas to be improved in some cutting-edge models, computing power and top talents, and enterprises can promote technological and industrial development through cooperation.

According to his introduction, Unitree has been promoting cooperation with partners at home and abroad. At the ecological conference held last month, overseas partners accounted for more than one third of the on-site participants. He hopes that enterprises will pay more attention to the expansion of the future industrial scale, and create new demands and revenues through cooperation.

Li Yanhong: The Most Urgent Issue for AI Is the Return on Application, Organizations and Workflows Also Need to Keep Pace

Li Yanhong, Founder, Chairman and CEO of Baidu Group, participated in two dialogues in Chinese and English respectively on the same day, discussing how AI can be implemented in the Asia-Pacific region.

He divides the AI industry into four layers: chips, cloud infrastructure, models and applications. The investment and technological progress in the previous links ultimately need to reflect value through specific applications. In his view, what is more worthy of attention is how AI enters the actual scenarios of various countries, companies and organizations.

For the AI development in 2026, Li Yanhong chose the keyword "most urgent". In the past few years, AI technology has iterated rapidly, and investment in computing power, cloud infrastructure and other fields has continued to grow, but the penetration of the application layer is lower than expected. After many enterprises use AI, their revenue and profits have not improved significantly. If this problem cannot be solved for a long time, the bubble that the outside world worries about may become a reality.

Taking self-driving taxis as an application sample, he pointed out that such services use chips, cloud and models at the same time, and can directly serve travel needs. He also mentioned that directions such as AI accelerating new drug discovery and helping produce short dramas also have opportunities to translate technological progress into value that users can perceive.

According to Li Yanhong's introduction, Apollo Go is expanding its open road operation scope in Dubai, and carrying out cooperation with platforms such as Uber and Lyft. He hopes that more countries and economies will open up roads, so that self-driving technology can be more widely applied.

For how small and medium-sized enterprises can make good use of AI, he focuses on organizational capabilities. After motors enter factories, the assembly line needs to change the production mode; after enterprises introduce large models, they also need to rearrange positions, knowledge and workflows.

Combined with Baidu's practice, he put forward three directions: breaking position boundaries, precipitating tacit knowledge, and forming a closed workflow loop.

Tasks that were previously undertaken by front-end R&D, back-end R&D and testing teams separately can now be completed by one person who may play multiple roles with the assistance of AI. At the same time, the experience accumulated by employees and unwritten business rules need to be recorded for AI to call; when multiple people and multiple agents work together, the division of labor, data flow and feedback mechanism also need to be clarified.

Li Yanhong also cited examples of port operations and supermarket restocking. How quay cranes load containers involves complex combinatorial optimization; which products in stores sell fast and when restocking is needed depends on specific operational data. He believes that the combination of general models and industry data can better handle these practical problems.

AI has also changed the operation mode of technology enterprises themselves. Li Yanhong said that companies like Baidu used to be regarded as asset-light enterprises. After purchasing a large amount of computing power, their business is becoming more asset-heavy, and their dependence on capital has increased accordingly. He mentioned at the meeting that in the latest quarter, Baidu Intelligent Cloud's growth rate was close to 100%, and the call volume of ERNIE large model and tokens also grew rapidly.

In terms of regional cooperation, he focuses on the matching of computing power, models, demand and data, and also points out that there is an AI development gap between different economies. All parties need to reduce the application threshold through policy and industrial cooperation, so that more enterprises and economies can use advanced technologies and obtain actual benefits from them.

The discussions of these entrepreneurs all focus on the question of how AI is used: robots need to adapt to different environments and complete practical work, and enterprises applying AI need to adjust processes and measure returns. Manufacturing, after-sales service, organizational capabilities and international cooperation, together with model progress, have become important contents of this summit's discussion on AI and robots.

This article is from the WeChat official account "Core Intelligence Headlines", written by Mianmian, edited by Mark, and published with authorization from 36Kr.