2026-11-18 16:06
Introduction: Wang Xingxing bets on generalization breakthrough, but scalability remains constrained by manufacturing and security challenges. Lei Jun reveals that Xiaomi's embodied AI robots have already completed eight-hour factory shifts;
Lei Jun revealed, Xiaomi's embodied AI robots have already completed eight-hour work shifts in factories; Wang Xingxing believes the generalization capability of robot models is approaching a critical breakthrough point; Li Yanhong warns that if AI applications fail to generate revenue and profit growth, the industry may face bubble risks.
On November 17, the 2026 Asia-Pacific Economic Cooperation (APEC) Business Leaders Summit opened at the Qianhai International Convention Center in Shenzhen. Lei Jun, Wang Xingxing, and Li Yanhong participated in discussions on topics including digital economy, AI innovation ecosystems, robot deployment, large model applications, corporate restructuring, and international cooperation.
Lei Jun, founder, chairman, and CEO of Xiaomi Group, first addressed the gap between AI development and commercial deployment.
He believes that over the past four years, large models have advanced rapidly, transforming many people’s lives and work, but the number of truly deployable scenarios remains limited. In this cycle, relatively few companies have achieved commercial closure and profitability.
What Lei Jun sees as the next step is deeper integration of AI with terminal devices and real-world applications. In his view, once AI enters the physical world, it will unlock significant industrial opportunities, warranting a fresh examination of various industries through new technologies and mindsets.
Regarding AI smartphones and AI vehicles, he notes that no sufficiently revolutionary or benchmark-defining products have emerged in recent years, but there is potential for major breakthroughs in the coming years.
In manufacturing, Lei Jun explained that Xiaomi has already applied AI to optimize every stage of factory operations and is now exploring the possibility of industrial agents—systems capable of autonomous production, self-diagnosis, and continuous iteration.
Robots are a key focus area. Lei Jun stated that Xiaomi’s embodied AI robots have entered factory "internships," capable of completing eight-hour shifts, during which they may autonomously recharge twice. He estimates the industry’s potential could surpass that of the automotive sector.
On Xiaomi’s own AI strategy, he mentioned foundational large models like MiMo, aiming to leverage these technologies to enhance user experience. Xiaomi continues to advance its “human-device-home ecosystem,” while maintaining an open ecosystem. According to Lei Jun’s remarks at the summit, Xiaomi’s IoT platform has integrated 1.16 billion devices and boasts numerous partners.
On regional collaboration, Lei Jun emphasized that economies are not only vital hubs for supply chains and industrial production but also significant consumer markets. Differences and complementarities create fertile ground for cooperation. He believes mutual understanding and trust are essential foundations for driving collaboration.
The founder of Unitree Robotics, Wang Xingxing, described 2026 as “the most exciting”.
He believes the public has only witnessed partial progress in AI and robotics. Underlying R&D is advancing rapidly, and many technological advancements have not yet been fully reflected in visible products. He advises dedicating more time daily to using AI and proactively adapting to technological change.
For robots, Wang Xingxing is most concerned about their ability to genuinely perform tasks. He attributes rising industry enthusiasm to the demonstrated value of language models in daily life and office work, fueling expectations that robots can leverage AI to complete real-world tasks.
The most pressing technical challenge today is model generalization. He envisions a single robot using the same model across different homes and work environments, able to begin functioning immediately without requiring retraining for each new context. Based on global technological progress this year, he believes the industry is nearing a breakthrough threshold, suggesting it may be imminent.
Wang Xingxing further asked: If robot AI models achieve major breakthroughs next year, what should enterprises prepare for?
His response covers manufacturing, reliability, and after-sales service: The difficulty of producing tens of thousands annually versus future demand potentially reaching millions or even tens of millions is vastly different. Scaling up requires simultaneously addressing product stability and post-delivery service. He also noted that future robotic compute demands may far exceed common expectations, necessitating sustained investment in safety.
On international collaboration, Wang Xingxing observes that different countries and regions each have strengths and weaknesses in manufacturing, cutting-edge models, compute power, and talent. While China excels in manufacturing, gaps remain in some advanced models, compute infrastructure, and top-tier talent. Enterprises can drive technological and industrial progress through cooperation.
He shared that Unitree has consistently promoted partnerships with domestic and international collaborators. At last month’s ecosystem conference, overseas partners accounted for over one-third of on-site participants. He hopes enterprises focus more on expanding future industrial scale and co-create new demand and revenue through collaboration.
Li Yanhong, founder, chairman, and CEO of Baidu Group, participated in two dialogues—one in Chinese and one in English—discussing how AI can be implemented across the Asia-Pacific region.
He categorizes the AI industry into four layers: chips, cloud infrastructure, models, and applications. Investments and technological advances in earlier layers must ultimately deliver value through concrete applications. In his view, greater attention should be paid to how AI integrates into real-world scenarios across nations, companies, and organizations.
For AI developments in 2026, Li Yanhong chose the keyword “most urgent.” Over recent years, AI technology has evolved rapidly, with continuous growth in compute power and cloud infrastructure investment. However, application penetration has fallen short of expectations—many enterprises using AI have not seen noticeable improvements in revenue or profits. If this issue persists, the feared bubble may become reality.
Using autonomous ride-hailing taxis as a case study, he pointed out that such services combine chips, cloud computing, and models to directly address mobility needs. He also highlighted promising areas such as AI-accelerated drug discovery and content creation for short-form videos, where technological advancement can translate into tangible user value.
According to Li Yanhong, Apollo Go is expanding its open-road operations in Dubai and collaborating with platforms like Uber and Lyft. He hopes more countries and economies will open roads to enable broader adoption of autonomous driving technology.
For small and medium-sized enterprises (SMEs) seeking to leverage AI, he emphasizes organizational capability. Just as electric motors required changes to factory assembly lines, introducing large models demands reconfiguration of roles, knowledge systems, and workflows.
Based on Baidu’s experience, he proposes three directions: breaking down role boundaries, codifying tacit knowledge, and establishing closed-loop workflows.
Tasks previously handled separately by front-end, back-end R&D, and testing teams can now be managed collaboratively by individuals with AI assistance. Meanwhile, employees’ accumulated experience and unwritten business rules must be documented for AI access. When multiple humans and intelligent agents collaborate, clear division of labor, data flow, and feedback mechanisms are essential.
Li Yanhong also cited examples from port operations and supermarket restocking. Optimizing container loading by quay cranes involves complex combinatorial optimization; determining which retail items sell quickly and when to restock depends on granular operational data. He argues that combining general-purpose models with domain-specific data enables better handling of such practical challenges.
AI has also transformed the way tech companies operate. Li Yanhong noted that companies like Baidu were once considered lightweight assets, but heavy investments in compute power have made their operations increasingly capital-intensive. At the summit, he mentioned that Baidu Intelligent Cloud grew nearly 100% in the most recent quarter, with rapid increases in Wenxin Large Model and token usage.
On regional cooperation, he focuses on aligning compute capacity, models, demand, and data across economies. He acknowledges existing disparities in AI development among different regions. All parties need to lower application barriers through policy and industrial collaboration, enabling more enterprises and economies to access advanced technologies and derive real benefits.
The discussions among these entrepreneurs all converge on one core question: how AI is used. Robots must adapt to diverse environments and perform real tasks; enterprises applying AI must adjust processes and measure returns. Manufacturing, after-sales support, organizational capabilities, and international collaboration are now as crucial as model advancements, becoming central themes in this summit’s discourse on AI and robotics.
Source: XIN Intelligence Headlines
Disclaimer: Contains third-party opinions, does not constitute financial advice
The AlphaGo Moment in Cybersecurity, or Starting from These 466 Agent Matchups
13 days ago
HuggingFace's "AI Duck" Goes Viral, Boosting Chinese Chip Supplier Rockchip
09-02
Yushu Surges 629% on First Day of Trading, Humanoid Robots Sell 5,500 Units
08-19
14 Executives Depart, Horizon Unleashes a Robot Legion
08-11
Yushu's 61 Billion IPO: Who Will Make Money?
08-07
AI is not a search bar—it's a supercomputer costing just a few dozen dollars
08-05
Former Huawei Genius Youth: Building 5,000 Reliable Robots This Year Won't Exceed 3 Companies
08-05






