Physix Frontier · News Briefing Card (TMTPost · Nov 4, 2026)

Wang Xingxing: 2027 May Be Robot Industry Tipping Point; Unitree Explores Physical AI Self-Evolution

KEY FACTS

At the 2026 World Internet Conference Wuzhen Summit, Wang Xingxing predicted that 2027 may mark a tipping point for the robotics industry or models.
Unitree Robotics has internally conducted preliminary research on autonomous evolution of physical AI models, where the model writes code and iterates training itself, while humans are responsible for evaluation.
Wang Xingxing stated that current robots can achieve nearly 100% success rates in fixed workstation tasks, but fine manipulation remains the main bottleneck.
Unitree Robotics' overseas business accounts for nearly 50%, and more than one-third of attendees at its October partner conference came from overseas.
Unitree's new-generation dexterous hand has 22 degrees of freedom, and it has released a full-size humanoid robot approximately 1.7 meters tall.

KEY DATA

nearly 50%Overseas product export share

22Dexterous hand degrees of freedom

approximately 1.7 metersFull-size humanoid robot height

at least one-thirdOverseas attendee share at partner conference

PHYSIX OBSERVATION

Wang Xingxing is betting the tipping point on 2027, which in essence is a gamble that model self-evolution can close the millimeter-level shortcomings in physical interaction. If models can truly write code and iterate autonomously, robot R&D will shift from manual tuning to a data flywheel, but the fact that humans are left with only a scoring role also means the evaluation criteria themselves become the new bottleneck. Chips and energy were called out, indicating that competition has moved upstream from algorithms to the manufacturing foundation.

Source: TMTPost report