Simulation data is igniting physical AI, and institutions estimate the embodied-data market could eventually exceed 100 billion yuan, the 21st Century Business Herald reported on November 4. On the same day, UBTECH released Thinker-Sim, a one-stop system covering real-scene digitization, scaled simulation-data synthesis and robot-capability evaluation.
Unitree's Wang Xingxing said on November 3 the firm will focus on the robot "brain" and physical AI while seeking global cooperation, signaling that leading hardware makers are shifting the narrative重心 toward brains and physical AI. Overseas, Google DeepMind's Genie 4 world model extended its consistency window to two minutes with native audio, and Gemini Robotics 2 adapted to new robot bodies with fewer than 200 examples.
Guojin Securities noted the core variable of this physical-AI round has shifted from hardware performance to the emergence of generalization. With global high-quality real physical-interaction data at only about 500,000 hours against a need of at least 10 million hours, simulation-synthesized data has become the industry's agreed shortest path, then migrated to real machines.
Source: 21st Century Business Herald
