Striding AI, a Beijing-based startup building world foundation models for general-purpose robots, has raised a nearly $100 million angel round backed by local corporations, entrepreneurs and venture funds — one of the largest angel financings ever recorded in robotics.
Full-stack bet on embodied intelligence
Founded by CEO Song Yao, Striding AI is developing what it describes as a robotic foundation system: world action models, human-in-the-loop reinforcement learning, data infrastructure and its own humanoid hardware, all engineered as one closed loop spanning perception, planning, execution, feedback and recovery. "We believe that breakthroughs in physical AI emerge from the continuous co-evolution of data, models, and infrastructure," Song said in the company's announcement. Early internal testing, the company says, showed its reinforcement learning approach improving task success rates by up to three times.
Retail first, then everything else
Commercial deployments will start in retail — shelf restocking, inventory counting, product organization and checkout assistance — because the sector offers repeatable workflows and abundant operational data for training. The company then plans to expand into food service, agriculture, logistics and healthcare, with skills learned in one environment transferring to the next.
China's world-model wave
The raise reflects Beijing's intensifying push into embodied AI, where world models — systems that let robots understand how actions change the physical world — have become the centerpiece of the next phase of the AI race. Striding AI's full-stack approach mirrors strategies pursued by Western rivals, though at a fraction of the valuations seen in deals like Agility Robotics' $2.5 billion SPAC merger. It also extends a torrid stretch for robot-intelligence funding that included Proception's $11 million seed round this week.
Reporting based on coverage from AI Insider and the Association for Advancing Automation (A3).