Infiforce

Robotics · Embodied AI · China Tech

Funding

China's Infiforce raises nearly $150M for ego-native robot world models

Betting that first-person human video can replace costly teleoperation data collection, a wager also being made by well-funded labs abroad.

• Beijing-based embodied-intelligence startup Infiforce closed a combined Series A and A+ round worth nearly RMB 1 billion (roughly $140-150M), led by Dunhong Capital and an unnamed top state-owned capital platform.
• Infiforce, founded by former 15-year Alibaba VP Bai Huiyuan, builds AtomBrain, an 'embodied brain' meant to give robots the ability to understand and adapt to the physical world across different robot bodies.
• Proceeds will fund R&D on AtomBrain's causal world model, the company's DataGrid full-stack AI infrastructure, and large-scale deployment validation for its multi-form-factor robot lineup.
• Infiforce has already run pilots in more than 30 Chinese cities across over 100 real-world scenarios, spanning commercial services, warehousing/logistics, and industrial manufacturing.
• The company says it will soon release what it calls the world's first embodied model built around high-quality first-person 'Ego' data, aimed at letting skills transfer across robot bodies and tasks rather than being locked to one.
• CCV founding partner Zhou Wei said the firm keeps backing Infiforce because it has moved from world-model research to industrial validation, calling it a 'complete loop' from technology to commercial deployment.
• Infiforce's ego-centric video approach echoes a broader industry shift, seen at firms like Dyna Robotics, toward training robots on abundant human video instead of scarce, expensive teleoperation data.