Chinese embodied-AI startup Infiforce has closed nearly $150 million across Series A and Series A+ rounds to industrialise a first-person data pipeline it believes will unlock generalist robot brains. The company said the raise, led by Dunhong Asset Management with state-owned platforms and Zhejiang University Sci-Tech Innovation Group, funds three pillars: its AtomBrain embodied-intelligence system, expansion of the DataGrid data platform, and multi-form robot deployments in factories, warehouses and commercial services.
Ego Data As A Training Substrate
Rather than lean primarily on remote teleoperation or third-person demonstration, Infiforce is training on what it calls Ego data — first-person recordings of humans doing physical work, paired with motion, spatial and environmental signals. That approach, the startup argues, produces more training data per dollar than piping everything through robots, while giving its Atom world models a viewpoint closer to what a robot actually experiences.
Company-reported benchmarks put its AtomVLA policy at a 97 percent success rate on the Libero robot-learning benchmark, HiMem-WAM at 97.7 percent, and its third-generation AIM world model at 93.1 percent on RoboTwin 2.0. Those figures come from Infiforce's own research disclosures rather than field deployments.
AtomBrain, DataGrid And A Multi-Form Fleet
AtomBrain is designed as a common intelligence layer that can drop into different bodies. The hardware lineup includes the AstroDroid wheeled humanoid, UltraDroid general-purpose robot, Little Atom bipedal robot and specialised Force systems. DataGrid, meanwhile, threads collection hardware, handheld grippers, egocentric recorders and teleoperation into a closed feedback loop where deployments generate the next generation of training data.
Where The Money Goes
Infiforce says its robots are being tested or commercialised across more than 30 Chinese cities and more than 100 scenarios, including manufacturing, logistics, warehousing and commercial services. It is also collaborating with CRRC High-Tech on embodied intelligence for infrastructure and has contributed to a proposed national specification on crowdsourced embodied-intelligence data. The round joins a Chinese funding wave for physical-AI startups this year — see our coverage of the LG and Nvidia physical AI pact, River AI's $1.1B raise and the broader Q2 uptick in North American robot orders.
Reporting based on coverage from The AI Insider and Infiforce's investor announcement.