Humanoid robot firms raise billions as industry grapples with teleoperation data limits

Flexion is developing a reinforcement‑learning and sim‑to‑real platform for humanoid robots and warns that the sector’s reliance on teleoperation for data collection is unsustainable. Human‑operated demonstrations grow only as fast as labor can be hired, yet the data needed to train physical AI systems is over 100,000 times smaller than that used for language and vision models. The approach also suffers from low‑quality inputs, as operators cannot feel or judge depth, limiting robot performance.

Meanwhile, LimX Dynamics announced a pre‑IPO financing round that raised nearly $200 million, bringing its total capital raised in the past six months to about $400 million. Investors from the United States, China, Europe, the Middle East and North America participated. The company plans to use the funds to deepen high‑level cognition integration, expand manufacturing capacity, and accelerate global deployment of thousands of autonomous humanoid robots. LimX’s technology stack includes a whole‑body motion control foundation model, vision‑language‑action capabilities, and an operating system for reasoning and physical action.