Motion Brain raises nearly 100 million yuan in Pre-A round: Fudan professor and former Intel chief scientist create 'brain' for robots

Motion Brain (眸深智能) completed a nearly 100 million yuan Pre-A extension round, with investors including Jinxing Capital, Chuanghehui Capital, and Xuhui Capital. Founded in January 2025 by Fudan University professor Chen Tao, former Intel China chief scientist Zhang Yimin, and serial entrepreneur Mu Zelin, the company aims to install native general-purpose brains on 1 billion robots. The team, with backgrounds from HiSilicon, Intel, and NVIDIA, is among the few with operator-level adaptation capabilities. Since 2022, the team has developed the MLD (Latent Motion Diffusion) model and MotionGPT, which decomposes human poses into 3000 motion tokens, enabling zero-shot generalization. In 2026, they released STI-WM, a Spatio-Temporal Integrated World Motion Model that reduces real-robot data needs by 90% while achieving 99% motion accuracy. The T2MB model enables offline self-evolution, improving task accuracy by up to 25%. Notably, NVIDIA's DAIR lab cited Motion Brain's MLD and MotionGPT in its latest ARDY model. The company has reduced end-side inference latency from 200ms to 10ms and cost from 200,000 yuan to 10,000 yuan by model compression and adaptation to domestic chips such as HiSilicon, Horizon Robotics, and Enflame. Motion Brain has generated revenue, with audited collections of 10 million yuan in 2025 and 30 million yuan in H1 2026, and expects 50 million yuan for the full year. Customers include industrial inspection, property, and sanitation companies.