Source Robotics (Archon Robotics) launches with hundreds of millions in seed funding for general full-body embodied brain.
The AMW Read
New entrant with a novel whole-body approach and top-tier founding team; seed round size and open-source plan signal segment-level ambition, but market divergence remains unresolved.
Source Robotics (Archon Robotics) launches with hundreds of millions in seed funding for general full-body embodied brain.
Professor Li Hongyang of the University of Hong Kong has launched Source Robotics (Archon Robotics), a startup focused on developing a general full-body embodied AI brain for humanoid robots. The company secured a seed round of "hundreds of millions of yuan" (RMB 100M+ ~ $14M+) from investors including ZhenFund, Gaorong Capital, IDG Capital, and 5Y Capital. The funds will be used for R&D of a whole-body humanoid base model, multimodal motion data collection, talent expansion, and open-source model release within 2026.
Why it matters: This launch enters the robotics segment at a critical divergence moment. The team's background in autonomous driving (UniAD, CVPR 2023 Best Paper) and mass-produced ADS systems signals a potential leapfrog in whole-body training data and coordination, addressing the current structural gap where most embodied solutions lack whole-body motion data. The company's emphasis on an open-source base model positions it as a new entrant aiming to accelerate the whole-body intelligence ecosystem.
Grounded take: Source Robotics exemplifies the talent-and-capital convergence pattern seen in Chinese robotics: top academic researchers spinning out with strong investor backing to tackle the hard problem of general whole-body manipulation. The challenge lies in data acquisition at scale - the article notes real humanoid data costs hundreds to nearly a thousand yuan per hour. If the team can leverage its autonomous-driving experience to synthesize or simulate whole-body training data efficiently, it could move the segment beyond desk-grasping tasks. This is a bet on data infrastructure as the true moat in embodied AI.
