The Robotics Arms Race Between Anthropic and OpenAI

Physical Intelligence is far from a typical robotics venture. Co-founded by investor Lachy Groom and backed by over $1 billion in capital, the company is valued at roughly $11 billion. Its π0.5 model has become a staple in robotics research, making it a prime asset for any firm looking to ground large language models in physical reality. Although the specific acquisition rumor proved inaccurate, reports confirm that Anthropic and Physical Intelligence held exploratory talks earlier this spring.

Strategic interest in robotics stems from the belief that true superintelligence requires more than internet text; it demands an understanding of the physical world. OpenAI previously disbanded its robotics group in 2021, only to revive its hardware ambitions this year with a new San Francisco-based lab. Anthropic, meanwhile, has focused on research projects like Project Fetch, which tests how Claude can assist in programming robotic hardware.

The competitive landscape is further complicated by overlapping interests. OpenAI is a direct investor in Physical Intelligence, sharing a cap table with firms like Khosla Ventures and Thrive Capital. This financial entanglement raises questions regarding potential rights of first refusal or information clauses, which could grant OpenAI significant leverage should the startup decide to sell. As both Anthropic and OpenAI move toward their respective IPOs, the race to secure the best robotics expertise has become a critical front in their battle for dominance.