A weekend rumor that Anthropic would buy Physical Intelligence, a robotics startup last valued near $11 billion, drew a denial, yet the two firms did hold talks this spring.
Key Points:
- Reports confirm Anthropic and Physical Intelligence discussed a deal this spring, despite a public denial.
- Physical Intelligence has raised more than $1 billion and builds the widely used π0.5 robot model.
- OpenAI already holds a stake in Physical Intelligence, complicating any rival bid.
Physical Intelligence Talks
The frenzy began over the weekend, when tech blogger Robert Scoble posted on X that Anthropic was in the process of buying the startup. Later accounts confirmed the two sides really did weigh a purchase this spring.
Scoble missed the specifics, it seems, but caught that something real had stirred.
The denial itself was mild, just a Slack message from CEO Karol Hausman with a gif of a character from The Office shaking her head no. Co-founder Lachy Groom, for his part, stayed silent on the reports.
Physical Intelligence is no minor shop, having raised more than $1 billion already. It weighed another $1 billion round at an $11 billion valuation this spring and builds the π0.5 model that many research labs now run. The company took shape in San Francisco about two years ago, started by Groom, former Google researchers and professors from Stanford and Berkeley.
Also Read: Dow Jones Industrial Average Volatility Tests Bitcoin's Decoupling Story
Robotics Race Widens
Both labs have spent 2026 shopping hard, and both now aim to go public. Anthropic closed four known acquisitions and filed for an IPO on Jun. 1, while OpenAI bought at least 17 companies since 2023 and filed a week later, teeing up two of the largest U.S. debuts on record.
The pull toward robots is simple, since physical-world understanding may be a prerequisite for smarter systems, and no amount of internet text can substitute for it. There is a catch, though, because OpenAI already holds a stake in the target, alongside Khosla Ventures, Thrive Capital and Founders Fund. Whether that early money came with a right of first refusal on any sale remains an open question.
Anthropic's Robot Bets
Anthropic has not built a hardware lab to rival OpenAI's. It leans instead on published safety research that probes how far frontier models can push into real, physical tasks.
That work included Project Fetch last November, when staff tested how much Claude could help non-experts program a robot dog. A second phase in June found a newer model finished the same tasks about 20 times faster than the best human-plus-Claude team a year earlier. Buying a seasoned team would let Anthropic skip years of that slow, iterative groundwork.
Read Next: Meteorite That Hit A Home In New Jersey Fuels AI-Driven Hunt For Life