Anthropic's Robotics Ambitions Put OpenAI's Stake in Physical Intelligence Under the Microscope

What began as a weekend post on X by tech blogger Robert Scoble quickly ignited one of the more consequential rumors to ripple through AI circles in recent months: that Anthropic was in talks to acquire Physical Intelligence, a well-funded robotics startup headquartered in San Francisco. The rumor spread with unusual speed, partly because Physical Intelligence is no small player. Founded roughly two years ago by investor-operator Lachy Groom alongside former Google researchers and professors from Stanford and Berkeley, the company has raised over $1 billion and was reportedly in discussions this past spring for an additional $1 billion round that would have valued it at $11 billion. Its π0.5 model has become a widely referenced foundation in robotics research circles.

The denial that followed was, to put it diplomatically, underwhelming. Physical Intelligence CEO Karol Hausman reportedly addressed employees via Slack with a gif of a character from The Office shaking her head — not exactly the kind of firm, public rebuttal that tends to put rumors to rest. Lachy Groom, for his part, did not respond to press inquiries. And as reporting from The Information later clarified, the rumor wasn't entirely fabricated: Anthropic and Physical Intelligence actually did engage in acquisition conversations during the spring, meaning Scoble's post may have been wrong on the specifics while pointing at something real.

The timing is notable in an industry already dizzy with deal-making. Anthropic has completed four known acquisitions this year, while OpenAI has moved far more aggressively, acquiring at least 17 companies since 2023. Both companies are also now preparing for IPOs — Anthropic filed confidentially on June 1, with OpenAI following a week later — which only intensifies the pressure to demonstrate tangible, monetizable capabilities beyond raw model performance. Robotics represents one of the more credible frontiers for that kind of expansion, with the physical world increasingly seen as a necessary proving ground for systems aspiring toward general or superintelligent capability.

Anthropics's interest in physical-world AI isn't entirely new. The company published Project Fetch last November, testing how effectively Claude could help non-experts program a robot dog. A follow-up phase in June found that a newer Claude model completed the same tasks roughly 20 times faster than the best human-assisted team from the prior year — a striking benchmark that suggests internal appetite for robotics is growing. What Anthropic lacks, however, is a dedicated hardware and robotics engineering team of the kind OpenAI has been quietly assembling in San Francisco since 2024, when CEO Sam Altman formally announced an OpenAI Robotics division focused initially on infrastructure robots and eventually on personal robotics.

The most complicated dimension of this story may be OpenAI's own position. The company is an existing investor in Physical Intelligence, sharing early backers like Khosla Ventures and Thrive Capital — and Founders Fund is reportedly involved in Physical Intelligence's most recent funding round as well. That investment relationship raises real questions about whether OpenAI holds information rights or a right of first refusal that could complicate any acquisition by a direct competitor. If Physical Intelligence is genuinely in play, OpenAI may have both the financial stake and the strategic motivation to ensure it doesn't land in Anthropic's hands.