A weekend rumor that Anthropic was acquiring robotics startup Physical Intelligence spread unusually fast for what should have been just another entry in a year already packed with AI acquisitions, and it turned out to have more substance behind it than a typical internet rumor usually does.
Part of the reason it caught on has to do with who's involved. Physical Intelligence isn't an obscure lab; it was co-founded by investor-operator Lachy Groom, has raised more than a billion dollars, was reportedly in talks this spring for a new round at an eleven-billion-dollar valuation, and its foundation model is regarded as one of the more widely used "robot brains" in robotics research today. That combination of profile and capital made a buyout claim plausible enough to travel well beyond the usual AI-rumor crowd.
According to reporting from The Information, Anthropic and Physical Intelligence did hold acquisition talks earlier this spring, meaning the tech blogger whose post set off the weekend frenzy may have gotten some specifics wrong without being wrong that discussions had actually taken place. Physical Intelligence's public response to the swirl wasn't exactly forceful: CEO Karol Hausman reportedly told employees the reports weren't accurate via a Slack message featuring a gif of a character from The Office shaking her head no, a denial that reads more like a shrug than a firm rebuttal. Groom did not respond to a request for comment.
Both Anthropic and OpenAI have been unusually acquisitive this year, snapping up developer tooling, AI services shops, and product-testing startups aimed at converting raw model capability into enterprise revenue faster than the other side can match. Anthropic has made four known acquisitions so far; OpenAI has made at least seventeen since 2023. Both companies are also now preparing to go public, having confidentially filed for IPOs within a week of each other, setting up what could become two of the largest U.S. stock market debuts in history.
The more interesting question sitting underneath the rumor is why an AI lab focused on language models would want a robotics company at all. The likeliest explanation is that physical-world understanding may turn out to be a prerequisite for genuinely capable AI systems, and that no volume of internet text can substitute for the kind of grounding that comes from acting in and observing the physical world directly.
OpenAI's own history with robotics is instructive here. The company built an early robotic hand capable of solving a Rubik's Cube, then shut its entire robotics group down in 2021, with a co-founder later saying the approach was missing pieces needed for genuine advanced intelligence. That team quietly returned in 2024, building a humanoid robotics lab in San Francisco before the company's robotics efforts became public again this year with a stated near-term focus on robots for infrastructure work and a longer-term ambition around personal robots. Anthropic, by contrast, hasn't built anything resembling a hardware lab, but has published research through an internal group that stress-tests frontier model capabilities for safety purposes, including work testing how much its models could help non-experts with physical tasks, with a later phase reportedly finding a newer model completed similar tasks roughly twenty times faster than the best human-plus-model team from the prior year.
Buying an existing robotics team would let Anthropic skip years of foundational work it hasn't otherwise invested in. There's a complication, though: Physical Intelligence's investor base overlaps heavily with OpenAI's own, including firms like Khosla Ventures, Thrive Capital, and Founders Fund, and OpenAI itself holds a stake in Physical Intelligence as an investor. That raises a real question about whether OpenAI's early investment came with information rights or a right of first refusal over any sale to a direct competitor, the kind of protective provision strategic investors sometimes negotiate for exactly this scenario. If Physical Intelligence genuinely is in play, OpenAI, as an existing shareholder already close to Groom and clearly motivated to keep pace with Anthropic in robotics, may have a stronger claim to it than Anthropic does. OpenAI did not respond to questions on the matter.