A rumor that Anthropic is buying robotics startup Physical Intelligence tore through AI Twitter over the weekend, gaining speed even after the target company’s own chief executive said it wasn’t happening. The denial did not stop the story. It may have made it more interesting.
The claim traces to a post from tech blogger Robert Scoble, whose account of an Anthropic acquisition set off a wave of reposts and commentary before anyone at either company confirmed a word of it. Physical Intelligence CEO Karol Hausman told employees the reports were false, according to The Information, using a Slack message with a gif of a character from “The Office” shaking her head no. Company co-founder Lachy Groom did not respond to a request for comment from TechCrunch, sent Monday night.
Physical Intelligence is not a company that rumors attach to lightly. Co-founded by Groom, a Silicon Valley investor-operator with an expanding profile, the startup has raised more than $1 billion and was reportedly courting another billion-dollar round this spring at an $11 billion valuation. Its π0.5 model has become one of the more widely cited robot foundation models in the field, the kind of asset that would let a buyer skip years of robotics groundwork rather than build it from scratch.
Here is the detail that keeps the rumor from being pure noise: Anthropic and Physical Intelligence did hold acquisition talks this spring, The Information reported separately. Scoble’s post may have gotten the outcome wrong while still landing near something real, which is exactly the pattern that makes a denial hard to read as final. A “no, we are not being acquired” from Hausman rules out an active, signed deal. It does not rule out past conversations, a revived approach, or interest from a different suitor entirely, and Hausman’s statement addressed none of those possibilities directly.
The plausibility of the rumor is doing real work here. Anthropic has spent the past year publishing research through an internal group that stress-tests frontier models against physical tasks, including a project last November testing how much Claude could help a novice program a robot dog, and a follow-up in June that the company says cut completion time roughly twentyfold versus the best human-and-Claude pairing a year earlier. Robot foundation models are widely viewed inside the industry as the next capability frontier once internet-scale text stops yielding gains, which gives a rumor like this a plausible motive even without confirmation.
There is a complication that gets less attention than the denial itself. OpenAI is an investor in Physical Intelligence, alongside Khosla Ventures and Thrive Capital, and Founders Fund was reportedly part of the startup’s most recent funding round. A sale to Anthropic would hand a strategic asset to OpenAI’s chief rival, which raises the question of whether OpenAI’s investment carries information rights or a right of first refusal, the kind of protective term investors negotiate for exactly this situation. TechCrunch asked OpenAI about it and got no response.
None of this confirms a deal. It confirms that the rumor cycle now moves on partial information and outruns the fact-checking that would normally settle it, and that a Slack gif is what currently passes for an official denial in this market. Anyone tracking robotics consolidation should watch whether Physical Intelligence opens a new funding round in the next quarter, since a fresh raise would be the clearest signal that no sale is imminent.
Reported by TechCrunch (Connie Loizos) on July 21, 2026.
