A weekend rumor that Anthropic was buying robotics startup Physical Intelligence tore through AI Twitter this past weekend, and it kept spreading even after the startup’s CEO shook his head no. According to TechCrunch AI, the rumor wasn’t fully accurate, but it wasn’t fully wrong either. Anthropic and Physical Intelligence did hold acquisition talks this spring, per The Information, which means tech blogger Robert Scoble got the details wrong without being wrong that something real had happened.
What stands out here isn’t the rumor itself. It’s why a robotics deal set off this much noise in a year when AI acquisitions barely register anymore.
Why robotics, why now
Anthropic and OpenAI have both been on buying sprees, snapping up developer tools, services shops, and testing startups to turn model capability into enterprise revenue. TechCrunch AI reports Anthropic has made four known acquisitions this year. OpenAI has been far more aggressive, with at least 17 since 2023. Both are also preparing to go public, having confidentially filed for IPOs in early June that could become two of the largest U.S. stock debuts ever.
So a robotics play is a different kind of bet. The likeliest reason both labs want in: physical-world understanding may be a prerequisite for superintelligent systems, and no amount of internet text can teach a model what it feels like to move through the real world.
OpenAI’s own history proves the point. It built a robotic hand that solved a Rubik’s Cube, shut the whole robotics group down in 2021, then quietly rebuilt a humanoid lab and made “OpenAI Robotics” official in late May. Co-founder Wojciech Zaremba later said the original approach was missing pieces needed for real superintelligence. They came back anyway.
Two different playbooks
Here’s the contrast worth watching:
- OpenAI is building hardware. A humanoid lab in San Francisco, a near-term focus on robots for infrastructure work, and a stated long-term goal of a personal robot for everyone.
- Anthropic is testing, not building. Its internal safety group ran Project Fetch, checking how much Claude could help non-experts program a robot dog. A second phase in June found a newer model finished the same tasks roughly 20 times faster than the best human-plus-Claude team from a year earlier.
Buying a team like Physical Intelligence would let Anthropic skip years of hardware work. And Physical Intelligence is no obscure shop. It was co-founded by rising investor-operator Lachy Groom, has raised more than $1 billion, was reportedly in talks this spring for another billion at an $11 billion valuation, and its π0.5 model is one of the more widely used robot brains in research.
The complication nobody’s talking about
This is the part that makes the story more than gossip. Physical Intelligence’s investor base looks a lot like OpenAI’s, including Khosla Ventures, Thrive Capital, and Founders Fund. More pointedly, OpenAI is itself an investor in Physical Intelligence.
That’s significant because it raises a real question: did OpenAI’s early stake come with information rights or a right of first refusal over a sale to a competitor? Strategic investors negotiate exactly those protections for exactly this scenario. If Physical Intelligence is truly in play, OpenAI, already a shareholder and already close to Groom, may have the stronger claim. TechCrunch AI says OpenAI didn’t respond to questions about it.
What to take from this
- Watch robotics as the next frontier line. The competition is shifting from chatbots to embodied AI, and the labs are telling you so with their hiring and their checkbooks.
- Cap-table entanglements are becoming strategic weapons. When rivals invest in the same startups, an early check can quietly block a competitor’s acquisition later.
- Denials mean less than they used to. A gif of a character shaking her head is not a legal statement, and “not true” often means “not true yet.”
Expect more of these fights over the next year or two, and expect them to get less deniable. More details are available at the original source.