Anthropic secretly pursued buying Physical Intelligence, and OpenAI may hold a blocking stake

Anthropic's Robotics Talks Expose AI's Shared Investor Problem

Physical Intelligence CEO Karol Hausman denied the rumor with a GIF: a character from The Office shaking her head no, delivered to employees over Slack 1. It was not a vigorous rebuttal. And it did not need to be, because The Information had already confirmed the substance underneath: Anthropic and Physical Intelligence, the robotics startup co-founded by investor-operator Lachy Groom, held acquisition talks this spring 1.

The talks are the headline. The complication buried underneath is that a clean deal between these two companies may be structurally impossible, and the reason has nothing to do with price or product fit.

Physical Intelligence's early investor base looks remarkably like OpenAI's own cap table. Khosla Ventures and Thrive Capital, two of the firms most prominently backing OpenAI, are also investors in Physical Intelligence 1. Founders Fund, another major OpenAI backer, was reportedly involved in Physical Intelligence's newest funding round earlier this year 1. And OpenAI is itself an investor in Physical Intelligence 1.

That last fact is where the situation gets tangled. If OpenAI's early investment came with information rights or a right of first refusal over a sale to a competitor, the kind of protective provisions that strategic investors routinely negotiate for exactly this scenario, then OpenAI could potentially block its biggest rival from acquiring a company it holds a financial stake in 1. Whether those provisions actually exist has not been confirmed publicly. TechCrunch asked OpenAI directly and the company did not respond 1. But the structural reality holds regardless of the legal specifics: the same venture firms are funding both sides of a potential acquisition and a potential block.

This is not unique to Physical Intelligence. The top AI labs draw from an overlapping pool of mega-funds, and as those labs push into adjacent territory, the conflicting claims multiply. Anthropic has made four known acquisitions this year. OpenAI has acquired at least 17 companies since 2023 1. Both confidentially filed for IPOs this summer, Anthropic on June 1 and OpenAI a week later 1. Each acquisition attempt now navigates a shared set of boardrooms where the same partners sit on both sides of the table.

Robotics is where this gridlock becomes concrete. OpenAI shut down its original robotics group in 2021 after building a robotic hand that could solve a Rubik's Cube, with co-founder Wojciech Zaremba later saying the approach was missing pieces needed for real superintelligence 1. The company relaunched the effort, with CEO Sam Altman officially announcing OpenAI Robotics in late May and describing a near-term focus on robots for infrastructure work and a personal robot for everyone as the long-term goal 1. Anthropic has taken a different path, publishing research in which its Claude model helped non-experts program a robot dog, with a June follow-up finding a newer model completed the same tasks roughly 20 times faster than the best human-plus-Claude team from the previous year 1.

Physical Intelligence, which has raised more than $1 billion and whose π0.5 model is apparently among the more widely used robot brains in robotics research, would let whichever lab acquires it skip years of ground-up development 1. But the path to that acquisition runs through investors who are equally, or more, committed to OpenAI winning the same race.

For founders and investors in AI, the takeaway is structural. Your cap table is simultaneously your competitive moat and your cage. The firms that funded your runway also funded the company that wants to buy you and the company that wants to block that sale. Embodied AI is the next frontier where these conflicts surface, and the deals that matter most may be the ones no one can cleanly execute.