Did Anthropic Try to Buy Physical Intelligence? What We Know

A weekend post on X by tech blogger Robert Scoble claiming Anthropic was acquiring robotics startup Physical Intelligence spread rapidly across AI circles, even after Physical Intelligence's CEO denied it via a Slack message. According to The Information, the rumor wasn't entirely wrong: the two companies did hold acquisition talks this spring. Physical Intelligence has raised over $1 billion, was reportedly in talks for another $1 billion round at an $11 billion valuation, and counts OpenAI itself as an investor, which creates a thorny conflict if a rival deal moves forward.

What happened

Detail	Fact
Rumor source	Robert Scoble post on X, weekend of July 19-20, 2026
Confirmed by	The Information: talks did occur this spring
Physical Intelligence funding raised	Over $1 billion total
Reported new round valuation	$11 billion
Anthropic acquisitions in 2026	4 known
OpenAI acquisitions since 2023	At least 17
Anthropic IPO filing	Confidentially filed June 1, 2026
OpenAI IPO filing	Confidentially filed one week after Anthropic

Tech blogger Robert Scoble posted on X over the weekend claiming Anthropic was acquiring Physical Intelligence, a San Francisco robotics startup. The post spread fast. Physical Intelligence CEO Karol Hausman responded by telling employees in Slack that the reports were not true, attaching a gif of a character from “The Office” shaking her head. That was the extent of the denial.

The Information then reported that Anthropic and Physical Intelligence had, in fact, held acquisition talks this spring. Scoble may have gotten the timing or status wrong, but something did happen. Lachy Groom, Physical Intelligence’s co-founder, did not respond to TechCrunch’s request for comment.

Why does Physical Intelligence matter to Anthropic?

Physical Intelligence was co-founded roughly two years ago by Groom alongside former Google researchers and professors from Stanford and Berkeley. Its pi0.5 model (pronounced “pi-zero-point-five”) is reportedly one of the more widely used neural network architectures for controlling robots in research settings. The startup’s early backers include Khosla Ventures and Thrive Capital, both also investors in OpenAI.

Anthropic has no hardware lab to speak of. What it does have is a research group that stress-tests its frontier models for safety. That group ran “Project Fetch” last November, where Anthropic staff tested how well Claude could help non-experts program a robot dog. A second phase in June 2026 found that a newer model completed the same tasks roughly 20 times faster than the best human-plus-Claude team from the prior year. That’s promising research. It is not a robotics team.

Buying Physical Intelligence would let Anthropic skip years of team-building. The broader logic: physical-world experience may be a prerequisite for genuinely superintelligent systems, and no amount of text scraped from the internet can replicate it.

OpenAI’s conflicting position

Here’s where it gets complicated. OpenAI is itself an investor in Physical Intelligence. That means if Anthropic were serious about an acquisition, OpenAI would not just be a competitor watching from the sidelines. It would be a stakeholder with a potential say in the outcome.

Large strategic investors sometimes negotiate for information rights or a right of first refusal (the right to match any acquisition offer before a competitor can close). Whether OpenAI secured those terms is not publicly known. TechCrunch asked OpenAI directly; the company did not respond.

OpenAI’s own robotics trajectory adds context. It built an early robotic hand that could solve a Rubik’s Cube, then shut down its entire robotics group in 2021. Co-founder Wojciech Zaremba later said the approach was missing pieces needed for real superintelligence. The team quietly reformed in 2024, building a humanoid robotics lab in San Francisco. In late May 2026, CEO Sam Altman made it official, announcing “OpenAI Robotics” was hiring and describing near-term focus on infrastructure robots, with a personal robot for everyone as the long-term goal.

Both Anthropic and OpenAI have now filed confidentially for IPOs, setting up what could be two of the largest U.S. stock market debuts in history. That context makes every significant acquisition more consequential: each company is building the story it will tell public investors.

For more on how AI labs are integrating capabilities into products, our coverage of OpenAI’s frontier model testing shows how far these systems have come and how much further they’re being pushed. If you’re thinking about what AI integration looks like for a smaller business, our AI integration service covers practical starting points.

Our take

The Slack gif denial is a tell. When a CEO wants to kill a rumor dead, they issue a statement, not a reaction gif. Hausman’s response suggests either that the talks are genuinely over and he didn’t feel they deserved a serious denial, or that the situation is still fluid enough that a strong statement felt premature. Neither reading is reassuring for anyone hoping for a clean “nothing to see here.”

The more interesting story is OpenAI’s investor position. If OpenAI holds standard protective provisions, Anthropic may have spent months in negotiations with a company it could never actually close. That would be an expensive way to learn that your rival already had the inside track. Watch for whether Founders Fund, listed as involved in Physical Intelligence’s newest round, plays any coordinating role. The investor overlap between these companies is not accidental, and it will shape how the robotics consolidation plays out.

For businesses tracking AI developments, the practical takeaway is simpler: robotics is becoming a serious battleground for the frontier labs, not a side project. Products built on top of these models, from automation workflows to workflow integrations, will eventually have a physical-world layer. Plan for it.

Source: TechCrunch · AI