## Did Anthropic Really Try to Acquire Physical Intelligence?

Yes — and the complication is that OpenAI may be positioned to block it. According to a report from The Information, Anthropic and [Physical Intelligence (π)](https://humanoidintel.ai/companies/physical-intelligence) held acquisition talks earlier this spring. A weekend rumor that a deal was imminent was denied swiftly by Physical Intelligence CEO Karol Hausman — via a Slack message featuring a GIF from *The Office* — but the denial addressed timing, not the existence of talks. The distinction matters enormously. Physical Intelligence has raised more than $1 billion, counts former Google researchers among its founders, and its π0.5 model is widely used across the robotics research community. More critically: OpenAI is itself an investor in Physical Intelligence, and may hold protective provisions — information rights or right of first refusal — that give it direct leverage over any sale to Anthropic. That investor dynamic, not the rumor itself, is the story worth tracking.

---

## Why Every Major AI Lab Now Wants a Robotics Foothold

The strategic logic is not complicated, even if execution is. Both Anthropic and OpenAI are operating under the same core thesis: that [physical AI](https://humanoidintel.ai/glossary/physical-ai) — systems that understand and manipulate the real world — may be a prerequisite for building superintelligent systems. Text scraped from the internet has limits. Real-world interaction, [dexterous manipulation](https://humanoidintel.ai/glossary/dexterous-manipulation), and embodied feedback loops provide a qualitatively different training signal.

OpenAI's own history makes this tension concrete. The company built a robotic hand capable of solving a Rubik's Cube, then shut down its entire robotics division in 2021. Co-founder Wojciech Zaremba later acknowledged the approach was missing key elements for real superintelligence. The team quietly returned in 2024, establishing a humanoid robotics lab in San Francisco. In late May, CEO Sam Altman formalized the effort — announcing OpenAI Robotics with a stated near-term focus on robots for infrastructure work.

Anthropic's path has been different. The company has published robotics-adjacent research through an internal safety group rather than building hardware. Last November, Anthropic ran Project Fetch, testing how much Claude could help non-experts program a robot dog. A second phase in June found that a newer Claude model completed the same tasks roughly 20 times faster than the best human-plus-Claude team from the prior year. That's a striking internal benchmark — but it's still software capability grafted onto someone else's hardware. Acquiring Physical Intelligence would buy Anthropic the robotics team, the institutional knowledge, and the π0.5 model's existing footprint in the research community — compressing what would otherwise be years of hiring and infrastructure build-out.

---

## The OpenAI Investor Problem Is the Real Variable

Here is where the story gets structurally interesting and genuinely unresolved. Physical Intelligence's early investor base includes Khosla Ventures, Thrive Capital, and Founders Fund. All three are also major investors in OpenAI. More directly: OpenAI itself holds equity in Physical Intelligence.

Large strategic investors frequently negotiate protective provisions — information rights that allow them to see deal terms before they're finalized, or rights of first refusal that give them the option to match any acquisition offer. The specific terms of OpenAI's investment in Physical Intelligence have not been disclosed publicly. But if those provisions exist, Anthropic would be attempting to acquire a company whose largest strategic investor is its primary competitor — and that competitor would have both the financial resources and the obvious motivation to intervene.

The source material notes that OpenAI is already close to Physical Intelligence co-founder and investor-operator Lachy Groom, already attempting to out-position Anthropic in robotics, and was contacted for comment on these questions but did not respond. That non-response is itself data. Companies with nothing to protect generally say so.

---

## IPO Timelines Are Compressing the Decision Window

There is an urgency layer here that transcends robotics strategy. The source material reports that Anthropic confidentially filed for an IPO on June 1, followed by OpenAI filing a week later. Both companies are moving toward what could be among the largest U.S. stock debuts in recent history. Pre-IPO, the incentive to acquire transformative capabilities — and to deny competitors the same — is acute. Post-IPO, shareholder scrutiny and dilution concerns make large acquisitions harder to execute quickly.

That timeline compression explains why a rumor with no confirmed deal still carries a newsworthiness score of 100: the acquisition window, if it exists, is narrow and closing. Physical Intelligence's [vision-language-action model](https://humanoidintel.ai/glossary/vision-language-action-model) capabilities and its position as a foundational AI layer for humanoid robot developers make it arguably the most strategically valuable independent robotics AI company currently available. The question is not whether someone will try to acquire it — it's who gets there first, and whether OpenAI's cap table position constitutes a structural veto.

---

## What This Means for the Broader Humanoid Stack

For humanoid robot manufacturers — everyone from [Figure AI](https://humanoidintel.ai/companies/figure-ai) to [Agility Robotics](https://humanoidintel.ai/companies/agility-robotics) to the wave of Chinese entrants — the Anthropic-PI dynamic is a vendor risk signal. Physical Intelligence's π0.5 model is embedded in research pipelines across the industry. If it gets absorbed into either Anthropic or OpenAI, the model's availability, licensing terms, and development roadmap become hostage to a parent company's competitive interests. Labs currently building on π0.5 should be stress-testing their dependency assumptions now, not after a deal closes.

More broadly, this episode clarifies that the AI-for-robotics layer is no longer a niche research problem. It is a strategic asset that foundation model companies are willing to pay acquisition prices for. That raises the floor for every independent robotics AI company — including [Skild AI](https://humanoidintel.ai/companies/skild-ai) — and accelerates consolidation pressure across the sector.

---

## Key Takeaways

- Anthropic and Physical Intelligence held real acquisition talks this spring, per The Information — the weekend rumor was denied, but the underlying talks were not.
- Physical Intelligence CEO Karol Hausman denied the rumored imminent deal via a *The Office* GIF on Slack; he did not deny that talks occurred.
- OpenAI is an investor in Physical Intelligence and may hold protective provisions — right of first refusal or information rights — that complicate any Anthropic acquisition.
- Physical Intelligence's investor base (Khosla Ventures, Thrive Capital, Founders Fund) overlaps substantially with OpenAI's investor base, creating alignment pressure against an Anthropic deal.
- Both Anthropic (filed June 1) and OpenAI (filed a week later) have confidentially filed for IPOs, compressing the window for pre-IPO acquisitions.
- OpenAI formally relaunched its robotics division in late May under CEO Sam Altman, with a stated focus on infrastructure robots.
- Physical Intelligence's π0.5 model has broad adoption in robotics research; its acquisition would create vendor dependency risk for humanoid manufacturers currently building on it.

---

## Frequently Asked Questions

**Did Anthropic actually try to acquire Physical Intelligence?**
According to The Information, yes — the two companies held acquisition talks earlier this spring. Physical Intelligence CEO Karol Hausman denied that a deal was imminent (via a Slack message with an *Office* GIF), but reporting indicates the talks themselves were real, not fabricated.

**Why would Anthropic want a robotics AI company?**
Anthropic operates under the thesis that physical-world understanding is necessary for building superintelligent systems. Physical Intelligence offers a team with deep robotics AI expertise, an existing foundation model (π0.5) with broad research adoption, and co-founders from Google's robotics research community — capabilities that would take Anthropic years to build internally.

**Could OpenAI block an Anthropic acquisition of Physical Intelligence?**
Potentially. OpenAI is an investor in Physical Intelligence and may hold protective provisions such as a right of first refusal over a sale to a competitor. The specific terms of OpenAI's investment have not been disclosed, but if such provisions exist, OpenAI would have the legal mechanism and obvious motivation to intervene.

**What is Physical Intelligence's π0.5 model?**
Physical Intelligence's π0.5 is a foundation model for robot learning widely used in robotics research. The source material describes it as broadly adopted across the robotics research community. Its acquisition by either Anthropic or OpenAI would transform it from an independent research resource into a proprietary asset.

**How does this affect humanoid robot companies building on Physical Intelligence's technology?**
Any acquisition introduces vendor risk. If π0.5 is absorbed by a foundation model company with its own competitive robotics agenda, licensing terms, model availability, and development priorities could shift materially. Humanoid manufacturers with deep dependencies on Physical Intelligence's stack should assess alternative or redundant AI layer options now.