Physical Intelligence Valuation Nears $11B With No Revenue Plan

A valuation that doubled in four months

Physical Intelligence, a two-year-old company building AI models to control robots, is raising about $1 billion at a valuation above $11 billion. That would roughly double the $5.6 billion price set only four months earlier.2 The company was founded by AI academics and former Google DeepMind researchers. Its pitch is that general-purpose models can let robots handle many different tasks, rather than one robot being engineered for one job.2

The earlier price came from a $600 million round in late 2025.12 By March 2026 the company was reportedly back in the market.1 That pace is unusual even by the standards of the current AI funding cycle. It suggests investors see robot foundation models as a category where an early lead could turn into a lasting advantage.

Conflicting accounts of the deal

Reports do not agree on the details. Bloomberg described the $1 billion round as under discussion, citing people familiar with the matter.2 One later analysis said the round had not closed as of its late-August writing.1 A robotics trade outlet, however, reported that roughly $1 billion had closed at a valuation north of $11 billion. It said Founders Fund led the round, with Lightspeed Venture Partners joining as a new investor and Thrive Capital and Lux Capital following on.4

The accounts of the prior round also differ:

• Round label: One source calls the $600 million raise a Series B backed by OpenAI, NVIDIA, and Jeff Bezos.1 Another calls it a Series A.4
• Named backers: A third source lists Khosla Ventures, Sequoia Capital, and Thrive Capital among the investors at the $5.6 billion mark.3
• Cumulative funding: That same source says total funding passed $1 billion at that valuation. That figure appears to describe money raised to date, not the new round.3

The safest reading is this. The $5.6 billion valuation is confirmed. A roughly $1 billion raise at about $11 billion has been widely reported. Whether and when it formally closed depends on which account you trust.

No robots, and no revenue timeline

Physical Intelligence does not build robots.1 Figure AI and Unitree are competing on humanoid hardware. Physical Intelligence is instead trying to build the software brain that could run on many machines.1 One outlet compared it to "ChatGPT but for mechanical arms" learning tasks such as folding laundry and peeling vegetables.3

The company is also open about not having a commercialization timeline. Co-founder Lachy Groom, a Stripe veteran, says he does not give investors answers on when they will see a return. He has called it "a weird thing" that backers tolerate this.3 For a company valued in the billions, that is a notable admission. It shows how much of the valuation rests on expected future technical progress rather than current sales.

There are signs of a business model forming. Dealroom reports that Physical Intelligence licenses its models to several robot makers on a non-exclusive basis, including Ultra, a robots-as-a-service startup.5 Licensing to many hardware makers would fit the company's role as a supplier to the whole industry. Still, nothing reported here indicates how much revenue those deals bring in.

Investors are paying most for the AI layer

The wider funding picture helps explain the price. Ultra disclosed no valuation for its latest round. Dealroom's headline puts that round at $50 million, while its text says $62 million. In either case, it was far smaller than the money flowing to the AI providers Ultra relies on.5 Elsewhere, XPeng's robotics unit raised more than $900 million at a post-money valuation above $6.3 billion. Chipmaker SiMa.ai raised $150 million at $1.45 billion.5

Dealroom reads this as venture capital valuing the intelligence and compute layers of robotics well above hardware and deployment businesses. In that view, pricing power sits with foundation-model providers even as robots-as-a-service companies build recurring revenue.5 The trade press makes a similar point, describing Physical Intelligence's round as part of an intensifying race in robot foundation models.4

The bet investors are making

The likely explanation for the premium is a familiar venture logic. If one model becomes the default brain for robots, as large language models have for software, its maker could earn platform-scale returns. In that case, a lack of near-term revenue matters less than a technical head start.

The risks match that upside. Physical-world AI is harder to scale than text, because training data is costly and deployment involves expensive hardware. Physical Intelligence's model also depends on robot makers it does not control. A valuation that doubles in four months leaves little room for slow progress.

For now, investors appear willing to fund the research and wait for the commercial answers Groom says he does not provide.