Robot AI maker Physical Intelligence's value grew from $2.4B to $5.6B

Physical Intelligence, a San Francisco robotics AI company founded in 2024, was valued at $5.6 billion in its most recent priced funding round, a $600 million Series B led by CapitalG on 20 November 2025, according to the 311 Unicorn Index. That is more than double the $2.4 billion post-money valuation reported after its $400 million round in November 2024.

The company builds general-purpose AI foundation models for robots, which it calls pi. A foundation model is a large model trained once on broad data and then adapted to many tasks; here the aim is one model that can drive different machines rather than one model per robot. This article sets out what the available funding records show, where those records disagree, and what remains unconfirmed.

What the latest confirmed figures show

The 311 Unicorn Index, last updated on 16 October 2026, lists Physical Intelligence at $5.6 billion as of 20 November 2025, based on the CapitalG-led $600 million Series B. A Series B is typically a company’s second major institutional round after its seed and Series A. The index counts only the valuation set by the latest priced round or company-run share sale, as announced by the company or its lead investor or reported by a major news organisation; talks that have been reported but not closed do not count.

The same index notes that talks at a valuation of over $11 billion were reported in March 2026, citing Bloomberg, and marks them as “not confirmed closed”. None of the sources reviewed for this article shows a later round or later reported talks. Until a closed round is announced or reported, $5.6 billion remains the most recent priced valuation in the record.

Stage	Date	Amount raised	Valuation
Seed	March (year as reported, 2024)	Not stated in sources	$400 million
Round reported by CNBC	November 2024	$400 million	$2.4 billion post-money
Series B, led by CapitalG	20 November 2025	$600 million	$5.6 billion
Reported talks (Bloomberg)	March 2026	Not stated	Over $11 billion, not confirmed closed

How the valuation moved between 2024 and 2025

In November 2024, coverage of the company’s $400 million raise said it lifted the post-money valuation to $2.4 billion, a figure the write-up says was confirmed by CNBC. Post-money means the company’s value including the new cash just invested. The same piece describes this as a large step up from the company’s March seed round, which valued it at $400 million. The sources give the seed round’s valuation only; none states how much money was raised in that seed round.

A separate report on the 2024 deal gives the same headline numbers: $400 million in funding at a $2.4 billion valuation. About a year later, the CapitalG-led Series B set the valuation at $5.6 billion. On those figures, the company’s valuation rose roughly sixfold from seed to the 2024 round, and a little more than doubled again in the following year.

The 2024 coverage frames the company as unusually young for its size at the time: it says the team had spent the previous eight months building a general-purpose AI model designed for robotics.

Who has backed Physical Intelligence

The investors named for the 2024 round include Amazon founder Jeff Bezos, OpenAI, Thrive Capital and Lux Capital, with additional backing from Bond Capital. The same coverage says Khosla Ventures and Sequoia Capital also support the company. A second write-up lists Bezos, OpenAI, Thrive Capital, Lux Capital and Bond Capital as the round’s backers, and describes Bezos as a lead angel investor investing personally through Bezos Expeditions.

For the 2025 Series B, the 311 Unicorn Index names CapitalG as lead investor. The sources do not list the other participants in that round.

OpenAI’s presence on the cap table is worth noting for readers who follow the model labs: one of the best-known developers of language models is a named backer of a company building models for physical machines. The sources do not describe any technical partnership between the two.

What the company is building: the π0 robot model

Physical Intelligence describes its mission as bringing general-purpose AI into the physical world. Coverage of the 2024 round says its team includes veterans of Tesla, Google DeepMind and X, and that its general-purpose robotics model is meant as a step towards artificial general intelligence, or AGI, described as AI that aims to match or exceed human intelligence across a broad range of tasks.

The model named in the sources is π0 (pi-zero), described as a general-purpose foundation model that powers various types of robots to perform complex tasks such as folding laundry and bagging groceries. The 2024 coverage says the company’s longer-term aim is to let users command robots to perform a wide range of tasks, from household chores like laundry to bussing tables or assembling boxes.

These descriptions come from funding coverage, not from a technical report or model card, and the sources give no benchmark results, model sizes or performance numbers for π0. Claims about what the model can do should be read as the company’s stated goals as relayed by that coverage.

Where the funding records disagree

The three data sources do not line up neatly, and readers comparing figures online will run into the same gaps.

• 2024 valuation. Two sources put the 2024 post-money valuation at $2.4 billion. GetLatka lists the valuation as $2.8 billion.
• Total raised. GetLatka says the company has raised $400 million in one round, a Series A in November 2024. That total does not include the $600 million Series B in November 2025 recorded by the 311 Unicorn Index, even though GetLatka’s page was updated on 10 September 2026.
• Round names. GetLatka labels the November 2024 raise a Series A; the other 2024 coverage does not assign it a series letter.
• Headcount. A 2024-round report gives a team of 40 to 80 people. GetLatka, citing LinkedIn, records 136 employees in September 2025 and 168 in November 2025.

Revenue is not public. GetLatka says Physical Intelligence has not published its revenue and that it has no sourced estimate. GetLatka lists Karol Hausman as CEO.

What it means for builders

For developers who work with AI models through APIs, the funding record matters mainly as a signal of where capital is flowing. A company focused on general-purpose models for robots has reached a multibillion-dollar valuation within roughly two years of its founding, backed by investors that include OpenAI.

The practical questions remain open in the sources:

• Access. None of the sources describes an API, a licence, open weights or a commercial product that outside developers can use today. The company’s website is listed as physicalintelligence.company, but the sources do not say what, if anything, it offers to external builders.
• Evidence of capability. The tasks cited, such as folding laundry and bagging groceries, come from funding write-ups. There are no independent evaluations in the sources.
• Business model. With revenue undisclosed, the sources do not show whether the company plans to sell models, robots or services.

The idea itself is the one to track: a single foundation model that can run on different kinds of robots, in the same way a single language model now sits behind many different apps. If that approach works, robotics could start to look more like software development, where teams build on a shared model rather than training their own from scratch. The sources describe that as the company’s aim, not as a demonstrated result.

What to watch next

• Whether a new round closes. Bloomberg’s March 2026 report of talks at over $11 billion has not been confirmed as closed. A closed round, announced by the company or its lead investor or reported by a major outlet, would replace $5.6 billion as the reference figure.
• Corrections to funding databases. GetLatka’s total of $400 million raised and its $2.8 billion valuation differ from other sources; updated records would settle the discrepancy.
• Technical disclosures. A model card, technical report or independent evaluation of π0 would give builders concrete evidence of what the model can do, beyond the task examples in funding coverage.
• Access for outside developers. Any announcement of an API, licence or partner programme would be the first point at which the company’s models become directly relevant to most builders.