In September 2025, Figure AI raised over $1 billion in Series C funding at a $39 billion post-money valuation. Eighteen months earlier, its Series B had valued the company at $2.6 billion — a 15x mark-up in a year and a half for a company that does not yet sell a humanoid robot at scale. Figure is not the outlier. It is the template for the most capital-intensive, lowest-revenue sector in venture right now.

The data

The numbers across the leaders are staggering in both directions. Humanoid and physical-AI startups raised $3.2 billion globally in 2025 — more than the previous six years combined, and the pace accelerated into 2026. Skild AI raised $1.4 billion led by SoftBank in January 2026 at a $14 billion valuation, roughly tripling its worth in seven months. Physical Intelligence — backed by Jeff Bezos and OpenAI — went from a $400 million Series A at a $2.4 billion valuation in late 2024 to seeking $1 billion at an $11 billion valuation in 2026.

Now hold those valuations against revenue. Skild AI has generated roughly $30 million in revenue against that $14 billion valuation — a multiple north of 450x. Physical Intelligence has no commercial product and no disclosed commercialisation timeline, with its co-founder emphasising a research-first approach. The bet investors are making is not on this year’s income statement. It is on a thesis: that the value in robotics sits in the software “brain” — a single foundation model that can drive any robot body — rather than in the hardware itself. Skild calls its model “omni-bodied”; Physical Intelligence builds vision-language-action models that turn a camera feed and a spoken instruction into joint movements.

Why it matters

The software-layer thesis is seductive because it promises the economics of AI (one model, infinite deployments) grafted onto the trillion-dollar addressable market of physical labour. But it imports a problem that pure-software AI never had: the real world doesn’t grade on a benchmark. A language model that is 95% reliable is useful. A robot that mishandles a task 5% of the time in a warehouse or a home is a liability, not a product. The gap between an impressive demo video and a machine an enterprise will pay for, repeatedly, is exactly where the previous generation of robotics companies stalled.

For founders outside the sector, the lesson is in the capital structure. These rounds are being raised at valuations that require near-flawless execution against a decade-long horizon. That is fine for Figure or Skild with billions in the bank, but it sets a reference price that strands everyone in the middle — the robotics startup that needs three more years and a real revenue ramp will find the bar has moved to “show me a foundation model,” not “show me a working product.”

The Charaka View

Manthan Intelligence’s pattern data has a name for this shape: the revenue-density inversion. In AI-native software, we have tracked companies reaching extraordinary revenue per employee — the value compounds faster than headcount. Humanoid robotics is the mirror image: enormous capital and valuation per dollar of revenue, with the income curve deferred years into the future. Both can be rational, but they fail differently. Software failures are cheap and fast; capital-intensive hardware failures are slow and expensive, and they take nine-figure rounds down with them.

Our read is that the winners here will be decided not by who has the best model in 2026, but by who first crosses the reliability threshold that turns a demo into a deployable unit with real switching costs. Until a humanoid company shows recurring revenue from machines doing real work — not pilots, not videos — the valuations are pricing a future that physics has not yet agreed to deliver. That is a bet worth watching closely, and sizing carefully.

This analysis draws on TechCrunch on Figure AI’s $39B Series C, TechCrunch on Skild AI’s $14B round, Crunchbase News on Physical Intelligence, PYMNTS on its 2026 raise, and CNBC on the humanoid funding wave. Human editorial oversight applied.

This analysis is informational and does not constitute investment advice, a research report, or a recommendation to buy, sell, or hold any security.

Charaka Notes by Manthan Intelligence. Subscribe