Physical AI’s record half was mostly four cheques

Physical AI startups raised $47.4 billion across 521 deals globally in the first half of 2026, according to Crunchbase data published August 18. The comparisons are the headline: $12 billion across 470 deals in the second half of 2025, and $26.4 billion across 436 deals in the first half of 2025. The whole of 2022 through 2024 produced $41.9 billion — several billion less than this single half.

Crunchbase’s definition is broad. Physical AI here covers robotics, autonomous vehicles, aerospace, drones, industrial automation and sensors. That matters for reading the number, because the money did not land evenly across those lines.

Dollars quadrupled; deal count did not

Deal count went 436 → 470 → 521 across the three halves. That is roughly 19% growth over a year, the kind of curve a healthy sector produces. Dollars over the same span went $26.4B → $12B → $47.4B. Those two series are not describing the same phenomenon.

The reconciliation is at the top of the table. Waymo’s $16 billion Series D in February, at a $126 billion valuation and co-led by Alphabet, Dragoneer, DST Global and Sequoia, took close to one third of every venture dollar in the half by itself. Anduril added $5 billion in May at a $61 billion valuation — double where it stood less than a year earlier. Shield AI raised $2 billion in March at $12.7 billion, co-led by Advent International and JP Morgan Chase. Saronic took $1.75 billion in March at $9.25 billion. Those four rounds total $24.75 billion, more than half the half.

Three of the four are defence companies. The fourth is Waymo. None of them builds a humanoid, and none of them is a robotics company in the sense the term is usually used in this sector.

What the shape of the money says

A sector where the median company is thriving produces rising deal counts at rising average sizes. A sector where a handful of category winners are being pre-funded to industrial scale produces this: flat-ish deal flow, one round worth a third of the total, and valuations doubling inside twelve months at the top.

Joe Fath, partner and head of growth at Eclipse Capital, told Crunchbase that funding is shifting away from experimentation and toward companies that can hit production milestones, land customers and scale efficiently — in his words, “customers value operational efficiency, reliability, and revenue, not technical sophistication alone.” The data supports that reading, and it has an uncomfortable corollary for the middle of the market: capital that concentrates on proven scaling is capital that is not available to the company still proving the machine works.

The exits point the same direction. Crunchbase notes they have been more concentrated in aerospace, defence and drones than in robotics. The most notable robotics exit was Mobileye paying about $900 million for Mentee Robotics — a trade sale of a humanoid team into an automotive supplier, not a robotics company reaching public markets on its own numbers.

Why it matters

Two conclusions are safe from this data. First, the pool of capital available to physical AI is genuinely larger than it was, and that is not an artefact of one round — strip Waymo out entirely and H1 2026 still leaves $31.4 billion, roughly 19% above H1 2025. Second, the average robotics company’s odds did not improve by anything like the headline multiple, because the increment went to four addresses.

For anyone raising into this market, the operative question is not whether physical AI is well funded. It is which of Crunchbase’s six categories a given company actually sits in, and whether it can point to production milestones and named customers. That is the filter the money is currently applying.