Physical AI funding in H1 2026: dollars up fourfold, deal count nearly flat

Crunchbase data shows global venture funding into physical AI — robots, aircraft and intelligent machines tied to the physical world — reached US$47.4 billion across 521 deals in the first half of 2026. In the second half of 2025 the figures were US$12 billion across 470 deals; in the first half of 2025, US$26.4 billion across 436 deals. Dollars are up nearly fourfold on the prior six months while deal count rose only about 11%.

What average deal size is saying

Divide the Crunchbase figures directly: in H2 2025 the average physical AI deal was about US$25.5 million; in H1 2025, about US$60.6 million; in H1 2026, about US$91 million. This is not a broadening of the funded population but a concentration of capital in the leaders. The largest example is the US$12 billion financing of Prometheus, a physical AI company co-founded by Jeff Bezos and, per Crunchbase, the biggest North American deal of the half; next came US$1.7 billion for Travis Kalanick's Atoms.

Valuations are also moving fast. Crunchbase reports that Germany's Neura Robotics raised a US$1.4 billion Series C led by stablecoin issuer Tether, said to value it at US$7 billion, with a US$1 billion order pipeline; Shenzhen-based Astribot raised a US$148 million Series B led by a Wuxi-area innovation investment fund at a US$1.5 billion valuation. In H1 2026, robotics and new-generation AI labs were the leading sectors for additions to the Crunchbase unicorn board, which recorded 195 new unicorns in total.

The payers are no longer pure venture funds

Both examples share a trait: the round was led by an institution with strategic or industrial motives, not a purely financial venture fund. There is data behind the pattern. Global Corporate Venturing reports that roughly one in five startup rounds now includes a corporate backer, with 5,038 corporate-backed deals in 2025 worth US$233.8 billion, up about 70% year over year. PitchBook finds corporate investors present in 87.9% of US AI venture deal value in 2026 to date, even as their share of deal count declines.

For physical AI the structure is logical. Hardware companies need lines, tooling, certification, a supply chain and first industrial customers — things a manufacturing group or a locally anchored industrial fund can supply faster than pure cash. The consequence for founders: they are assessed on order book and deployment capability, not user-growth curves.

The risk is working capital, not the model

The weakness of this cohort is cash trapped in inventory, tooling and the long collection cycles of industrial customers. Equity used to buy fixed assets is the most expensive capital available. With the average deal now around US$91 million, the gap between an emerging-market hardware seed round and the capital needed to reach the next round widens, because later-stage buyers are now used to profiles with large order books.

• Signed orders or quantified letters of intent — Neura Robotics disclosed a US$1 billion order pipeline when raising at US$7 billion.
• A bill of materials and a cost-down curve at 100 and 1,000 units.
• Who owns the intellectual property and the tooling once a strategic investor is on the cap table.
• A blended capital structure: equity for R&D, leasing or debt for equipment and inventory.