Physical AI Startups Just Raised $47.4 Billion in Six Months  -

Physical AI startups raised $47.4 billion across 521 deals in H1 2026, up nearly 80% from a year earlier, according to Crunchbase. Castelion, a hypersonic weapons startup, closed over $1 billion in Series C financing at a reported $13 billion valuation, highlighting the shift toward hardware and infrastructure. The funding surge is concentrated in a few large deals, with five transactions accounting for 73% of the $267 billion invested in Q1.

Physical AI Startups Just Raised $47.4 Billion in Six Months - Physical AI startups raised $47.4 billion across 521 deals in H1 2026, up nearly 80% from a year earlier. Castelion's $1 billion Series C at a $13 billion… For the last two years the AI funding story has been about models and the software wrapped around them. The first half of 2026 just moved the centre of gravity, and the number is big enough that it's worth stating plainly. Physical AI startups /category/startups , meaning robotics /category/robotics , autonomous vehicles, aerospace, drones, sensors and industrial automation, raised $47.4 billion across 521 deals in the first half of 2026. That's up nearly 80 percent from the same period a year earlier, and almost fourfold from the second half of 2025, according to Crunchbase. The Deal That Tells the Story Castelion, a hypersonic weapons startup, closed more than $1 billion in Series C financing at a reported $13 billion valuation, including a $250 million revolving credit component. It's the clearest signal of the shift. A company building hardware with substantial capital expenditure now attracts the kind of cheque that used to be reserved for software with near-zero marginal cost. ALSO, an autonomous small-vehicle company, added $150 million. Rillet raised $100 million at a $1 billion valuation for accounting software with agents embedded in the ledger. The deals cluster around a single theme: investors are paying up for companies that control hard-to-replicate systems, infrastructure, manufacturing capacity or regulated distribution, not for another layer of AI software. The Two-Speed Market The aggregate numbers look euphoric, and they're hiding something. PitchBook and NVCA put US startup funding above $400 billion for the first half, with AI companies and rounds of $100 million or more taking the overwhelming majority. But the distribution is brutal. In the first quarter, five transactions alone accounted for 73 percent of the $267 billion invested. This is not a broad reopening of venture capital. It's a sharp repricing of a small group of companies investors believe can become unusually large, and the definition of "unusually large" now includes building physical things. KPMG recorded $227.4 billion across 8,440 global deals in Q2, the second-highest quarterly total on record, and a disproportionate share of it went to the very biggest financings. What Changed The honest answer is the buyers changed. Software-era investors are now willing to finance factories and vehicles when the underlying market has strategic urgency and a large, reliable customer. Defense procurement, industrial automation, autonomous transport, these have demand that doesn't depend on getting a consumer to click something. Performance and supply availability matter more than gross margins. There's a caution buried in the same week of data. Newly listed defense contractor Lyntris fell 11.4 percent on its NYSE debut after pricing below range. Private investors will pay extraordinary prices for perceived scarcity and the pull of future procurement contracts. Public investors still price against earnings and rates. The gap between those two opinions is where the next correction in this cycle will live. The signal is still the signal, though. Capital that once flowed to software is now flowing to machines that do things in the real world. $47.4 billion in six months is not a blip. It's a direction. Sources: Crunchbase physical-AI funding estimates, August 2026; Tech Startups funding roundup, August 19, 2026; KPMG and PitchBook/NVCA venture data, August 2026. Get AI news in your inbox Daily digest of what matters in AI.