Physical AI Funding Hit $47.4B and Got Top-Heavy • The Weighted Average

Robotics, autonomy, and industrial-AI startups raised $47.4 billion across 521 deals in the first half of 2026, according to Crunchbase’s physical-AI funding analysis — nearly four times the $12 billion raised across 470 deals in the second half of 2025. The headline reads like a category-wide boom. The deal count says otherwise: transactions rose 11% while dollars rose 295%.

Divide it out and the real story appears. Average deal size went from $25.5 million to $91.0 million, a 3.6× jump in six months. Capital is not spreading across physical AI; it is concentrating inside it.

The averages hide two megadeals

Crunchbase attributes much of the spike to a handful of transactions. Waymo’s $16 billion Series D at a $126 billion valuation, closed in February, accounted for nearly a third of all physical-AI dollars in the half by itself. Anduril raised $5 billion in May at a $61 billion valuation, double the $30.5 billion it carried less than a year earlier — a round Crunchbase separately logged as a defense-tech venture funding record. Shield AI landed a $2 billion Series G at a $12.7 billion valuation in March, and Saronic raised $1.75 billion at $9.25 billion. Strip Waymo and Anduril alone and the remaining 519 deals share roughly $26.4 billion, or about $50.9 million each — still double the prior half-year average, but a very different market from the headline.

The concentration also skews toward defense and autonomy rather than the general-purpose robotics the category name implies. Crunchbase notes that exits have clustered in aerospace, defense, and drones more than in robotics, citing SpaceX’s $75 billion June IPO at a $1.77 trillion valuation, HawkEye 360’s $416 million debut, and Mobileye’s roughly $900 million acquisition of humanoid startup Mentee Robotics — a deal Mobileye’s own newsroom ties to its physical-AI push. A founder pitching warehouse manipulation is competing for attention in a category whose capital and liquidity are both being defined by autonomous vehicles and defense hardware.

Set the number against the wider market and the proportion clarifies further. Crunchbase separately reported that global venture funding hit a record $510 billion in H1 2026, exceeding all of 2025 combined. Physical AI’s $47.4 billion is therefore 9.3% of every venture dollar deployed worldwide — a meaningful slice, but one that shows the category riding a rising tide rather than reallocating capital away from software AI. That ratio is the number nobody published: it requires the category report and the global report together.

What a builder should actually take from this

For operators, the useful signal is not the total but the shift in what capital is buying. Edison Partners general partner Ryan Ziegler told Crunchbase the appeal lies in high-value analog industries where physical AI becomes mission-critical infrastructure, with hardware acting as “the distribution model for creating a data intelligence flywheel.” Eclipse Capital’s Joe Fath said funding is moving away from experimentation toward companies that hit production milestones and scale efficiently. Both descriptions point the same way: the money is going to deployment, not demos. Waymo’s own product cadence illustrates the shift — its engineering and operations updates now read as fleet-scaling and city-expansion notes rather than research milestones, which is what a company deploying a $16 billion round looks like from the outside.

That has a direct procurement consequence this quarter. A buyer evaluating a robotics or autonomy vendor should expect the well-funded ones to have moved past pilot pricing — and should read a $91 million average round as a warning that vendor economics now assume multi-year contracts and hardware amortization, not month-to-month trials. It also means smaller vendors in the long tail are competing against balance sheets they cannot match, which raises the odds that a promising pilot partner is acquired or repriced mid-deployment. Practically, that argues for contract terms that survive a change of control: assignment clauses, source or data escrow for anything embedded in a production line, and a documented exit path if the vendor’s roadmap is absorbed into an acquirer’s. The same capital-concentration dynamic drove the archive’s read of Bezos’s Prometheus reaching a $41 billion valuation on an industrial-AI thesis and the compute-side scramble behind OpenAI’s $20 billion move to break Nvidia’s grip.

What could break the thesis. Crunchbase’s category definition is broad — robotics, autonomous vehicles, aerospace, drones, industrial automation, and sensors — and a taxonomy that includes SpaceX and Waymo will always produce lumpy half-year totals. One $16 billion round landing in January rather than February would move the comparison by a third. The 3.6× average-deal figure is arithmetic on two half-year snapshots, not a trend line; a single quiet quarter would flatten it.

The deeper risk is that concentration signals late-cycle enthusiasm rather than maturation. Ziegler’s own argument is that build costs have fallen because compute, foundation models, and simulation are cheaper and sensors are commoditized. If capital requirements are genuinely dropping, a tripling average check size is the opposite of what efficiency should produce — which suggests the money is chasing a small number of proven winners rather than funding a broad build-out. The evidence that would settle it is H2 2026 deal count. If transactions climb toward 700 while average size falls back under $50 million, the category is broadening. If the count stalls again near 520, physical AI is four companies and a long tail, and operators should plan vendor risk accordingly. Today’s lead on agent search-API economics makes the software-side version of the same point: commoditizing capability does not automatically commoditize price.