Physical AI took $47.4B in six months. The risk moved to bills of materials.

Venture investors put $47.4 billion into physical AI companies across 521 deals in the first half of 2026, according to Crunchbase data, which is $5.5 billion more than the $41.9 billion the category took in over the whole of 2022 through 2024 combined ^[1][4][17]. Crunchbase counts robotics, autonomous vehicles, aerospace, drones, industrial automation and sensors in that definition ^[5], meaning the fastest-growing pool of AI capital is now going to companies with bills of materials rather than companies with inference bills.

The growth rate is real but the distribution is narrow. Dollars were up almost 4x from the $12 billion raised in the second half of 2025 and nearly 80 percent from $26.4 billion in H1 2025 ^[1][2][3]. Deal count moved far less: 521 deals is up 10.9 percent on H2 2025 and 19.5 percent on H1 2025 ^[19]. Average deal size therefore jumped to roughly $91 million from about $25.5 million six months earlier ^[18].

Strip out one round and the picture changes again. Waymo's $16 billion Series D in February, co-led by Alphabet, Dragoneer, DST Global and Sequoia at a $126 billion valuation, accounted for nearly a third of all physical AI venture dollars in the half ^[6]. Without it, the remaining 520 deals split $31.4 billion, an average of about $60.4 million, essentially flat against the H1 2025 average of $60.6 million ^[20][18]. Add Anduril's $5 billion in May at a $61 billion valuation, double its $30.5 billion mark from less than a year before ^[7], Shield AI's $2 billion Series G at $12.7 billion ^[8], and Saronic's $1.75 billion Series D at $9.25 billion ^[9], and four rounds account for $24.75 billion, or 52.2 percent of the half ^[21].

Exits skewed the same way, toward aerospace and defense rather than robotics ^[10][11]. SpaceX raised $75 billion in its June IPO at a $1.77 trillion valuation ^[10], which is 1.58 times the entire private funding total for the category in the half ^[22]. HawkEye 360 raised $416 million and drone maker Aevex $320 million in public debuts ^[11]. The clearest robotics outcome was Mobileye's roughly $900 million purchase of Tel Aviv-based Mentee Robotics, which the buyer explicitly tied to its own physical AI push ^[12].

The bull case, as put to Crunchbase News by Edison Partners general partner Ryan Ziegler, is that hardware is getting cheaper and more accessible while AI's ability to process sensor data at scale improves; his example is LIDAR scanners shipping in mobile phones ^[13][14]. Ziegler's interest is in traditionally analog sectors such as manufacturing, supply chain, utilities, agriculture, transportation and government, which he says resemble vertical software businesses with large deal values and multi-year deployments ^[15]. That is a defensible thesis for sensing. It says less about actuators, batteries, airframes and hulls, where the input that constrains a Saronic or an Anduril is factory throughput, not model quality. Nothing in the Crunchbase data addresses component cost or capacity, and no one quoted claims it does.

Three things to watch. Whether deal count catches up with dollars in H2, or whether the $91 million average deal is a permanent feature of a capital-intensive category ^[18]. Whether valuations keep doubling inside twelve months as Anduril's did ^[7]. And whether the next Mentee-style acquisition is priced for capability or for manufacturing capacity ^[12].