Thesis
The other 99.5%: why the next wave of AI is physical
Knowledge-work AI has its winners. In about three years, capital picked a leader at nearly every layer of the stack: chips and cloud, frontier labs, data, and the applications that lawyers, engineers and analysts now use every day. I watched it happen from inside Harvey. It was the fastest technology adoption I've seen, and it isn't slowing down.
It is also a wave with a ceiling. Goldman Sachs estimates that software accounts for less than 0.5% of global GDP. The companies that won that market are well known, well funded and valued accordingly. For an investor writing early checks today, most of the upside from that wave is already spoken for.
Where the rest of the economy lives
The other 99.5% is physical: factories, ships, aircraft, power grids, ports, mines and the supply chains that connect them. AI that can perceive, decide and act in that world, through machines, sensors and the systems that build them, is a far larger opportunity. Goldman's own industrials team says AI's impact there is only beginning.
It is also a harder market, which is why it is still early. Hardware takes longer to build. Customers buy more slowly. Mistakes have physical consequences. Venture capital is only starting to adjust: physical AI drew roughly one in every eight dollars of AI venture funding in the first half of 2026, and more than half of that went to four late-stage companies, by our analysis of Crunchbase data.
We believe the generational companies of this wave are being founded now, and that most of them are at pre-seed through Series A.
Why defense comes first
Every new technology needs a first customer willing to pay for capability before the unit economics are proven. For physical AI, that customer is defense. Global military spending reached a record $2.9 trillion in 2025, according to SIPRI, and the need for autonomy, sensing and resilient manufacturing is urgent, not theoretical. Defense budgets absorb early costs that commercial buyers won't, and the neo-primes now building in the sector are proving the path from a single program to a platform.
The commercial market follows. A drone that inspects a pier, a sensor that watches a power grid and software that rebuilds a factory floor serve both buyers. That is what dual-use means in practice, and it is why we see defense as the beachhead rather than the boundary.
What changes for investors
Backing physical AI well takes different skills from backing SaaS. Founders need help landing their first large customer, financing hardware without giving away the company, and working through regulated and government markets. The investors who add the most value will be the ones who know how those customers buy, and who sit close to them.
That is why 32 North is in San Diego, home to the largest concentration of military assets in the world, and why we back the founders building physical AI.