Thesis

Platforms doing essential physical work earn in good markets and bad.

Our focus is physical AI: the infrastructure, intelligence and machines through which AI performs physical work. Robots are one form among several, beside the intelligence that runs them and the energy, inference hardware and compute beneath them. We back two kinds of company: evergreen physical AI infrastructure, and high-margin, high-growth platforms in production. We start from work that customers must have done: a turbine needs repair whether technology stocks rise or fall. Where skilled workers retire and few replace them, machines keep essential work going. The customer pays for an outcome, and its value comes from the difficulty or danger of the task.

Labor is short, and the shortage is structural. US manufacturing could need as many as 3.8 million more workers by 2033, and 1.9 million of those jobs could go unfilled (Deloitte and The Manufacturing Institute, April 2024). Since March 2006, average hourly earnings in US manufacturing have risen 80 percent while the producer price of computers has fallen 61 percent (US Bureau of Labor Statistics). AI has sharply cut the cost of writing software, so moats are harder to hold there. In physical work the moats are defensible supply chains, complex hardware and software systems, economies of scale, and customer workflow integration, with its switching costs. Capabilities commoditize, and a deployed workflow does not.

Much of the AI market rests on one chain of spending, from model labs to cloud providers to chips, and one shortfall cascades through it. We pass if a company needs that financing cycle to continue without a break. We also pass on a valuation far above what public markets would pay, and on software or models with no direct role in physical work. A resilient company keeps its pricing power because its product is hard to substitute, and its value holds whichever AI model or lab wins.

We look for customer evidence, technical evidence, production economics, a compounding advantage, a capital strategy and a credible exit route. When financing tightens, well-capitalized buyers can acquire distressed assets, teams and data at deep discounts. Some resilient companies will be those buyers. When listings close, such buyers also become the main route to an exit.

We believe machines will do a growing share of essential physical work in the next decade. A platform in production that displaces labor reduces the customer's total cost of work that must be done. Its customers keep buying through a difficult market. Revenue that holds in good markets and in bad lets the company keep operating on its own cash flow. That quality strengthens the investment whichever way the AI cycle turns.

Our ambition is to back the physical AI companies that become the incumbents of the next decade. They will perform necessary work, reduce human exposure to dangerous tasks and keep essential systems running. When machines do essential work as a matter of course, physical AI becomes an engine of economic growth. We are moving from warehouses that run with the lights off to a world that can keep working after dark.