Wall Street for the Physical Buildout: what YC wants and how to validate it

What YC's request says

• Banks currently underwrite a robot much like a forklift.
• Founders are already inventing new instruments, such as loans against robots that each earn about $500k a year.
• Options on GPU capacity are one example of a hedging product YC mentions.
• Past buildouts like railroads, electricity, and telecom produced large financial companies.

Why now

Every infrastructure buildout needs capital that matches its assets. Railroads created bond markets, electrification created utility finance, and telecom created tower and spectrum financing. Physical AI is starting the same cycle, but the financial tools have not caught up.

The gap is underwriting. A traditional lender does not know how to value a robot that does skilled work, a GPU whose resale price moves with each chip generation, or a data center lease tied to AI demand. Whoever models those assets well can lend profitably where banks will not.

What makes it defensible

Data. A lender that tracks real utilization, revenue per asset, and resale prices across many deployments can price risk better than anyone else, and that advantage compounds with every loan.

This is also a business where software alone is not enough. You need capital partners, credit expertise, and the discipline to say no to bad deals while the market is hot.

Three startup ideas to start from

How to validate it before you build

1. Talk to robot and GPU buyers about how they finance purchases today and what they were denied.
2. Build a depreciation and resale model for one asset class and stress test it against a downturn.
3. Line up a credit partner or warehouse facility before you promise loans.
4. Close a few small deals by hand and track repayment before automating anything.

The founder trap

Growing the loan book faster than your understanding of the asset. In lending, the losses show up later than the growth, so early discipline matters more than speed.