testing free note xx

Trinity Capital is keeping its loans on its books; Little is betting that even if some of the AI companies vanish, there will still be plenty of demand for the chips that secure the loans. Let’s say one of the neoclouds is forced into bankruptcy because it’s gotten its chips’ depreciation wrong, or for some other reason. Most of their customers may very well continue running their programs while banks repossess the servers and then sell them for pennies on the dollar. This is not the end of the world for the neocloud’s lenders or customers, though it’s probably annoying.

That situation will, however, bite Nvidia twice: first by flooding the market with its old chips, and second by reducing its number of customers. And if something happens that makes several of these companies fail at once, the situation is worse.

So how vulnerable is Nvidia?

The risky business of banking on GPUs

Part of what’s fueling the AI lending boom is private credit firms, which both need to produce returns for their investors and outcompete each other. If they miscalculate how risky the GPU loans are, they may very well get hit — and the impact could ripple out to banks. That could lead to widespread chaos in the broader economy.

Earlier, we talked about understanding interest rates as pricing risk. There is another, perhaps more nihilistic, way of understanding interest rates: as the simple result of supply and demand. Loans are a product like any other. Particularly for lenders that don’t plan on keeping them on their own books, pricing risk may not be a primary concern — making and flipping the loans are.