The loop
Nvidia CFO Colette Kress told analysts on August 26 that labs Nvidia finances will make up roughly 25% of its business next year. The company has invested nearly $50 billion in AI labs and lined up commitments for more than $500 billion.
Here's how it works: Nvidia invests in a lab or arranges financing. The lab uses the money to build a data center. The data center is filled with Nvidia chips. The purchase is recorded as Nvidia revenue. Nvidia's cash pile and stock price grow. It invests again.
Kress used the phrase "circular financing" herself, before any analyst did. She said Nvidia sees it differently.
The deals
Nvidia partnered with six Wall Street firms — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — to raise the $500B. It's backing OpenAI with up to $105 billion in financing for an Ohio data center. For another unnamed lab, Nvidia will guarantee nearly two gigawatts of compute.
The $500B is not signed yet — Nvidia's filing says it's "subject to definitive agreements." It's an intention, not money in place.
Why Nvidia says it's not a problem
Kress gave three reasons: outside lenders still assess each deal independently, Nvidia isn't making direct loans, and if a customer fails, the equipment can be moved to another buyer.
The risk
If one lab can't pay, Nvidia loses the sale and the investment at the same time. The chips may be movable, but the timing isn't. And 25% of next year's revenue now depends on customers whose business model is: borrow to buy chips, then hope the AI pays for itself.