Liz’s Viewpoint

OpenAI and Anthropic need to borrow vast sums of money, but they have neither profits nor a track record demonstrating their creditworthiness. Enter Jensen Huang.

Nvidia’s $250 billion guarantee for OpenAI will allow the cash-burning AI lab to lease space in the largest data center ever built. OpenAI does not have an investment-grade credit rating, so Nvidia is effectively lending it its own. Broadcom did the same for Anthropic several weeks ago; at the time, I described it as being like asking your parents to co-sign the lease on your first apartment.

Nvidia’s guarantee for OpenAI is exactly that: OpenAI wants to sign a lease, and the property owner, SoftBank, is uncomfortable with the tenant risk. Huang is therefore signing as guarantor.

It is a classic credit-enhancement technique: take something risky and wrap it in something safer. It usually works well. Sometimes it ends catastrophically. Comparisons to an AI version of “Too Big to Fail” are proliferating on Fintwit.

The idea of Nvidia as the new AIG is accurate in one respect. The company most responsible for the 2007 bubble was not a bank issuing bad loans, but the insurer guaranteeing them and spreading that risk throughout the financial system. Risky mortgages went into AIG and came out bearing a AAA label. Risky AI deals are now going into Nvidia, Broadcom and Google, only to emerge similarly cleansed.

Spreading risk can sometimes be prudent; that is how mutual insurance works. But the practice can also pull into the heart of a crisis companies that might otherwise have remained outside it. AIG did not need to become involved in the mortgage crisis. Nvidia must participate in AI development, but it is putting its balance sheet to the test to finance its customers. OpenAI will install Nvidia chips in the Ohio data center.

The situation recalls General Electric and General Motors, which were nearly brought down by their respective finance divisions in 2008.