Jensen Huang Explained $500 Billion of Wall Street Money in Five Words

Five words: "In AI, compute is revenue."

That's how Nvidia (NASDAQ: NVDA) CEO Jensen Huang summed up Monday's announcement. Goldman Sachs, BlackRock, KKR, and three other financial giants signed memorandums of understanding to build financing platforms meant to mobilize more than $500 billion of third-party money for artificial intelligence (AI) infrastructure.

The sentence is the whole credit case. After all, lenders finance buildings, power plants, and toll roads because those assets produce income a loan can be repaid from. Huang's argument is that a rack of chips now belongs on that list.

The release completes the thought, calling Nvidia compute "broadly adopted, flexible across models and workloads, fungible and transferable across customers and operators, and continuously improved through CUDA software."

In plain terms, if a borrower fails, the hardware can be pulled and put to work for someone else, and software updates keep old chips earning.

The platforms are aimed at the hyperscalers, AI labs, and enterprises building data centers -- financing that lets them buy compute without tapping their own balance sheets. Demand isn't the problem the platforms solve. Nvidia's revenue grew about 71% over the trailing 12 months. The bill for the next wave is what's outgrowing the buyers.

Image source: Nvidia Corporation.

The repossession test

So the test worth applying is the lender's. What would you actually hold if a default happened?

A warehouse of processors, whose value depends on two things -- demand for computing at that moment, and how quickly newer chips make older ones cheap. If a borrower failed because AI demand cooled, the collateral would be cooling with it.

Scale is part of the test, too. More than $500 billion is roughly two years of Nvidia's total revenue (the trailing-12-month figure is about $253 billion).

And nowhere does the announcement assign the credit loss. The platforms are described as independent, their capital as third-party. The document leaves it there.

The CUDA line is arguably the strongest part of the claim. Code written for one generation of Nvidia hardware runs on the next, so an older chip keeps its buyers -- which makes a rack of its processors better collateral than most machines its age.

Other machines already get financed this way. Aircraft and rail cars secure their own debt because their value outlives their first owner. Chips have never been a standard entry on that list, because chips have historically been treated as fast-depreciating hardware.

As long as AI demand keeps growing, Huang's five words hold and the collateral goes untested. Collateral only matters when something goes wrong. So far, nothing has.

Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has positions in and recommends BlackRock, Goldman Sachs Group, KKR, and Nvidia. The Motley Fool has a disclosure policy.