Nvidia hands the AI build-out to Wall Street with a $500bn financing push
Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR will raise third-party capital for data centres, with Nvidia backstopping part of the value of the chips pledged as collateral.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- Six asset managers will mobilise more than $500bn of third-party capital through independent compute financing platforms, announced on 10 August.
- Nvidia will cover up to 25% of the shortfall if GPUs pledged as collateral fail to hold their expected value, according to TechCrunch's reporting.
- The structure keeps the borrowing off Nvidia's balance sheet while giving customers a route to finance hardware they could not buy outright.
- Jensen Huang says the aim is to bring independent, long-term institutional capital into AI infrastructure, after bond markets balked at circular financing.
Nvidia said on 10 August that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR have agreed to establish independent compute financing platforms intended to mobilise more than $500bn of third-party capital for AI infrastructure. The company framed the platforms as a way to fund the build-out of data centres without that funding sitting on its own books.
The mechanism that makes the arrangement work is a residual value guarantee. TechCrunch reported on 13 August that Nvidia has agreed to cover up to 25% of the difference if the GPUs pledged as collateral fail to hold their expected price. That commitment converts a fast-depreciating asset into something a lender can underwrite, which is the precondition for institutional money entering at this scale.
Jensen Huang, Nvidia's chief executive, has spent the days since defending the structure in public. Bond markets had grown uneasy about the circularity of a chip supplier financing its own customers, and Huang argued that the platforms do the opposite, by bringing independent, long-term institutional capital into the market. He also pointed to resale: when needs change, he said, a factory can be used by another customer, another cloud or another operator.
That second argument is the more consequential one. A guarantee is only as good as the secondary market standing behind it, and no deep market for used AI accelerators currently exists. Nvidia is effectively underwriting one into being, and the residual value it is promising to defend is the residual value of its own product.
The obvious objection is that the risk has not disappeared, only moved. Nvidia's exposure is capped at a quarter of any shortfall rather than eliminated, and the guarantee falls due precisely when demand softens — the moment the company would least want to be writing cheques.
For two years the constraint on AI infrastructure has been described as chips. This says the constraint is now capital, and that the build-out has run past what hyperscaler cash flow can cover. Bringing in pension and insurance money changes who absorbs the loss if demand disappoints, and moves a share of AI's downside onto balance sheets with no particular insight into how quickly a GPU stops being useful.