Morning Edition · Thursday, August 13, 2026Published at 2:26 AM EDT · New York
The chipmaker is guaranteeing part of the resale value of its own accelerators, moving compute financing off its own balance sheet and into structured credit markets.

Nvidia said it has signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create independent financing platforms intended to mobilize over $500 billion of third-party capital for data centers, power and the rest of the AI buildout. The company describes the arrangement as turning AI factory compute into an investable asset class, with dedicated capital pools lending across its network of AI labs, enterprises and cloud providers.
Two structural details matter more than the headline number. First, the financing vehicles sit outside Nvidia's balance sheet, so the debt that funds GPU purchases is held by credit investors rather than by Nvidia or, in many cases, by the buyer directly. Second, Nvidia is guaranteeing a portion of the resale value of its own chips, which is what makes an accelerator usable as loan collateral at a rate a lender will accept. A GPU with a guaranteed floor price fits a depreciation schedule a credit committee can underwrite. A GPU without one does not.
That guarantee is also where Nvidia's own risk sits. Nvidia sells the equipment, sets the product roadmap that makes older chips obsolete, and now backs part of what that equipment will be worth after use. If demand for Hopper and Blackwell generation chips holds up and utilization stays high, the resale assumptions prove conservative and the deals hold together. If a newer chip or a cheaper way to run inference shrinks the resale market for older accelerators, the guarantee turns what would have been a customer's loss into Nvidia's loss.
The agreements are memorandums of understanding, not completed funds, and the $500 billion is a target for capital raised over time, not money committed today. Bloomberg reported the plan before Nvidia confirmed it, and Chief Executive Jensen Huang said in a CNBC interview that he approached only these six firms and none turned him down. Debt is expected to be the main financing tool. The announcement does not specify pricing, loan terms, or how much of the resale value Nvidia is actually backing.
Start a discussion in Townsquare.
More from this edition
Nvidia, which converts its customers' capital constraint into bookable demand while keeping the debt off its balance sheet, and the six private-credit and asset managers, who gain a large new fee-bearing asset class with a vendor-backed floor under the collateral, as BlackRock's Larry Fink compared the structure to the creation of mortgage-backed securities.
The core announcement is confirmed by Nvidia, Blackstone and Apollo, but the load-bearing detail sits outside those releases: secondary reporting puts the guarantee at up to 25 percent of any residual-value shortfall, a cap the article omits, and Jensen Huang disputes the circular-financing characterization by arguing Nvidia's share of the risk is smaller than what lenders absorb, a claim no disclosed term sheet yet tests.
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
The main limit on the AI buildout is shifting from chip supply to the cost and structure of capital, and Nvidia is responding by backing its own collateral. Private credit managers gain a large new source of loans with a vendor-backed floor under them. Neoclouds and second-tier AI labs gain access to debt they could not raise on their own credit alone. The risk runs through resale prices for used chips: if the secondhand market for older accelerators weakens, the losses land on Nvidia's guarantee and on the credit funds holding the loans, not on the operators who bought the hardware.
What to watch
Observations to monitor, not financial advice.
Synthesized from: NVIDIA Blog · NVIDIA Newsroom · Blackstone
Comments
1Aug 14, 12:58 AM · edited
Nvidia's residual value guarantees on accelerators impose a direct cost on its own product cadence: each successor GPU that reprices current inventory triggers claims against those guarantees.