Lenders were reported on October 1, 2026, to be questioning whether NVIDIA GPUs could retain enough value to support long-term loans tied to the company’s proposed AI infrastructure financing platforms. Some were seeking stronger protections. The concern centers on whether equipment will keep generating revenue and what it could be worth if a borrower cannot repay.
Why lenders are questioning NVIDIA’s AI financing plan
A GPU used as collateral gives a lender an asset it may be able to recover or sell if a borrower defaults. Its residual value is the expected value left at the end of a loan or other period. For long-term financing, lenders also assess whether customers will keep paying for the computing capacity that generates revenue to cover the debt.
NVIDIA has argued that top-tier GPUs can generate revenue for up to a decade. One portfolio manager said banks typically underwrite GPUs using a 3–4-year depreciation schedule. Depreciation is an accounting measure of how an asset’s cost is allocated over time; it does not, by itself, state when hardware stops operating or earning revenue.
The measures differ. NVIDIA also cited studies in which some large cloud companies extended server depreciation periods from 3–4 years to 5–6 years. That is a server-accounting measure, not a direct estimate of how long an individual GPU will generate revenue.
What NVIDIA announced with its financing partners
On August 10, 2026, NVIDIA announced memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to develop independent AI-compute financing platforms. NVIDIA said the platforms were intended to mobilize more than $500 billion in third-party capital over time for AI infrastructure. The company said the partnerships were subject to final agreements.
The proposed platforms were intended to create dedicated pools of capital for NVIDIA customers, including AI labs, enterprises and AI cloud providers. The $500 billion figure is the announced goal for capital mobilization—not a completed financing transaction.
How guarantees could fit into the proposed financing
NVIDIA has said its financing partners independently assess each opportunity, including customer commitments, expected cash flow and residual value. Those factors help lenders judge whether repayments are supported by both the computing equipment and the revenue it may generate.
Some possible deal terms were reported to include residual-value guarantees capped at 25%, while other prospective transactions could involve stronger guarantees, customer contracts or NVIDIA-backed support. These are potential structures, not terms established for every transaction. A customer contract can give lenders another source of expected payments to assess alongside the collateral.
