Lenders Want Bigger Nvidia Guarantees on $500B GPU-Backed AI Loans, Reuters Reports

According to banking sources cited by Reuters on October 1, 2026, lenders doubt that Nvidia's chips can serve as long-term collateral.

They want Nvidia to extend its guarantees across all deals in the financing structure — or have the loans backed by revenue from investment-grade technology customers — before the structures become bankable.

At the heart of the dispute is the useful life of an AI accelerator. Nvidia chief executive Jensen Huang has said the company's top-tier GPUs can remain productive for up to a decade, generating revenue long after purchase as software extends their usefulness across workloads.

Banks, however, are taking a far more conservative view. Lenders typically underwrite Nvidia's GPUs over a three-to-four-year depreciation schedule, pricing the asset on a much shorter working life in a market where new chip generations arrive roughly every year.

That gap matters because collateral value determines how much debt lenders will extend and on what terms — and the longer a GPU is expected to generate cash flow, the more cheaply the asset can be financed.

Nvidia unveiled the financing initiative in August 2026, signing memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create compute-financing platforms intended to mobilize more than $500 billion in third-party capital over time.

Goldman Sachs chairman and chief executive David Solomon told CNBC that Huang had approached the bank with the idea. Nvidia said it may provide a residual-value support mechanism for up to 25 percent of an opportunity, assessed carefully on a project-by-project basis.

Lenders say that ceiling is too low. Three banking sources who reviewed the plan but are not part of the original financing group told Reuters that Nvidia may need to guarantee every deal — or anchor each one to the contract payments of a strong credit — to close the gap between a chip's accounting life and its claimed economic life.

There is at least one signed, disclosed example of Nvidia sharing the risk. In an August 17 filing, Nvidia revealed residual value guaranty agreements with SB Energy tied to leases for roughly 4.25 gigawatts of IT capacity at a site in Portsmouth, with the company's aggregate payment obligation capped at $105 billion.

The tenant is OpenAI: if it becomes insolvent and defaults, or simply fails to make its lease payments, Nvidia is obligated to cover the shortfall between the lease's guaranteed minimum value and whatever the site can actually be re-leased or sold for.

The standoff reflects a broader shift toward AI infrastructure funded increasingly through debt, as cloud operators and data-center developers borrow billions before new capacity starts generating revenue.

Some chip-backed structures have worked: CoreWeave closed an $8.5 billion GPU-backed loan rated A3, largely because lenders rely on Meta's contract payments, and Broadcom backstopped more than 80 percent of a $35 billion financing structure tied to Anthropic.

Nvidia shares rose about 1 percent on Thursday following the report, suggesting investors see the pushback as a negotiation rather than a breakdown.

But until lenders are convinced that a data center full of GPUs can be underwritten the way they underwrite an aircraft fleet — with decades of resale history — the scale of Nvidia's $500 billion ambition will remain subject to hard bargaining over who carries the risk.