Nvidia's effort to arrange up to $500 billion in financing for AI data-center builders is running into a Wall Street dispute over how long its GPUs hold their value. Bankers estimate a useful life of three to four years, against Nvidia's decade-long projection, and the gap could reshape loan terms for leveraged neoclouds such as CoreWeave and Nebius.
Nvidia's plan to funnel up to $500 billion in financing to its AI customers is colliding with a Wall Street dispute over how long its GPUs hold their value, a question that could reshape lending terms across the industry.
In August, Nvidia announced a partnership with six Wall Street asset managers to back AI infrastructure build-outs, with the GPUs themselves serving as collateral on the loans. That structure leaves neoclouds such as CoreWeave and Nebius, which rely heavily on leverage to expand data centers, dependent on how much value their hardware retains over time.
A gap in useful-life estimates
Nvidia estimates a GPU's useful life at about a decade, while bankers put it at three to four years, according to a Reuters report cited by Nvidia's rebuttal. Nvidia published a blog post arguing hyperscalers' own accounting undercuts that shorter estimate.
Hyperscalers stretch their depreciation schedules
Nvidia pointed out that Alphabet, Microsoft, Amazon, and Meta Platforms have extended the depreciation life of their hardware from three or four years to five or six years. It also cited pricing moves: CoreWeave signed new contracts at an annualized rate of $40 million per megawatt in September. Nebius, meanwhile, raised hourly compute rental prices by 17% to 21% for various Nvidia GPUs, including four-year-old chips.
Stronger guarantees are already in the pipeline
Still, the banks hold the leverage in these deals. Reuters reports tens of billions of dollars in loans now carry stronger guarantees than Nvidia's original August announcement outlined, and those guarantees could come from Nvidia's own capital reserves or from the revenue of investment-grade tech firms such as Microsoft, Alphabet, Amazon, or Meta Platforms. Nvidia is also reportedly in talks with insurance companies to hand off some of the financing risk tied to loans for CoreWeave and Nebius specifically.
Why the payback math matters
The stakes vary by borrower. Amazon says it will break even on its AI accelerator chips within three years. Alphabet's breakeven, meanwhile, is just two years. Both hyperscalers are also installing more of their own custom accelerators, which carry even shorter payback periods than Nvidia's GPUs.
If Wall Street's shorter estimate holds, it's the neoclouds leaning hardest on GPU-backed debt that stand to feel it first.
Source: The Motley Fool
Trading involves risk.
Risk Disclaimer
