The rapid expansion of AI has created a new financing market around the expensive infrastructure needed to run it. AI companies and cloud providers have increasingly borrowed money to buy Nvi
The rapid expansion of AI has created a new financing market around the expensive infrastructure needed to run it. AI companies and cloud providers have increasingly borrowed money to buy Nvidia GPUs and build data centres, sometimes using the chips as collateral for those loans.
Nvidia has also become involved in helping customers secure financing, working with major financial firms to develop structures that can fund the growing demand for AI computing. That effort reached a much larger scale when Nvidia announced partnerships with firms including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to help mobilize more than $500 billion in third-party capital for AI infrastructure.
Source: Reuters
The plan is now facing a basic question from Wall Street: how much are those AI chips really worth as collateral? Banks and asset managers want stronger guarantees because they are unsure how long Nvidia’s GPUs will generate enough revenue to support the loans, and how much the hardware would be worth if a borrower defaulted.
Reuters reported that lenders generally depreciate the chips over about three to four years, while Nvidia argues that some AI systems can generate revenue for much longer. That difference is creating tension as Nvidia tries to bring more capital into AI infrastructure financing.
Wall Street has seen this exact structure before, a dominant hardware supplier financing the purchases of its own equipment, and it has a specific, well documented failure mode. In the telecom boom of the late 1990s, Lucent Technologies built a vendor-financing portfolio that reached $ 8.1 billion by lending money directly to the same customers buying its networking equipment, with many of those loans going to startups.
When the telecom bubble popped, one of those borrowers, Winstar Communications, defaulted on its obligations after Lucent had extended it 1 billion dollars under a 1998 financing agreement, then filed for bankruptcy and sued Lucent for 10 billion dollars, claiming Lucent’s own failure to keep funding the deal is what pushed Winstar under.
Lucent’s stock collapsed from a high of 82 dollars a share in December 1999 to about 9 dollars by mid 2001, and the company posted 16.2 billion dollars in net losses across four quarters, with executives directly blaming vendor financing for a majority of the damage.
CreditSights analyst Glenn Reynolds said what made Lucent’s position alarming wasn’t the loans already drawn, it was the undrawn commitments still sitting on the books, because the company “did not have the balance sheet capacity to fulfill all the draws” if customers called them in at once.
Nvidia is spreading the lending across six outside financial firms rather than funding everything itself. Also, CEO Jensen Huang capped Nvidia’s own backstop at 125 billion dollars, a quarter of the 500 billion dollar target, rather than guaranteeing the whole amount the way Lucent effectively tried to.
Wall Street’s skepticism goes deep because nobody has settled how fast a GPU loses its worth, including the hyperscalers whose own accounting depends on the answer.
Amazon shortened the useful life it assigned to its servers in 2025, while Meta extended its own useful life assumption in that same quarter, two of the industry’s biggest buyers moving in opposite directions on the identical hardware category at the identical time.
Microsoft’s CFO has put the company’s policy at a two- to six-year range for computing equipment, acknowledging openly that the number is closer to an estimate than a measurement. Also, CoreWeave had H100 GPUs return from expired 2022 contracts and get immediately released at 95% of their original pricing, which argues for durability.
However, Nvidia’s CEO said at the company’s March AI conference, joking that “when Blackwell ramps, nobody will want Hopper,” openly predicting his newest chip generation would gut the resale value of the one investors are currently being asked to lend against. This statement makes it harder for Nvidia to insist, in the same financing pitch, that lenders should treat those same chips as collateral stable enough to count on for years.
Meanwhile, NVIDIA has launched the Open Agent Safety Platform with more than 100 industry partners, bringing together OpenShell and Sentry to provide security controls for AI agents from development through deployment.
Enjoyed this? BookmarkDeFi Planet, explore related topics, and follow us onTwitter,LinkedIn,Facebook,Instagram,Threads, and CoinMarketCap Community for seamless access to high-quality industry insights
Take control of your crypto portfolio with DEFI PLANET PRO, DeFi Planet’s suite of analytics tools.
The post Wall Street Questions Nvidia’s AI Chips as Collateral appeared first on DeFi Planet.
