Nvidia this week revealed that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR are willing to commit up to $500 billion to build AI data centers. The eye-popping figure drew most of the attention, but the more consequential development is Nvidia’s plan to create a secondary market for aging GPUs by guaranteeing a portion of their future value.
To win over those institutional investors, Nvidia has agreed to use its own money to back the residual value of chips used as collateral in these financing deals. The arrangement has drawn comparisons to Lucent Technologies, the telecom equipment maker that collapsed after lending customers money to buy its gear during the dotcom bust. Nvidia CEO Jensen Huang has pushed back on that framing, taking to X and business television to explain how the company’s risk is limited.
Also read: Microsoft merges Copilot apps, cuts Group Chats, Deep Research, and Mico character
Under the terms, if GPUs used as collateral fail to retain their book value during a liquidation event, Nvidia will cover up to 25% of the shortfall. That means if a data center owner defaults on a loan and the lender sells the hardware for less than expected, Nvidia absorbs part of the loss. The structure is designed to make aging chips more attractive as collateral and, in turn, keep Nvidia’s revenue flowing even as its hardware cycles through multiple owners.
How the financing structure works
The plan is deliberately different from Lucent’s model. Nvidia is not lending its own capital at scale; it is bringing in independent, long-term institutional money and simply backstopping a slice of the chips’ future value. Huang described the initiative on X as a way to address concerns about circular financing, writing: “We are bringing independent, long-term institutional capital into the AI infrastructure market.”
Also read: Apple in talks to pay publishers for news content to power Siri AI: report
CoreWeave, the neocloud provider, originated the practice of using Nvidia chips as loan collateral, and the model has since expanded to other firms like Nebius, Firmus, and Lambda. Nvidia has also committed billions toward frontier AI labs OpenAI and Anthropic. Bloomberg has calculated that the company has been working on another $750 billion worth of circular deals this summer alone.
The financial engineering comes as traditional funding sources for AI infrastructure have begun to strain. Some hyperscalers have loaded up on debt—Oracle being a notable example—while Google has issued new equity tranches and Meta has burned through substantial cash. The anxiety has reached such a pitch that Microsoft CEO Satya Nadella recommended the book “1873” during his latest earnings call, a history of railroad-era financial speculation that ended in economic collapse.
Why Nvidia is betting on a used chip market
Beyond the immediate goal of financing data center construction, Huang is pushing a longer-term vision: treating AI servers as durable, “investable infrastructure” rather than quickly depreciating assets. He has likened his “AI factories” to railroads or airlines, arguing that when one customer’s needs change, the hardware can be redeployed by another cloud operator or enterprise.
“This broad ecosystem gives NVIDIA compute a deep market of potential users and offtakers, helping protect residual value,” Huang said. That framing requires a functioning secondary market for older GPUs, which is precisely what the value guarantee is meant to establish. If successful, it would sustain demand for Nvidia hardware well beyond its first deployment cycle.
For startups and enterprises, the implication is significant. A liquid market for used AI hardware could lower the barrier to entry for companies that cannot afford the latest flagship chips, much as the current shift toward affordable open-weight models has broadened the AI software space. Researchers and smaller firms would gain access to a wider variety of hardware tuned to different computational needs.
The central risk is that the AI boom’s demand-supply imbalance does not persist. If enterprises and consumers temper AI usage, or if new technologies render current infrastructure obsolete, the value of those GPUs could plummet—and Nvidia’s guarantee would come due at exactly the wrong time. Financiers call this “wrong way” risk: the company’s obligations grow as its core business weakens.
Huang’s counterargument is that AI is in its early innings and that the infrastructure being built today will remain productive for years across multiple tenants. Whether that bet pays off will determine not just Nvidia’s balance sheet, but the shape of the AI hardware market for the rest of the decade.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. The AI infrastructure and semiconductor markets are volatile, and forward-looking statements about Nvidia’s financing plans involve significant uncertainty. Readers should conduct their own research before making investment decisions.
Be the first to comment