Nvidia’s newest AI infrastructure strategy is not only about raising capital for more data centers. It is also about shaping what happens to its chips after their first wave of use, and whether older GPUs can remain valuable enough to support a larger market.
A $500 billion financing push with a second goal
Nvidia announced this week that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR were willing to commit up to $500 billion to build AI data centers. The size of that number explains why it attracted attention, but the more important mechanism sits underneath the headline.
The company is trying to make aging AI hardware easier to finance, own, resell, and reuse. In practical terms, Nvidia wants its GPUs to behave less like equipment that quickly loses relevance and more like assets that can support long-term infrastructure investment.
That matters because data center owners can use Nvidia chips as collateral. If lenders believe those GPUs will still be worth enough later, financing becomes easier. If lenders worry the chips could lose too much value, the entire structure becomes harder to support.
How Nvidia’s guarantee works
To bring large financial firms into the plan, Nvidia has agreed to back part of the future value of its chips with its own money. The promise applies when GPUs used as collateral do not hold their expected value.
Specifically, Nvidia is promising to cover up to 25% of the difference if the collateral falls short. If a data center owner defaults on a loan, and the lender has to sell the chips but cannot get the value assumed on the books, Nvidia would contribute toward that gap.
This is why the plan is both clever and risky. It gives lenders more confidence that Nvidia hardware can support AI data center loans. At the same time, it creates a direct link between Nvidia’s balance sheet and the future resale market for its own chips.
The company is not simply selling GPUs and walking away. It is trying to make the financing market believe those GPUs will remain useful, tradable, and wanted after their first deployment.
The risk inside the structure
The danger is what financiers call “wrong way” risk. Nvidia’s potential obligations would rise in the same scenario where demand for its hardware weakens. If GPUs lose value because AI infrastructure demand cools, Nvidia could face pressure at the same time its revenue is likely being squeezed.
That concern has already affected how the plan is being discussed. The bond markets got so worried that Nvidia CEO Jensen Huang used X and business TV to explain why he believes the company’s exposure is limited.
Critics have also compared the structure to Lucent Technologies, the telecommunications equipment provider that rose and crashed with the dotcom bubble after lending customers money to buy its products. The comparison is uncomfortable for Nvidia because the company has already committed billions toward buyers of its chips.
Those buyers include frontier AI labs OpenAI and Anthropic, neoclouds such as CoreWeave, which originated the use of Nvidia chips as collateral, and Nebius, Firmus, and Lambda. Bloomberg has calculated that Nvidia has also been working on another $750 billion worth of circular deals this summer.
Huang has tried to draw a clear distinction between that concern and the new structure. On X, he wrote, “Is this circular financing?” He then described the initiative this way: “This initiative is designed to address that concern. We are bringing independent, long-term institutional capital into the AI infrastructure market.”
Why the used GPU market matters
The deeper strategic point is that Nvidia is trying to build confidence in a secondary market for used AI hardware. If aging GPUs can move from one customer, cloud, or operator to another, they become easier to finance today.
Huang’s argument depends on AI infrastructure being treated as long-term “investable infrastructure.” He describes AI servers as “AI factories,” a framing that puts them closer to railroads or airlines than to PCs that quickly depreciate.
He also argues that Nvidia compute can have many future buyers. As he put it, “When needs change, the factory can be used by another customer, another cloud or another operator. This broad ecosystem gives NVIDIA compute a deep market of potential users and offtakers, helping protect residual value,”
If that vision holds, Nvidia benefits in several ways. More financial capital can support AI data center construction. Lenders may become more comfortable with GPU-backed collateral. Buyers may view Nvidia hardware as part of a broader ecosystem rather than a single-use purchase.
The potential beneficiaries are not limited to the largest AI builders. Startups, enterprises, and researchers could eventually gain access to a wider range of AI hardware. Some may choose older or more affordable Nvidia systems for specific needs, just as some users are beginning to choose affordable open-weight models alongside frontier models.
The bet behind the plan
The plan depends on continued demand for AI infrastructure. Today, the source article describes demand as far greater than capacity. The risk is that this imbalance may not last.
Enterprises and consumers could reduce their AI usage. New technologies could make existing infrastructure more efficient. Current AI infrastructure could also become obsolete faster than investors expect.
That is why the financing structure is so important. Nvidia is not only trying to fund more data centers after traditional approaches have become strained. Some hyperscalers have already taken on debt, issued new equity, or burned significant cash. Microsoft CEO Satya Nadella recently recommended “1873,” a book about railroad-era financial engineering that crashed the nation’s economy, during his latest earnings call.
Nvidia is using its market position to argue for a different future: one where AI compute remains useful across owners and over time. If it succeeds, aging GPUs could become a central part of the AI economy rather than a weakness in the hardware cycle.
If it fails, the same guarantee that makes lenders comfortable could become a source of pressure. That is what makes the plan unusual, smart, and dangerous at once.