Nvidia is trying to turn the most expensive buildout in tech history into an asset class. The catch: to bring Wall Street in, the chipmaker is offering to absorb part of the pain if the value of its GPUs falls.

On Monday, Nvidia said Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR were prepared to create more than $500 billion in capital pools for AI infrastructure. The arrangement would help fund data centers for Nvidia customers, with the company potentially providing residual-value support of up to 25% on individual projects.¹

The pitch is bigger than another financing round. Nvidia wants lenders to view its chips as durable collateral—and, ultimately, to create a functioning resale market for aging AI hardware. Under the proposed backstop, if a borrower defaults and GPUs fetch less than their assumed value, Nvidia would cover part of the gap. The company argues its compute can be redeployed: “When needs change, the factory can be used by another customer, another cloud or another operator.”²

That is the bullish reading: institutional money could lower borrowing costs, help smaller neoclouds compete with hyperscalers, and make Nvidia’s huge cash pile a moat that rivals cannot easily replicate. One analyst described the arrangement as risk-sharing rather than Nvidia carrying the whole burden: “They’re sharing the reward, but they’re also sharing the risk.”³

But the announcement also sharpened the bear case. Nvidia is backing projects that buy Nvidia hardware; if AI demand softens, chip values, customer spending and Nvidia’s guarantee obligations could all deteriorate together. That “wrong way” risk recalls the unease around vendor financing during the dotcom era, even if outside investors—not Nvidia alone—would provide most of the capital.

On Tuesday, Jensen Huang publicly pushed back on the circular-financing critique, saying the initiative was designed to bring “independent, long-term institutional capital” into AI infrastructure. Still, skeptics say securitizing GPU-backed debt spreads exposure to insurers, pensions and other investors just as the industry is assuming compute demand will remain relentless.⁴

For Nvidia, the plan is both defensive and expansive: keep AI construction moving, keep customers attached to its ecosystem, and persuade Wall Street that yesterday’s GPUs will retain value tomorrow. Whether that becomes a new infrastructure market—or the next overbuild—will depend on demand holding up long after the financing closes.