Nvidia is asking Wall Street to finance the next leg of the AI buildout—and offering just enough protection on its chips to make the wager more comfortable. The move could widen access to computing power, or bind the industry’s fortunes even more tightly to unproven demand.

This week, Nvidia said Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR could commit up to $500 billion to AI data-center projects. The crucial mechanism is not merely the headline sum: on selected deals, Nvidia may support up to 25% of a GPU’s residual value if the collateral disappoints in a liquidation.¹

That is intended to draw long-term institutional money into projects as hyperscalers and smaller “neocloud” operators lean harder on debt. Nvidia chief Jensen Huang has argued the arrangement is an answer to circular-financing fears because outside institutions, rather than Nvidia, would supply the bulk of the capital.²

Supporters see a strategic extension of Nvidia’s dominance. Its cash pile and ability to backstop customers give it an advantage that younger chip rivals cannot easily replicate, while helping neoclouds fund capacity and stay tied to Nvidia GPUs. As one analyst put it, the company is “sharing the reward, but they’re also sharing the risk.”³ The broader bullish case is that older chips retain a useful second life: if one customer no longer needs an AI “factory,” another cloud or operator can take the machines.

But the structure also creates what financiers call wrong-way risk. If demand weakens or more efficient technology makes today’s hardware less valuable, Nvidia could face collateral payouts precisely as chip sales come under pressure. The plan therefore revives comparisons to vendor-backed booms of the past, even if Nvidia’s exposure is capped and dispersed.

Skeptics say securitizing GPU-backed debt can turn a rational infrastructure bet into a system-wide vulnerability. “The bull case is there’s infinite demand for AI, and this is going to go on forever,” Seaport Global Securities analyst Jay Goldberg said. “Somebody at some point is going to say, ‘Whoa. Wait a minute, what are we doing?’”⁴

For Nvidia, the gamble is clear: turn its chips into financeable infrastructure before the AI boom loses its momentum.