Nvidia is trying to turn its GPUs from fast-depreciating technology into financeable infrastructure. The wager could unlock a vast new pool of money for AI data centers—or leave the chipmaker exposed precisely when demand cools.

On August 10, Nvidia said it was working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on financing platforms intended to mobilize more than $500 billion for AI infrastructure. The headline number, however, is not cash already committed: the arrangements remain subject to final agreements, with the funding mix, guarantees and first-loss terms still unknown.¹

The core mechanism is Nvidia’s offer to provide residual-value support of up to 25% on a project-by-project basis. If collateralized GPUs lose more value than expected after a borrower defaults, Nvidia could absorb part of the shortfall. That could lower borrowing costs and make data-center projects viable for neoclouds and other customers that lack Big Tech-sized balance sheets.²

Supporters frame that as a shrewd extension of Nvidia’s dominance. Its cash-rich balance sheet can help customers build facilities, secure future chip orders and bind an ecosystem of cloud providers and startups more tightly to its hardware. As one analyst put it, Nvidia is “sharing the reward, but they’re also sharing the risk.”³

But the risk is unusually pointed. Nvidia’s obligation rises when GPU demand and resale values fall—the same conditions likely to pressure its chip sales. Critics see echoes of vendor financing and broader AI “circularity”: suppliers helping fund the purchase of their own products. The financing may be backed by real collateral and spread across institutional investors, but it still assumes compute demand remains durable and that newer, cheaper or more efficient technology does not suddenly erode the value of today’s hardware.⁴

Nvidia chief executive Jensen Huang has argued that the structure is designed to bring in “independent, long-term institutional capital,” rather than simply recycle Nvidia money through customers.⁵ The distinction matters. If Wall Street treats AI factories like railroads, aircraft or power plants, Nvidia gains a new engine for expansion. If the buildout outruns usage, its backstop could become the bill for an overheated boom.