Nvidia just pulled off one of the more audacious financial moves in recent tech history. The chipmaker announced that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR are willing to commit up to $500 billion toward AI data center builds. But the real story isn’t the headline number — it’s what Nvidia had to promise to get those firms to the table.
The company is now guaranteeing, with its own money, that its GPUs used as loan collateral will hold their value. Specifically, if a data center owner defaults and lenders must liquidate the chips, Nvidia will cover up to 25% of the shortfall between book value and actual resale price. That promise sent bond markets into a tizzy, enough that CEO Jensen Huang took to X and business television to explain why the risk is manageable.
Financiers call this “wrong way” risk: Nvidia’s obligations grow precisely when demand weakens and its revenues get squeezed. The Lucent Technologies comparison has been floating around, and Huang clearly feels the shadow. Lucent lent its own customers money to buy telecom gear and cratered with the dotcom bust. Nvidia’s counterargument is that it’s not doing that — it’s bringing in independent institutional capital and merely backstopping a fraction of residual chip value.
Here’s the part that matters beyond Wall Street: Nvidia is trying to build a functioning secondary market for aging GPUs. Huang frames AI servers as “AI factories” — long-lived infrastructure like railroads or airlines, not disposable PCs. “When needs change, the factory can be used by another customer, another cloud or another operator,” he wrote. That vision, if it works, could reshape how startups and enterprises buy compute. Instead of chasing the newest silicon, they might tap into a tiered hardware ecosystem where last-gen chips remain viable. Huang has the market power to create that reality. Whether demand holds long enough to prove him right is the billion-dollar question.
