Nvidia just pulled six of the biggest names in finance into the AI infrastructure race. According to TechCrunch AI, Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR have signaled they’re willing to commit up to $500 billion to build AI data centers. That number grabbed the headlines this week. But the real story sits underneath it, and it changes how AI hardware might get bought, sold, and reused for years.
What Actually Happened
To get those financial giants on board, Nvidia agreed to do something unusual: guarantee the value of its own chips when they’re used as collateral in these deals. TechCrunch AI reports that if GPUs pledged against a loan don’t hold their expected value, Nvidia will cover up to 25% of the shortfall.
Here’s the mechanic in plain terms. A data center owner borrows money, using Nvidia chips as collateral. If that owner defaults and the lender has to sell the chips, but the chips can’t fetch the price on the books, Nvidia writes a check for part of the gap.
The move rattled bond markets enough that CEO Jensen Huang went on X and business TV to explain that Nvidia’s exposure is capped. “This initiative is designed to address that concern,” Huang wrote, framing it as a way to bring “independent, long-term institutional capital into the AI infrastructure market.”
Why This Matters
What stands out here is the quieter goal. Huang wants a working secondary market for aging GPUs. If used AI hardware holds its value, older chips stay in demand, and Nvidia keeps selling into an ecosystem instead of a one-time upgrade cycle.
Huang is selling a specific vision to make that happen. He calls his AI servers “AI factories” and wants investors to treat them like railroads or airlines, long-lived infrastructure, not fast-depreciating gear like a laptop. “When needs change, the factory can be used by another customer, another cloud or another operator,” he said, arguing that broad demand protects residual value.
This is significant for startups and enterprises. A healthy used-hardware market means more teams could tap a wider range of chips, each tuned to different needs, the same way many are now mixing affordable open-weight models with frontier ones.
The Risk Nvidia Is Taking
The plan carries a real danger that financiers call “wrong way” risk. Nvidia’s obligations grow exactly when demand weakens, which is the same moment its own revenue would get squeezed. Both pains would land at once.
Critics have reached for the Lucent comparison, and it’s not a cheap shot. Lucent lent customers money to buy its telecom gear, then crashed with the dotcom bubble. Nvidia has already committed billions toward its own buyers, including OpenAI, Anthropic, CoreWeave, Nebius, Firmus, and Lambda. Bloomberg calculated roughly $750 billion in circular deals tied to Nvidia this summer, per TechCrunch AI.
The difference this time: Nvidia is getting outside institutions to carry the bulk of the capital and risk, while it only backstops a slice of future chip value. That’s a smarter structure than Lucent’s, even if the shadow lingers.
The Bigger Picture
Traditional funding for AI buildouts is wearing thin. TechCrunch AI notes that hyperscalers have leaned hard on the usual levers already:
- Oracle has taken on heavy debt
- Google has issued new equity
- Meta has burned large amounts of cash
The mood is cautious enough that Microsoft CEO Satya Nadella recently recommended the book 1873, about railroad-era financial engineering that helped crash the economy. The unspoken worry: what if AI demand doesn’t keep outrunning capacity? If usage cools, or a new technology makes today’s infrastructure obsolete, the whole structure gets tested.
What To Watch Next
Nvidia is betting the AI boom has years left, and it’s using its position as the dominant chipmaker to lock in demand while the window is open. Watch whether this $500 billion commitment turns into actual signed deals, how bond markets settle after Huang’s reassurances, and whether a genuine resale market for older GPUs takes shape.
If it works, Nvidia reshapes how AI compute gets financed and keeps its chips valuable long past launch day. If demand falters, the same guarantees become a liability at the worst possible time. Full details are in the original TechCrunch AI report.