Wall Street's $500B Bet on Nvidia GPUs

The most important detail in Nvidia's eye-popping $500 billion data center financing plan isn't the headline number. It's the 25%.

That's the percentage of GPU resale value Nvidia has agreed to guarantee with its own money when its chips are used as collateral in the deals. If a data center owner defaults on a loan and the lender has to liquidate the GPUs, but the chips won't sell for what the books say they should, Nvidia will cover up to a quarter of the difference.

It is, in essence, Nvidia building a financial derivatives market on its own silicon.

Wall Street's new infrastructure asset class

The financing platform announced last week brings together six of the biggest names in institutional finance: Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. Together they've signaled willingness to commit up to $500 billion to AI data center construction, with Nvidia's residual-value guarantee as the catalyst.

Jensen Huang, Nvidia's CEO, is framing the deal not as a chip sale but as infrastructure financing. "When needs change, the factory can be used by another customer, another cloud or another operator," he said in a statement. "This broad ecosystem gives NVIDIA compute a deep market of potential users and offtakers, helping protect residual value."

Huang wants aging GPUs treated like railroad tracks or airline fleets, durable assets with a long tail of utility, rather than quickly depreciating technology that loses value the moment a better chip ships. If that framing sticks, it transforms how Wall Street values not just Nvidia but the entire AI hardware supply chain.

The Lucent-shaped shadow

The comparison that keeps coming up is Lucent Technologies. The telecom equipment giant lent its customers money to buy its own gear during the dotcom boom and crashed spectacularly when the bubble burst. Nvidia has already committed billions directly to customers, including frontier AI labs OpenAI and Anthropic, neoclouds like CoreWeave, and infrastructure firms Nebius, Firmus, and Lambda.

Huang is aware of the comparison. Asked about "circular financing," he posted on X: "This initiative is designed to address that concern. We are bringing independent, long-term institutional capital into the AI infrastructure market."

The key difference: Lucent carried the risk itself. Nvidia is getting others, the BlackRocks and Goldman Sachses of the world, to shoulder the bulk of the capital, merely agreeing to protect a fraction of the chips' future value. But the "wrong way" risk remains: Nvidia's obligations grow exactly when its revenues are most likely to be squeezed.

The Burry factor

Michael Burry, the investor who famously predicted the 2008 housing collapse (and was portrayed by Christian Bale in The Big Short), called the plan a "Wall Street stunt" on social media. "Meet the new boss," he wrote, "same as the old boss."

Burry's skepticism taps into a broader anxiety. Microsoft CEO Satya Nadella recently recommended a book about railroad-era financial engineering during his latest earnings call, the one whose speculative bubble and crash helped trigger a national economic crisis. The parallel is not lost on anyone watching the AI infrastructure buildout.

The China variable

There's another risk the financing plan doesn't address. Chinese AI chip makers are producing competitive alternatives at significantly lower prices. If data center operators can buy cheaper Chinese hardware instead of paying a premium for Nvidia's aging GPUs, the secondary market Nvidia is trying to build loses its foundation before it's fully constructed.

For now, Nvidia's pricing power remains formidable. But the $500 billion plan is effectively a hedge against the day it isn't.

The gamble that could define the AI era

Huang's bet is simple but audacious: AI compute demand will grow for years, not months, and aging hardware will find new homes rather than becoming e-waste. In that world, a guarantee is cheap, the chips keep their value naturally, and Nvidia never has to pay out.

If he's wrong. If AI adoption plateaus, or new architectures render today's GPUs obsolete faster than expected, or Chinese competitors undercut the market, the guarantee becomes a liability that compounds at exactly the wrong time.

But Huang is selling a vision of AI factories as permanent infrastructure, the way railroads and power grids are permanent. And when you're the company that controls the most critical component of that infrastructure, you have both the incentive and the window of opportunity to make that vision stick.

The $500 billion plan is the mechanism. Whether it's genius or folly depends entirely on whether AI demand is a bubble or a new industrial revolution. The answer is probably somewhere in between. And that's exactly what makes this the most interesting financial engineering story in tech right now.