Nvidia Asks Six Richest Firms on Earth for Half a Trillion Dollars; None Say No

On August 10, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish financing platforms intended to mobilize more than $500 billion of third-party capital for AI infrastructure. Jensen Huang later noted that he approached only those six firms, and that none of them turned him down.

Take a moment with that. The most selective fundraising process in financial history consisted of six phone calls and six yeses. Most of us cannot get six friends to agree on a restaurant. Huang asked six of the largest pools of capital ever assembled to help move half a trillion dollars toward machines that his company happens to manufacture, and the hardest part of the negotiation appears to have been scheduling.

The mechanics, for those keeping score at home: special-purpose entities will be able to issue tens of billions of dollars in debt at a time, using GPUs — the computers — as collateral, then lease that compute to Nvidia’s customers. Insurance capital, retirement capital, and institutional capital thereby flow into data centers without any of it touching Nvidia’s balance sheet, which remains as pristine as a showroom H200.

Huang describes compute as “an investable infrastructure asset,” language that formally promotes the graphics card from “component” to “asset class,” a career arc previously achieved by houses, railroads, and briefly, tulips.

Apollo president Jim Zelter called modern compute “a scarce, mission-critical asset class with compelling investment characteristics,” which is what it sounds like when $1.05 trillion of assets under management falls in love.

The load-bearing question — the one holding up the entire half-trillion-dollar edifice — is whether a GPU ages like a railroad or like a smartphone. Railroads run for a century. Smartphones get traded in when the camera bump changes shape.

The bulls have an answer: Bank of America analysts argue that Nvidia compute is “fungible and transferable across operators,” with the CUDA software layer “continuously extending useful life,” keeping resale values high and depreciation curves “benign.” Benign is a wonderful word to deploy about the depreciation of an object whose manufacturer releases a dramatically better version of it every twelve to eighteen months, on a schedule, with a keynote.

The skeptics also have an answer, and awkwardly, some of them are in the deal. Goldman Sachs — a participant — flagged “economic obsolescence from newer-generation products” as a challenge for long-term lending. That newer-generation product pipeline is, of course, Nvidia’s own roadmap. The collateral’s greatest depreciation risk is the borrower’s partner’s keynote schedule. Analysts also note the arrangement makes future GPU demand “more sensitive to credit conditions,” which is a polite way of saying the AI buildout now has a mortgage.

Per TechCrunch’s reporting, Nvidia has even agreed to backstop up to 25% of the collateral’s value if borrowers default and lenders have to liquidate — the chipmaker equivalent of cosigning the loan on a car you sold.

It’s worth noting what this all is, legally speaking: memorandums of understanding. Non-binding. Contingent on definitive agreements. The $500 billion is a mobilization target “over time,” not a wire transfer. The market, unbothered by such distinctions, nudged Nvidia’s stock up about 1%, because at Nvidia’s size, announcing history’s largest infrastructure financing alliance moves the needle roughly as much as a good earnings whisper.

Still, the vision has a certain grandeur. Huang talks about creating “an ecosystem of used AI hardware” — a CarMax for compute, where last generation’s frontier-model trainer becomes next year’s inference workhorse becomes, eventually, someone’s very overqualified Minecraft server. When needs change, he says, “the factory can be used by another customer, another cloud or another operator.”

The railroads got a century out of their rails. The GPUs are being asked for four to six years, tops, and half a trillion dollars is being arranged on the bet that this is long enough. Six firms heard that pitch. None of them said no. Either it’s the surest thing in modern finance, or nobody wanted to be the one who passed — and the beautiful thing about a memorandum of understanding is that you never have to find out which.

Sources: