Nvidia has spent the past three years selling the picks and shovels of the artificial-intelligence boom. Now it wants to help pay for the mine.
On Monday, Aug. 10, the chipmaker announced it had signed memorandums of understanding with six of the largest names in global finance — Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs and KKR — to build dedicated financing platforms aimed at mobilizing more than $500 billion of third-party capital for AI infrastructure. The money is meant to flow into the data centers, chips, power plants, cooling systems and networking gear that underpin the buildout, and, crucially, to help Nvidia's own customers finance the deployments they cannot cover with cash on hand.
The structure is designed to keep the burden off Nvidia's balance sheet. According to the companies, the platforms will "create dedicated pools of capital at significant scale at attractive rates for Nvidia customers." Chief Executive Jensen Huang told CNBC he approached only the six firms — and that none turned him down.
Compute as collateral
At the center of the plan is an unusual idea: treating Nvidia's silicon as a financeable, long-duration asset. In a post on X, Huang described the company's compute as "an investable infrastructure asset," the kind of thing that can be pledged as collateral the way a toll road or a power grid might be.
That is roughly how the mechanics are expected to work. People familiar with the plans told Bloomberg that special-purpose vehicles will issue debt — private placements as well as bonds — using compute power as collateral, then lease that compute to Nvidia's clients. Individual vehicles could raise tens of billions of dollars at a time, with the first deals coming to market within months. Because GPUs can, in theory, be reallocated to a different tenant if one buyer falters, the arrangement is pitched as lower-risk than a single-customer loan.
"We are bringing the world's leading long-term capital providers together to independently underwrite AI infrastructure," Huang said in the statement announcing the deal. "These financing platforms will help customers access scarce compute at scale and build the AI factories that will power every industry and country in the age of AI."
The Wall Street partners cast the effort as a natural marriage of supply and demand. "It's a big infrastructure build, and the capital markets are signaling that there's lots of capital available to support it," Goldman Sachs CEO David Solomon said on CNBC, adding that his firm is looking for ways of "getting the capital to the right places to extend this or accelerate this." BlackRock CEO Larry Fink argued the resulting debt would carry "high credit quality" and offer attractive yields to institutional investors who, in his words, are "overinvested in equities." Goldman is the only bank in the group and is positioning itself as lead bookrunner on the public debt deals.
Why the numbers are staggering
The scale reflects a capital problem the industry has not solved. Morgan Stanley estimates hyperscalers alone could spend roughly $3.5 trillion between 2026 and 2028, and the broader AI infrastructure buildout could exceed $8 trillion — figures that dwarf what even the richest tech companies can fund from operating cash flow. That gap has pushed the sector toward private credit, securitization and project finance, and it is why an announcement with a giant headline number but few structural details still landed as a milestone.
Nvidia's move also formalizes a role it has been drifting toward for months. The company had already been in talks to backstop as much as $250 billion to help OpenAI lease computing power from a $500 billion, 10-gigawatt SoftBank-affiliated hub in Ohio, and to finance a further $350 billion of OpenAI's chip purchases. Two years ago it committed to supporting the BlackRock-Microsoft-MGX vehicle now known as the AI Infrastructure Partnership. The new platforms knit those threads into a repeatable machine.
Why it matters: the circular-financing question
The announcement immediately reignited the debate that has shadowed Nvidia's deal spree — that a chipmaker financing its own customers' chip purchases risks manufacturing the very demand it reports. Nvidia has signed hundreds of billions of dollars in agreements across the AI ecosystem, and skeptics see a loop in which capital, revenue and valuations feed one another.
Huang has tried to draw a boundary. On X, he said Nvidia might provide financing support of "up to 25% of an opportunity" while leaving the rest to independent lenders: "Our role is to help unlock a very large pool of independent capital while maintaining disciplined risk exposure." Analysts are not fully persuaded. "In effect, they made Nvidia's product cheaper without really cutting GPU prices," Felix Wang, managing director of global technology at Hedgeye Risk Management, told Bloomberg. "But it also makes future demand more sensitive to credit conditions, credit volatility, and raises a lot of questions on what we consider to be real demand."
The deeper worry is one of fragility. Financing that leans on compute as collateral ties the AI buildout to credit markets, meaning a spike in rates or a wobble in AI economics could ripple through debt that was underwritten on the assumption of insatiable demand. "AI isn't necessarily a bubble, but the market needs an earnings reality check," said Terri Spath of Zuma Wealth. "We're very bullish on the earnings power of AI — where we exercise some caution is the price that investors pay for that growth." Traders at Wells Fargo and Mizuho said the plan did little to quiet those concerns.
What to watch next
The first test is whether headline ambition becomes real paper. The MOUs are non-binding, and no firm has committed a fixed dollar amount; the meaningful signal will be the initial special-purpose-vehicle bond and private-credit deals that insiders say could arrive within months, and the yields and credit ratings they command. Watch whether ratings agencies bless compute-backed debt as investment grade, how much risk Nvidia actually retains versus its stated 25% cap, and whether an OpenAI-scale transaction becomes the platform's marquee deployment. If the debt clears at attractive rates, it could unlock the trillions the buildout requires. If investors balk, the episode may become the clearest stress test yet of whether AI's financing engine can run as fast as its ambitions.
“In effect, they made Nvidia's product cheaper without really cutting GPU prices. But it also makes future demand more sensitive to credit conditions, credit volatility, and raises a lot of questions on what we consider to be real demand.”— Felix Wang, Managing Director of Global Technology, Hedgeye Risk Management