On Monday, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to create financing platforms intended to mobilize more than $500 billion for AI infrastructure. The money will largely come from “third-party investors,” allowing Nvidia customers to finance chips and data centers while keeping Nvidia’s own risk limited and off the balance sheet.  

Details of the arrangements, like the extent of each deal, are still unknown. But analysts have been watching for a deal like this—that treats AI compute into an infrastructure asset, like a toll road or power plant—that produces cash flows and therefore can support debt. As of now, many have feared the chips instead look like a rapidly depreciating, and thus depleting, pile of graphics processors that will need more and more capital to finance. 

“We have moved from an era in which companies bought chips and built data centers project by project to one in which AI factories can be financed as productive infrastructure,” Nvidia CEO Jensen Huang wrote Tuesday. In Huang’s formulation, the premise is simple: “In AI, compute is revenue.” 

Underneath that transformation is a second one: who is actually paying for the AI boom. 

A year ago, most of Big Tech could claim it was financing AI from its enormous cash flows, accrued from decades of executing software-level thin margins and massive profits. But now debt is taking over. Goldman Sachs estimates AI-related financing now accounts for nearly one-quarter of all gross U.S. investment-grade issuance, while AI investment itself is approaching $600 billion this year.

So Nvidia’s getting ahead of the whole debacle to find the next pool of money. The chain is straightforward. An independent financing vehicle can raise money to buy Nvidia GPUs and data-center infrastructure. An AI company then leases that compute or commits to using it, creating a stream of payments against whichthe vehicle can borrow. Apollo, KKR, and their peers can structure or manage that debt and place it with the enormous pools of institutional money—mostly insurance and retirement capital—that they oversee.

Bloomberg columnist Matt Levine distilled the long-term vision into three steps: Put more private investments into ordinary people’s retirement accounts, raise “a gazillion dollars” of private-credit and infrastructure funds, and use that money to build the data centers that AI will rent.  

There is a reason those pools of money are attractive. Data centers are expensive, long-lived, and long-standing projects that require financing over many years. Insurers and pension funds, conveniently, have long-dated obligations—annuities that may pay for decades, or retirement benefits owed decades into the future—and therefore look for long-duration assets whose cash flows can be matched against those liabilities. Private-credit and infrastructure managers act as the middlemen, turning projects like data centers into debt those institutions can hold.

Nvidia, however, has said it’s putting something of its own behind the bet: Huang said the company may provide residual-value support of up to 25% for some projects—effectively promising some protection against the possibility that the chips backing a financing are worth much less in the future than lenders expected.

Ben Thompson, who writes the technology strategy publication Stratechery, calls that “in a certain sense, a price cut”: Nvidia is using its own profits to reduce customers’ cost of capital and make Nvidia-based data centers easier to finance. 

And that is where the deal gets more interesting. The AI boom started with some of the richest corporations in history spending their own cash. Then came bonds. Now Nvidia is helping Wall Street turn compute itself into an investable asset capable of drawing on insurance floats, pension funds, and other long-duration savings.

Wall Street sees it as a positive sign. Morgan Stanley’s Joseph Moore said the arrangement alleviates concerns about circular financing because third-party investors would provide most of the capital, while Nvidia would participate only to a limited extent. Bank of America’s Vivek Arya mostly agreed and added Nvidia’s chips are unusually financeable because GPUs can be moved among operators and CUDA software can extend their own useful lives.

But Thompson’s concern is those pools of cash institutions have are fundamentally different from venture capital or tech stocks: They are designed, at least in part, to seek safety. 

“It’s one thing to spend all of your free cash flow; it’s another thing to tap the debt markets,” he writes. “And, beyond that, it’s a completely new nerve-racking thing to bring safety-seeking assets to bear.”