Nvidia CEO Jensen Huang is running into a problem: Many of his customers can't afford to buy his firm's coveted AI-powering chips.

That explains why Huang teamed up with an array of Wall Street firms on a $500 billion plan that will theoretically standardize chip financing, creating asset-backed pools of capital for AI companies -- while leaving Nvidia partly on the hook if things go wrong.

Executives involved in the strategic partnership Huang unveiled this week with Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs and KKR tout the effort as the launch of a new asset class akin to the securitization of everything from airplanes to credit cards to mortgages.

To critics, it is a system that will cover up weaknesses in some corners of the AI marketplace. The concern isn't about tech heavyweights -- Meta, Microsoft, Google -- with fortress balance sheets. Instead, it is about the many smaller AI labs, cloud companies and enterprises that have ravenous demand for Nvidia chips but face high interest rates if they want to finance purchases of them.

The AI boom, and Nvidia's financial momentum, can't keep up at this blistering pace unless those kinds of companies get the hardware they need.

Now, it will be up to Wall Street to convince debt investors of the merits of the new financing plan. No money has been raised yet under the new partnerships. The goal is to sell public and private debt to pension funds, insurance firms and sovereign-wealth funds, among others, and use the proceeds to establish dedicated "platforms" that can swing into action and finance AI chip deals.

One challenge: The chips powering data centers become obsolete over time as the technology advances.

"This is great for Nvidia, which needs its customers to have access to capital," said Jack Ablin, founding partner at the $260 billion family office Cresset, an investor in Nvidia. "But if you're a debt investor relying on compute power as collateral? I mean, historically, that's an asset that's had the shelf life of lettuce."

Huang has said Nvidia customers are still using old-generation chips, a sign that the hardware maintains its value long after it is released.

In a TV appearance, executives involved in the agreement said the extreme supply and demand mismatch for Nvidia's chips means they make excellent collateral, unlike fast-depreciating computer processors of the past.

BlackRock Chief Executive Larry Fink likened the current moment in financing of AI "compute" to the beginning of his career, when banks and investment firms pioneered the mortgage-backed securities market in the 1970s.

The exact terms of the AI funding arrangement are vague and will likely be dictated by what investors will agree to. But the idea outlined by executives involved is for loans to be backed by Nvidia's AI hardware that is being bought or leased.

If a customer defaults, the thinking goes, a new one will step in quickly. Investing in such an instrument could be attractive for a debt investor who thinks the AI industry will continue to grow but is concerned about the risk of lending to an individual customer, people involved in the deal said.

The question that underwriters will have to assess is what would happen if the AI build-out stalls and projects are delayed or canceled. Another risk is if technological breakthroughs by Nvidia undermine the value of its older chips faster than anticipated.

Nvidia, which has faced criticism for a series of circular funding arrangements where it invests in the companies that buy its chips or uses its balance sheet to guarantee their financing, has been looking for other approaches.

Huang said on X that Nvidia might in some cases use its balance sheet to backstop as much as 25% of the cost of a project through what he called a "residual-value support mechanism."

That would likely keep the obligations off Nvidia's books initially but put the chip maker on the hook for a portion of the project if an end user of its chips defaults. Under a residual-value backstop, a guarantor such as Nvidia agrees to make up the difference if the value of a given asset backing a loan falls below a certain floor.

Huang says the new funding structure addresses concerns about circular financing. Some skeptics still see it as a sort of circularity lite.

Adding third-party capital and credit decision-making "should arguably alleviate circularity concerns...while admittedly polarizing the debate, " Morgan Stanley analyst Joseph Moore wrote Tuesday.

Rival chip maker Broadcom provided similar support on a $35 billion financing package that Apollo, Blackstone and a group of banks arranged to lease chips to Anthropic in June, though for a much higher proportion of the offering. Broadcom has disclosed that its maximum exposure under the backstop is $29 billion, an amount that decreases as lease payments come in.

Raising money has become more expensive for AI companies across the board, but especially for weaker borrowers. They are being crowded out in the debt markets by a glut of supply from companies with investment-grade ratings like Meta Platforms, Google parent Alphabet and Amazon.com.

Issuance this year of AI-related bonds by big tech companies, data-center projects and chip-financing vehicles reached $344 billion in early August, according to Bank of America Global Research. That is up by more than $200 billion from what was issued in 2025. A Bank of America survey indicates that investors expect about $100 billion in additional bond volume from big tech companies in the remainder of the year.

AI cloud-computing provider CoreWeave raised a $2.6 billion loan facility this month, backed by its contracts with Anthropic, Jane Street and others, at yields 5.5 percentage points above benchmark interest rates, or over 9% currently. The company's bankers originally marketed the deal with a yield as much as 1.25 percentage points less than that. Investors demanded yields of almost 10% on a $3.5 billion junk-bond sale last month from Galaxy Digital to build a Texas data center leased to CoreWeave.

Michael Burry, an investor famous for shorting mortgage-backed securities who has lately bet against Nvidia and other AI stocks, said the deal points to vulnerability in the AI market. "Structuring credit is a natural part of the system," he said on Substack. "Structuring unnatural credits to prolong momentum late in the bull phase is where the worry comes in."