For months, Goldman Sachs, Blackstone, and Apollo Global Management worked quietly to structure debt deals that would help AI developers pay for Nvidia chips. Progress was slow. So Nvidia CEO Jensen Huang changed tack: he went public with the effort this week, announcing the group aims to collectively finance $500 billion in AI computing deals — a round figure Bloomberg notes has no obvious underlying calculation, but one designed to send a message.
Reassurance as Much as Financing
The announcement served a dual purpose. Beyond raising capital, Huang was seeking to reassure Nvidia's own investors that plenty of deep-pocketed firms stand ready to finance his customers — particularly AI startups like Anthropic and OpenAI, which are central to Nvidia's future demand. While bullish on AI spending broadly, Nvidia has also been working to diversify its customer base beyond hyperscalers such as Microsoft and Amazon, many of which are developing their own competing chips.
A Last-Minute Expansion
Just days before the announcement, Huang's $5.5 trillion company called back the original trio of financiers with news that three additional lenders — KKR, BlackRock, and Brookfield — were joining the group and committing to help finance a share of the debt. No deals had actually been signed by the time of the announcement, which sources say was left deliberately vague. With the partnership now public, major Wall Street firms are positioned to arrange hundreds of billions in financing for chip deals, while Nvidia itself will backstop a portion with its own guarantees.
Addressing Circular Financing Concerns
Investors have grown increasingly wary that Nvidia and other AI companies may be inflating industry demand through circular financing — arrangements where Nvidia invests in its own customers, such as CoreWeave. The financing venture initially unsettled debt investors worried about how much additional leverage it would expose the chipmaker to. That concern eased after Huang clarified Nvidia's backstop would cover no more than 25% of any given deal, with each project assessed case by case.
How the Structure Works
The $500 billion commitment has no fixed timeline, combining deals already in discussion with forecasts of near-term demand. Each lender retains the ability to vet individual borrowers for creditworthiness before committing capital. While much of the total will move through private credit markets, the sheer scale means public markets will also need to be tapped — likely through large-denomination bonds, some worth tens of billions of dollars each, issued by special vehicles that would lease chips to Nvidia's customers. Collateral backing the loans is expected to combine the underlying chips themselves with offtake agreements; if a customer can't pay, the chips could be leased out to other buyers, reducing the risk that any single default collapses the arrangement.
A "Debt Shopfront" Strategy
One person involved in the announcement described Huang's approach as effectively setting up a debt shopfront — a visible advertisement to both customers and skeptical investors that financing capacity exists at scale. The risk, this person noted, is reputational: if the deals don't materialize as announced or go poorly, it could damage the credibility of both Nvidia and its financing partners.
What's in It for Wall Street
For the participating firms, the venture promises fee income across multiple fronts. Goldman Sachs, the only partner with a dedicated banking arm, stands to benefit directly, while Apollo could unlock additional fees by expanding its trading operations and selling loan portions to other investors. Mercer's global head of real assets, Alan Synnott, said the announcement reflects genuine financing needs tied to the AI infrastructure buildout, predicting a range of strategies will emerge spanning infrastructure, real estate credit, and potentially private equity — offering investors multiple access points to the trade.
How It Compares to Broadcom's Earlier Move
Nvidia's high-profile announcement contrasts with a quieter, more concrete deal struck weeks earlier by rival chipmaker Broadcom, which tapped Apollo and Blackstone as anchor investors to finance more than 20 gigawatts of compute capacity for frontier AI labs including Anthropic and OpenAI through 2028. Unlike Nvidia, Broadcom already had $35 billion in financing secured through Apollo and Blackstone before going public with its plan, with Broadcom backstopping most of the debt to attract investors while Apollo structured the arrangement to keep it off Broadcom's balance sheet.
Other Banks Racing to Get In
The scale of the opportunity is drawing interest well beyond the original six partners. JPMorgan Chase's asset management arm is reportedly in discussions about joining the effort. And within minutes of Nvidia's Monday announcement, Morgan Stanley — already a significant AI infrastructure lender — announced its own framework aimed at facilitating $1.5 trillion in funding for US innovation and national security priorities, with AI and advanced computing topping its list.
The Skeptics' Case
Not everyone is convinced the underlying economics justify the scale of financing now flooding into the sector. Critics point out that current chip valuations are inflated by record AI-driven demand, raising the risk that today's aggressive infrastructure buildout could result in a meaningful oversupply of computing capacity years down the line — a concern that has shadowed much of the AI investment boom throughout 2026.
The Bigger Picture
Whatever the eventual scale of actual signed deals, Nvidia's announcement has already succeeded in one respect: pulling nearly every major name on Wall Street into public competition for a piece of the AI financing trade, from established credit giants to newcomers racing to stake their own claim. For continuing coverage, see Bloomberg Technology.
With no deals formally signed and a headline number that even insiders admit has no precise derivation, Nvidia's $500 billion plan looks for now like a bet that saying the number loudly enough will help make it real.