Bloomberg learned of Nvidia's problems with its AI project support program before Huang stepped down

Nvidia has faced challenges with its AI project funding program, and these issues arose even before the scandal involving CEO Jensen Huang’s affairs and his subsequent resignation, Bloomberg has learned. Months after striking a $500 billion partnership with six Wall Street giants, not a single deal under the agreement has materialized, sources told the agency. According to them, over the past month, some of Nvidia’s clients and partners have begun demanding that the company increase the size of the financial guarantees it was supposed to provide.

In August, Nvidia signed memorandums of understanding with six of the largest asset managers—Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR—to create a platform for financing computing power. The partnership aimed to “mobilize” $500 billion in capital for the chipmaker’s customers—AI model developers, companies, governments, and cloud providers—who could use these funds to build infrastructure based on Nvidia hardware.

Nvidia said it would provide residual value support covering up to 25% of the transaction amount. This is less than what its competitor Broadcom offers, and it is precisely this difference that is complicating the conclusion of agreements, according to some sources interviewed by Bloomberg.

According to one source, the earliest Nvidia could announce its first deal is January. This will happen provided that the dismissal of CEO Jensen Huang does not derail the negotiations or the entire project.

Bloomberg describes what has been happening as a “stunning turn of events” since August. The agency refers to this period as the peak of Huang’s “prime” on Wall Street.

Representatives from Nvidia, Apollo Global Management, BlackRock, Brookfield, Goldman Sachs, and KKR did not respond to a request from Bloomberg, and a Blackstone spokesperson declined to comment.

The market for debt financing to purchase chips has grown sharply over the past year: OpenAI, Anthropic, SpaceX, and other major players are seeking access to the expensive processors needed for specialized AI tasks. Spending has already reached the hundreds of billions of dollars, according to Bloomberg.

The guarantee allows lenders to receive compensation if the borrower stops making payments on the chip lease. In such cases, the assets are either leased out again or sold, and the equipment supplier covers the difference, Bloomberg explains. This improves the transaction’s credit rating and lowers borrowing costs.

In May, Broadcom provided $30 billion in guarantees as part of a $35 billion debt deal under which Anthropic leases tensor processing units (TPUs)—chips manufactured by Broadcom and Google. The portions of the debt backed by the guarantees received investment-grade ratings and lower interest rates, Bloomberg noted. A second deal, worth $63 billion, is also intended for Anthropic. According to the documents, Broadcom is providing guarantees for 72.5% of that amount, or about $43 billion.

If Nvidia is unwilling to raise the guarantee threshold above 25%, its chips could become less accessible to AI companies and cloud customers, and the deals could become less attractive to lenders, according to Bloomberg. This is especially true if, at the same time, similar deals are being offered on the market under more favorable terms.

On the other hand, a higher threshold would pose a greater risk for Nvidia. Furthermore, it would undermine the very concept of the platform—building infrastructure with third-party funding and fueling the next wave of the AI boom, the agency notes.

This article was AI-translated and verified by a human editor