Nvidia recruits Wall Street for a $500 billion AI compute push

Why it matters

Huang is turning financing into a distribution tool for Nvidia: cheaper capital can expand chip demand, while lenders inherit the risk that today's costly AI systems retain value.

Jensen Huang (@JensenHuang) recruited six of Wall Street's largest investment firms to help Nvidia customers finance more than $500 billion of AI infrastructure, an attempt to turn access to capital into another engine for Nvidia hardware sales.

Nvidia announced Monday that it signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. Huang also shared the announcement on X.

The wording matters. The $500 billion is a target for third-party capital that the proposed platforms would mobilize "over time." Nvidia did not announce a funded $500 billion pool, a schedule for deploying the money, individual commitments from the six partners, borrowing rates or a list of approved projects. The partnerships also remain subject to final agreements.

If completed, the platforms would create dedicated pools of capital for Nvidia customers, including frontier AI laboratories, enterprises and cloud operators. Nvidia said the financing would be offered at "attractive rates," another term left undefined in the announcement.

Huang is extending Nvidia's role beyond supplying processors, networking equipment and software. The Nvidia co-founder, who started Nvidia in 1993 after working at AMD and LSI Logic, is now trying to establish Nvidia compute as an infrastructure asset that large financial institutions can underwrite for years.

"In AI, compute is revenue," Huang said in Nvidia's statement. He argued that Nvidia systems can retain economic value because they are usable across models and workloads, transferable among operators and supported by Nvidia's CUDA software. Those are Nvidia's claims about its own equipment, and they form the investment case that Apollo and the other firms will have to test before providing capital.

Nvidia is financing the market around its chips

The arrangement addresses a constraint that has moved closer to the center of the AI buildout: Nvidia's customers need enormous amounts of debt and equity capital before they can buy chips, secure power and construct data centers. Helping those customers borrow can accelerate their purchases from Nvidia.

Nvidia's announcement was unusually direct about that incentive. The proposed platforms are designed to support Nvidia's growth across hardware sales and software adoption while creating long-duration, usage-linked revenue for infrastructure owners. Goldman Sachs CEO David Solomon said the partnership could help create a credit market backed by Nvidia compute.

That structure would push lenders to evaluate GPU systems partly as financeable assets rather than treating every data center loan as a bet on a single AI company's future cash flow. The transferability and useful life of the equipment will be central to that underwriting. If the original borrower fails, lenders need confidence that another operator can use the same systems at sufficient utilization and pricing.

The six partners bring large pools of insurance, pension, private-credit and infrastructure capital. Their participation gives Huang a route to expand the buying power of Nvidia customers without requiring Nvidia to supply all of the money itself.

The structure does not eliminate circular financing concerns

The proposed platforms arrive as Nvidia's financial ties to major customers draw closer scrutiny. The Wall Street Journal reported that Nvidia was discussing a roughly $250 billion financing backstop tied to an OpenAI data center project in Ohio. Nvidia also announced in September 2025 that it planned to invest up to $100 billion in OpenAI as the AI developer deployed Nvidia systems.

Monday's announcement does not identify OpenAI as a recipient, and Nvidia did not say it would guarantee the new platforms' loans. That distinction separates the plan disclosed Monday from a direct balance-sheet backstop. It does not remove Nvidia's economic interest in the outcome: financing Nvidia customers creates more potential buyers for Nvidia products.

The unanswered questions sit in the eventual contracts. The memorandums do not show which party will absorb losses, how Nvidia equipment will be valued as it ages, whether Nvidia will provide guarantees or repurchase commitments, or how much capital each investment firm will supply.

For Huang, the strategic objective is clear. Nvidia already designs the compute systems. Huang now wants Wall Street to build a financing market around them, widening the pool of customers that can afford AI infrastructure and tying the availability of capital more closely to Nvidia's platform. The $500 billion figure measures the ambition of that market. It does not yet measure committed money.