Nvidia's $500B AI financing deal, explained

Nvidia announced on 10 August 2026 that it is collaborating with six of the world's leading capital managers to raise more than $500 billion for AI data centers. However, the press release did not explicitly guarantee this amount. Instead, the figure of 25% backing by Nvidia came from comments made by its CEO Jensen Huang after the announcement.

The partners involved in this financing deal include Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. Among them, Goldman Sachs is the only bank and is responsible for leading public debt offerings and distributing returns through its asset-management arm.

The partnerships are represented by memorandums of understanding (MOUs), which are not legally binding agreements. The $500 billion target is also a long-term goal with no specific timeline attached to it. As of now, no formal agreements have been signed.

The structure of this financing deal involves a special-purpose entity that holds a pool of assets and the debt raised against it. This entity issues bonds, acquires computing resources, and leases them to Nvidia's customers. The GPUs serve as collateral for these loans. If a single AI company fails, the GPUs can be reassigned to other customers, making this structure less risky compared to lending to a single operator.

The 25% backing claim by Nvidia stems from the company's broad adoption of its GPUs, which can be used across various models and workloads. This claim is about the liquidity of Nvidia's chips, not an explicit financial commitment. The actual commitment regarding the 25% backing was made by Huang himself, who stated that Nvidia might provide financing support of up to 25% of a project, assessed on a case-by-case basis. Huang defended this arrangement on X, addressing critics of his statement.