Nvidia Corp. is expanding a circular financing network through off-balance-sheet deals and new investments to fund customer purchases of AI chips.

This strategy allows the company to sustain rapid growth in AI demand by providing the capital its customers need to buy expensive hardware. However, the practice has drawn scrutiny from Wall Street analysts who worry the arrangement masks a financing bubble.

Nvidia recently announced memorandums of understanding to provide $0.5 trillion [5] in financing for its customers. This effort is part of a broader web of AI deals that some reports value at approximately $750 billion [3]. Other estimates suggest that off-balance-sheet AI obligations across major hyperscalers have ballooned to $3 trillion [2].

Central to this expansion is a relationship with SB Energy, a subsidiary of SoftBank. Nvidia has invested $1.5 billion [1] in the energy firm, though some reports indicate the company is in talks to invest up to $3 billion [4]. SB Energy is currently building a new data-center campus in Ohio for OpenAI.

The arrangement creates a loop where Nvidia provides the capital or investment that enables its partners to build infrastructure and purchase Nvidia GPUs. This mechanism ensures a steady stream of revenue for the chipmaker while accelerating the deployment of AI hardware across the U.S. tech sector.

Critics argue that this circularity may inflate demand figures by relying on credit rather than organic capital. The company has attempted to quiet these accusations, but uncertainty remains regarding the long-term stability of these financial structures.

Nvidia is expanding a circular financing network through off-balance-sheet deals and new investments.

The scale of Nvidia's financing indicates a shift from being a mere hardware provider to acting as a primary financier for the AI ecosystem. By facilitating the loans and investments used to buy its own products, Nvidia reduces the immediate friction of high hardware costs but increases the systemic risk if the projected AI returns fail to materialize for the end users.