Nvidia announced a financing initiative in which Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR have agreed to commit up to $500 billion toward building AI data centers. The company is also proposing a mechanism to support a secondary market for aging GPUs by guaranteeing part of their future value.
Deal structure and GPU guarantees
Under the plan, Nvidia will put its own money behind chips that serve as collateral in these financing deals. The company said it will cover up to 25% of any gap if GPUs sold to repay loans do not fetch the expected price. Jensen Huang wrote on X that the initiative aims to bring “independent, long-term institutional capital into the AI infrastructure market.”
The scheme builds on earlier arrangements in which Nvidia committed financing to several buyers, including OpenAI, Anthropic and neoclouds such as CoreWeave. Bloomberg has calculated Nvidia has been working on roughly $750 billion of related circular financing this summer, according to reporting cited by observers.
Risks, comparisons and market context
Commentators have described the plan as unconventional, clever and risky. Financial-market specialists warn the guarantees create “wrong way” risk: Nvidia’s obligations could rise precisely when demand for its hardware weakens, potentially coinciding with pressure on the company’s revenues.
Some observers have compared the approach to past vendor-led financing models, including Lucent Technologies’ lending practices around the dotcom bubble. Nvidia has pushed back on that comparison by emphasizing that third-party institutional investors will supply most of the capital while Nvidia protects only a portion of chip residual value.
Nvidia framed the strategy as a way to make AI compute an investable infrastructure that retains residual value because hardware can be re-used by other customers, clouds or operators. The company noted that if GPUs used as collateral fail to meet book values upon liquidation, it will cover up to 25% of the shortfall.
Original source: TechCrunch AI