Nvidia unveils $500B AI data center plan with GPU value guarantee

Nvidia has secured commitments from Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to invest up to $500 billion in AI data centers, with the chipmaker agreeing to guarantee a portion of its GPUs' residual value. The announcement, made the week of August 13, 2026, sent shockwaves through bond markets, prompting CEO Jensen Huang to publicly defend the plan on X and business television.

The initiative aims to create a secondary market for aging Nvidia GPUs by assuring financial backers that the chips used as collateral will retain value. Under the terms, if a data center owner defaults and the GPUs sell for less than expected, Nvidia will cover up to 25% of the shortfall. This guarantee is designed to attract long-term institutional capital to AI infrastructure projects, addressing concerns about circular financing that have dogged the company.

Analysts have drawn comparisons to Lucent Technologies, the telecom equipment maker that collapsed after lending money to its own customers during the dotcom bubble. However, Huang argues the new scheme is different because independent investors bear most of the risk, with Nvidia only backstopping a fraction of potential losses. The company has already committed billions to buyers like OpenAI, Anthropic, CoreWeave, Nebius, Firmus, and Lambda, and Bloomberg calculated it was working on another $750 billion in circular deals this summer.

The financial maneuvering comes as traditional funding sources for AI data centers show strain. Hyperscalers have taken on significant debt, issued new equity, or burned through cash reserves—Oracle, Google, and Meta are cited as examples. Microsoft CEO Satya Nadella even recommended the book "1873," about railroad-era financial engineering that crashed the U.S. economy, during a recent earnings call.

Huang frames AI servers as "AI factories" akin to railroads or airlines—long-term infrastructure with deep secondary markets—rather than rapidly depreciating assets like PCs. He argues that when needs change, a factory can be repurposed by another customer, cloud, or operator, protecting residual value. This vision underpins Nvidia's bet that demand for AI compute will remain robust, even as new technologies emerge.

Should the plan succeed, Nvidia will unlock new capital sources for data center construction and sustain revenue as its installed base ages. The risk is "wrong way" risk: Nvidia's obligations would grow precisely when demand weakens, squeezing revenues at the worst possible time. For now, the company is using its market power to shape the future of AI infrastructure financing.