Nvidia backs $500B data center plan, guarantees used GPU values

Nvidia this week announced a landmark partnership in which Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR committed up to $500 billion toward building AI data centers. The centerpiece of the deal is a value guarantee: Nvidia agreed to back its own chips used as collateral for these ventures, promising to cover up to 25% of any gap between the chips' expected price and their liquidation value should a borrower default.

The move has drawn intense scrutiny from financial markets and commentary alike. CEO Jensen Huang felt compelled to explain the strategy on X and business television, defending it against concerns about circular financing and comparisons to Lucent Technologies, which famously lent customers money to purchase its own telecom equipment before crashing during the dotcom collapse. Huang acknowledged the worry directly on X: "Is this circular financing? This initiative is designed to address that concern. We are bringing independent, long-term institutional capital into the AI infrastructure market." The distinction matters: Nvidia is not lending the money itself; instead, it is inviting blue-chip financial firms to deploy capital while Nvidia absorbs a portion of the collateral risk.

The underlying economics reveal why Nvidia took this step. Traditional funding sources for AI infrastructure have begun to tighten. Hyperscalers have strained themselves—Oracle accumulated significant debt, Google issued new equity tranches, and Meta burned through cash. Microsoft CEO Satya Nadella recently cited "1873," a historical account of railroad-era financial engineering that precipitated a national economic crash, signaling alarm among industry leaders about the sustainability of current funding models. Bloomberg has calculated that Nvidia is working on another $750 billion worth of similar circular deals this summer.

But underneath the immediate financing puzzle lies a deeper strategic insight: Nvidia is deliberately attempting to nurture a secondary market for aging GPUs. As Huang framed it, AI servers are not quick-depreciating assets like PCs but long-term "investable infrastructure" comparable to railroads and airlines. "When needs change, the factory can be used by another customer, another cloud or another operator. This broad ecosystem gives NVIDIA compute a deep market of potential users and offtakers, helping protect residual value," he said. The guarantee of up to 25% price protection on collateral is designed to stabilize that secondary market and keep demand alive for hardware as it ages.

The risk to Nvidia is substantial. Financiers call this "wrong way" risk: Nvidia's obligations expand precisely when demand weakens. Should adoption slow, new technologies make current infrastructure obsolete, or enterprises and consumers temper their AI spending, both the collateral value and Nvidia's own revenues could deteriorate simultaneously. If today's AI boom, where capacity still lags demand, cools before AI becomes embedded as foundational infrastructure, the entire structure could unravel. Yet if Huang's vision holds—if AI infrastructure anchors long-term growth the way utilities and transportation did—then Nvidia's guarantee becomes a cheap insurance premium on a durable revenue stream.