Nvidia’s $500 Billion AI Data Center Bet Builds a Used GPU Market
Nvidia backs a $500 billion AI data center plan, guaranteeing up to 25% of GPU value gaps to unlock institutional funding and a broader used GPU market.
Summary
TechCrunch reported on August 13, 2026, that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR were willing to commit up to $500 billion to AI data centers. Nvidia will use its own money to cover up to 25% of any gap between the booked and liquidation value of GPUs pledged as collateral after a borrower defaults. The guarantee is intended to attract independent, long term institutional capital, sustain Nvidia sales and create a secondary market for aging chips. Bond market concern prompted CEO Jensen Huang to explain the capped exposure on X and business television.
The arrangement creates wrong way risk because Nvidia’s obligations could rise as GPU demand and revenue fall. Nvidia has already committed billions to buyers including OpenAI, Anthropic, CoreWeave, Nebius, Firmus and Lambda, while Bloomberg calculated another $750 billion of circular deals was under development this summer. Unlike Lucent Technologies, which financed customer purchases before its dotcom era collapse, Nvidia leaves most capital and risk with outside institutions. That funding matters as Oracle adds debt, Google issues equity and Meta burns cash, while Microsoft CEO Satya Nadella recommended “1873,” about railroad financing that crashed the US economy. Huang argues Nvidia’s “AI factories” resemble reusable railroads or airlines rather than rapidly depreciating PCs because another customer, cloud or operator can take them over. Success could give startups, enterprises and researchers affordable aging hardware for varied AI workloads and open weight models. Slower AI use, efficiency gains or replacement technologies could instead make today’s infrastructure obsolete and trigger defaults, falling collateral values and payouts.
Positives
- Up to $500 billion from Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR could expand AI data center capacity.
- Nvidia’s 25% value guarantee could make GPUs more acceptable collateral while leaving outside institutions responsible for most capital and risk.
- A secondary market for aging Nvidia GPUs could give startups, enterprises and researchers more affordable hardware for varied AI workloads.
- Reusable “AI factories” could move among customers, clouds and operators, supporting residual values when individual owners’ needs change.
- Independent, long term institutional funding offers another financing source after Oracle, Google and Meta stretched debt, equity and cash respectively.
Risks & concerns
- Nvidia’s 25% guarantee creates wrong way risk because payouts could increase precisely when weakening GPU demand squeezes its revenue.
- Bloomberg calculated Nvidia was developing another $750 billion of circular deals after committing billions to OpenAI, Anthropic and several neoclouds.
- Bond market concern forced Jensen Huang to defend the plan on X and business television.
- The collapse of Lucent Technologies after financing customer purchases remains a cautionary comparison for Nvidia.
- Slower AI adoption, efficiency gains or replacement technologies could depress GPU values, increase defaults and render current data centers obsolete.