Nvidia’s new $500B plan is risky but brilliant, especially for aging GPUs
Nvidia has outlined a plan under which major financial firms could commit up to $500 billion to AI data centres, while Nvidia guarantees part of the future value of GPUs used as loan collateral. The proposal matters because it could unlock new funding for AI infrastructure and create a stronger resale market for older Nvidia chips, but it also leaves Nvidia exposed if demand weakens and GPU values fall.
Nvidia would cover up to 25% of any shortfall if a borrower defaults and the collateralised GPUs sell for less than expected. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR are involved, while Nvidia has also backed chip buyers including OpenAI, Anthropic and several cloud providers. Critics compare the arrangement with Lucent’s dotcom-era financing, although Nvidia argues outside institutions would bear most of the capital and risk.
- Nvidia may help finance $500 billion in AI data centres.
- It guarantees part of ageing GPUs’ resale value.
- The plan could boost demand but increases downside risk.
New here? Start with this
Nvidia makes the specialised computer chips, known as graphics processing units or GPUs, that are widely used to train and run artificial intelligence systems. Building the data centres that house large numbers of these chips requires vast spending on buildings, power supplies and equipment, so technology companies often rely on loans and outside investors.
The proposal involves banks and investment firms lending against GPUs as collateral, meaning they could sell the chips to recover money if a borrower cannot repay. Nvidia’s promise to cover part of any loss would make those loans less risky for lenders, while potentially giving older chips a clearer resale value.
The financial firms involved are major investors in infrastructure and private credit, and the chip buyers include AI companies and cloud-computing providers. The arrangement matters because it could speed up AI data-centre construction, but Nvidia could face costs if demand for its chips falls and used GPUs are worth less than expected.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters argue that the plan is a pragmatic way to finance infrastructure whose upfront costs are unusually large but whose productive value can extend beyond an initial deployment. A limited Nvidia guarantee could give lenders sufficient confidence to fund data centres while establishing credible resale prices for older GPUs, reducing waste and broadening access to capable hardware. They contend that outside financiers still supply most capital and retain most risk, so the arrangement can accelerate AI investment without Nvidia assuming the full burden.
The case against
Critics argue that guaranteeing GPU collateral risks turning Nvidia from a supplier into a backstop for an overheated investment cycle, especially if AI demand or chip prices fall sharply. The arrangement may encourage lenders and buyers to rely too heavily on Nvidia’s support, potentially obscuring the true economic value of data-centre projects and amplifying losses if borrowers default. Comparisons with vendor financing in the dotcom era reflect concern that sales could be sustained by credit arrangements rather than durable end-user demand.