Nvidia backs $500 billion AI data center plan with chip value guarantee

Nvidia is set to facilitate the construction of AI data centers valued at up to $500 billion, drawing commitments from financial giants including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. This ambitious initiative is structured around a unique financial guarantee from Nvidia itself.

“

Nvidia has agreed to cover up to 25% of any loss in the value of its chips if they are used as collateral and their market value falls below what the books say they are worth.

The company will cover up to 25% of any potential loss in the value of its chips if they are used as collateral for loans and their market value drops below the book value. This means if a data center owner defaults and lenders liquidate the chips, Nvidia will absorb the difference, up to the specified limit.

Nvidia CEO Jensen Huang has pushed back against comparisons to past corporate financing schemes, such as Lucent Technologies' role in the dot-com bubble. He argues that the majority of capital and risk in this venture rests with independent financial institutions, not Nvidia. Huang frames the plan by likening AI hardware to essential long-term infrastructure, such as railways or airlines, rather than rapidly depreciating consumer electronics.

“

Unlike Lucent, the bulk of capital and risk is being shouldered by large independent financial institutions rather than by Nvidia itself.

A key element of this vision is the establishment of a robust secondary market for older Nvidia GPUs. This would ensure continued demand for the company's hardware even as newer generations emerge. However, the plan carries a "wrong way" risk, where Nvidia's guarantee obligations could increase precisely when its revenues are most likely to face pressure. Huang has publicly clarified that Nvidia's direct financial exposure is intentionally limited to mitigate market anxieties.

“

AI hardware is long-term infrastructure, comparable to railways or airlines, rather than a fast-depreciating consumer electronics category like personal computers.