Nvidia has unveiled a financing structure to draw funding of up to $500 billion to build artificial intelligence (AI) data centres. It will directly guarantee part of the value of its GPUs used as collateral. The move is seen as a strategy to secure large-scale AI infrastructure funding while invigorating the secondary market for older Nvidia GPUs.

On Aug. 13, local time, IT outlet TechCrunch reported that Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR, among others, have expressed willingness to invest up to $500 billion to build AI data centres.

The core of the structure is financing backed by GPUs. Data centre owners take out loans using Nvidia GPUs as collateral. If they fail to repay and the GPUs are sold off, and the collateral value falls below expectations, Nvidia covers up to 25 percent of the shortfall. It is a way to reduce lenders' risk by guaranteeing part of the GPUs' residual value.

Financial markets are assessing the structure as unusual. If GPU demand slows, used GPU prices would fall and collateral values would drop, which could increase Nvidia's guarantee burden. That creates a kind of reverse risk because reduced GPU sales and higher guarantee costs could occur at the same time.

Nvidia CEO Jensen Huang (젠슨 황) appeared to address such concerns directly. On X, formerly Twitter, he referred to a question asking, "Is this circular finance?" He explained that the plan is instead intended to resolve that concern. Nvidia says the structure leaves most of the funding and risk to external financial institutions, while the company protects only part of the future value of its GPUs.

Nvidia is not aiming only to secure data centre funding. The key is to create a market in which chips displaced by newer GPUs over time can be transferred to other operators, cloud companies or data centre operators and continue to be used.

Huang said that even if demand changes, data centres can be used by other customers, cloud providers or operators. He stressed that as Nvidia's computing ecosystem expands, it helps protect the residual value of GPUs. The strategy is to extend hardware life cycles by building an ecosystem in which used GPUs circulate, rather than ending with the sale of new GPUs.

Nvidia has already expanded large funding and supply commitments to AI companies and data centre operators. It has conducted deals worth billions of dollars with OpenAI and Anthropic, as well as CoreWeave, Nebius, Peramers and Lambda. CoreWeave is cited as a representative case of a financing structure that uses Nvidia GPUs as collateral.

In the background is the growing burden of raising funding needed for AI data centre investment. Oracle's expansion of large-scale debt, Google's issuance of new equity stakes and Meta's cash spending show that AI infrastructure investment is placing a significant burden on corporate finances.

Ultimately, the success of the plan depends on how long AI demand lasts. If AI adoption slows or more efficient technologies emerge, GPU prices and collateral values could fall at the same time. If demand for AI infrastructure continues, Nvidia is expected to secure a market in which used and ageing GPUs are traded, beyond selling new GPUs, and raise the residual value of its chips.