Jensen Huang says strong AI demand supports faster data center deployment and GPU-backed financing

Nvidia CEO Jensen Huang said strong AI demand, faster data center deployment and the lasting value of GPUs could help ease the financing burden of AI infrastructure investment.

Bringing computing capacity into service sooner would allow operators to start earning revenue earlier, while equipment with substantial residual value could provide collateral for borrowing. Huang argued that demand gives the industry a basis for addressing its supply and financing constraints.

In an October 21 interview with Yahoo Finance, Huang said shortening the gap between equipment delivery and operation would make data center projects easier to finance. He described demand as the essential condition for resolving the industry's other challenges.

A gigawatt-scale data center costs about $50 billion, Huang said. Even after computers arrive at the site, wiring and starting the systems can take four to six months, leaving capital tied up in equipment that has yet to generate revenue.

Moving some assembly and installation work into the supply chain could shorten that waiting period, he said. Discussing the DSX pod, Huang said faster deployment would bring forward revenue and reduce idle capital, easing the funding needs of AI cloud operators between equipment delivery and commercial operation.

Huang also pointed to the residual value of Nvidia computing equipment. Systems once assumed to have a four-year useful life could remain productive for eight years or longer, he said. Continued software improvements can extend their usefulness, while other businesses can take over capacity that an existing customer no longer needs.

That productivity, fungibility and durability could support loans secured against the equipment itself, Huang argued. He said Nvidia computing systems retain substantial value and suggested that Vera Rubin could have a useful life of seven to eight years, supported by software upgrades and CUDA.

Power remains a major bottleneck, Huang said. Supply constraints have shifted through wafers, advanced packaging, high-bandwidth memory, silicon photonics and connectors, and could persist for another couple of years. As AI factories expand, land, buildings, generator deliveries, grid connections and construction labor also constrain deployment.

Differences in power availability and permitting across markets give Nvidia flexibility, he said. Its broad customer base allows it to ship equipment to customers ready to use it while working with others to resolve local constraints and deliver systems once those obstacles clear.

Huang said AI is reaching a turning point in practical usefulness, with demand growing across open and closed models. More users performing useful tasks are driving token demand, he said, adding that capital markets and a range of financing instruments must help support infrastructure construction and make funding more accessible to customers.