Nvidia teams with major lenders on $500 billion compute financing plan
Nvidia and a consortium of private-credit firms and banks have signed memoranda of understanding to create financing structures that treat GPU-based compute as a collateralised asset class.
Nvidia announced that it has entered into memoranda of understanding with a group of lenders that includes Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. The agreements are intended to provide up to $500 billion of financing that would allow cloud-service providers and other "neocloud" operators to borrow against Nvidia GPUs and related software as collateral, effectively turning compute capacity into an investable asset.
"This is really the first time that technology chips have become an investable asset class," Nvidia CEO Jensen Huang told CNBC. "These are revenue-generating assets now. They’re productive, they’re long-lived, they’re fungible, they’re flexible."
"This is the very beginning, like what it was when I started in the mortgage-backed securities market in the 1970s," BlackRock chief executive Larry Fink added.
The plan builds on earlier chip-backed loan structures, such as the $35 billion financing package Broadcom arranged with Apollo and Blackstone earlier this year. Nvidia says the value of its GPUs is supported by ongoing software improvements through its CUDA platform, which it claims can extend the economic life of older chips. Industry data shows rental rates for legacy GPUs have risen, and some cloud providers have doubled prices for certain models during contract renewals.
Analysts note that the success of the financing depends on the continued demand for inference workloads and the ability of borrowers to generate cash flow from contracts with large AI model developers. Concerns have been raised about the depreciation schedule of GPUs, which some experts estimate at two to five years, while Nvidia has suggested a longer, decade-long useful life for certain models.
The consortium’s financing would likely be structured as asset-backed notes, with lenders receiving senior claims on the GPU collateral and possibly on associated customer contracts. Nvidia has indicated it would provide residual-value support of up to 25 percent on some deals to mitigate lender risk.
If the agreements move beyond memoranda to binding transactions, they could make Nvidia-based compute cheaper to finance compared with competing hardware, potentially influencing data-center design standards and reinforcing Nvidia’s market position.