Nvidia is asking insurers to help protect lenders from losses if smaller cloud providers default on loans backed by its AI chips.
According to the Financial Times Nvidia is discussing insurance for loans to “neocloud” companies. If one of them goes bust and its chips sell for less than the outstanding debt, insurance could cover some of the shortfall.
The talks are in the early stages and may come to nothing. Nvidia has shared data on chip depreciation and the future value of computing power with at least one insurer, and is working with broker Howden Re on a possible structure.
The idea could make lenders and chip-leasing firms more comfortable financing smaller providers. They lack the deep pockets of Amazon, Google and Microsoft, and their lenders have less room to absorb a bad bet.
Nvidia chief executive Jensen Huang has called AI infrastructure an “investable asset class”, comparing chips with durable equipment such as aircraft. The company has explored joining financing deals alongside insurers, hedge funds and asset managers.
The insurer talks are part of Nvidia’s wider effort to pull more outside money into AI infrastructure. The company has backed financing plans intended to unlock up to $500 billion from financial firms, and has guaranteed $105 billion in leases for a data centre being built for OpenAI.
One possible product is residual value insurance, which protects against equipment losing value. Insurers are beginning to explore cover for AI hardware as lenders try to work out what the chips will be worth after years of use.
A forthcoming Barkr AI study estimates that an eight-GPU Nvidia H100 system introduced in 2022 is still worth about $320,000, close to its initial value. It projects that figure could fall to roughly $30,000 after six years if computing capacity catches up with demand.
Those figures are estimates, not guaranteed resale prices. Lenders still need to judge how much revenue a chip system can generate and what it might fetch later.
Nvidia’s plan rests on persuading outside investors that AI hardware will retain enough value to support the debt used to buy it. Insurers may help spread that risk, provided they like the numbers.
