Nvidia has opened talks with insurers about sharing the risk of loans that use its AI chips as collateral, the Financial Times reported on Tuesday. The discussions are at an early stage and may not produce any deals, according to the FT, which The Next Web says it has not independently verified.

One structure under discussion would insure loans to smaller cloud providers — the so-called neoclouds — that pledge Nvidia silicon as security. The cover would protect lenders in the scenario that matters most in GPU-backed lending: a borrower defaults, the repossessed chips are resold, and the recovery falls short of the outstanding debt.

The chipmaker is working with reinsurance broker Howden Re on one possible structure, the FT reported. Nvidia has already handed at least one insurer data on how quickly its chips depreciate and on the expected future value of the computing power they deliver. That depreciation curve is the crux of the matter. GPUs are not real estate; successive product generations can erode resale value quickly, and any insurer pricing this risk needs Nvidia's own numbers to do it.

Nvidia has also weighed joining groups of insurers, hedge funds and asset managers to back financing deals, according to the FT. Insurers could then pass part of the risk down the chain to hedge funds and other investors — a layered structure familiar from catastrophe and trade-credit markets, now applied to accelerators.

'An investable asset class'

The company's pitch to financial institutions is straightforward: treat its hardware like any other balance-sheet asset.

"AI infrastructure is an investable asset class because it's uniquely productive, durable and fungible," Nvidia has said, as quoted on the FT News Briefing podcast.

On the same podcast, the FT's insurance correspondent Lee Harris described a parallel development: start-ups are now selling "residual value insurance" that protects large chip buyers against the risk of faster-than-expected depreciation. Harris also flagged two constraints on how far this market can scale. Many large traditional insurers are already at their limit for exposure to AI companies, and the global insurance industry is small relative to the scale of AI investment overall.

Collateral that keeps financing itself

The report landed a day after Nvidia added a record $150bn to its share buyback — a signal of the cash generation behind the push. The lending machinery it aims to de-risk is already running. In August, the neocloud Lambda borrowed $917m to buy chips from Nvidia, which is also an investor in Lambda. The company later signed $1bn of private debt for another Nvidia chip deal.

The commercial logic is circular by design. Nvidia finances or de-risks the purchase of its own products, supporting demand from capital-constrained neoclouds while giving lenders collateral-backed comfort. Insurance would extend that comfort one step further down the risk chain, converting chip depreciation — normally a liability for buyers — into a tradeable, insurable risk.

Whether insurers bite depends on the same data Nvidia has been sharing: depreciation rates and residual compute value. If the talks mature into binding structures, chip-collateralized lending could expand well beyond the handful of neoclouds that can currently raise billion-dollar debt facilities on their own.