Nvidia courts insurers to underwrite residual-value risk on AI chip loans

What's the deal?

Nvidia has held early-stage talks with insurers about structures that would shift some of the risk of lending against its GPUs — including residual-value insurance if neocloud borrowers default and pledged chips cannot be resold for enough to repay lenders — according to people familiar with the discussions, reported by the Financial Times . The conversations may not lead to deals. Broker Howden Re is among the parties Nvidia has worked with on developing a structure; Nvidia has also shared chip-depreciation and compute-price data with at least one insurer.

Context

Jensen Huang has framed chips as an “investable asset class” akin to aeroplanes — durable, productive hardware that can support layered financing. The insurer push sits alongside Nvidia’s recent willingness to backstop Wall Street structures aimed at unlocking hundreds of billions for AI infra and to guarantee large leases for OpenAI data-centre builds. Niche residual-value insurers such as Forward Compute argue coverage can level the field for smaller neoclouds against hyperscalers by removing counterparty risk for compute buyers. Talks are at an early stage; Nvidia said AI infrastructure is uniquely productive, durable and fungible and that capital partners help expand compute access.

Source

Financial Times — Nvidia turns to insurers to spread the risk of AI build-out (Lee Harris & Ryan McMorrow, 29 Sep 2026). Discussions are early and may not result in transactions.