Arbol, the technology-led underwriter of parametric risk transfer, climate, and weather re/insurance business, has launched a new parametric insurance product that aims to provide protection against GPU residual value risk as artificial intelligence (AI) compute investments continue to accelerate.
The parametric cover offering, named GPU Residual Value Protection, is designed to provide protection for anyone carrying GPU value on its books, including hyperscalers, neoclouds, and other compute owners, as well as lessors, lenders, and residual value guarantors, against GPUs losing value faster than planned.
A GPU, or Graphics Processing Unit, is a high-value specialised computer chip that has become so integral to the artificial intelligence and data centre revolution.
Arbol explained that the parametric product will also pay in cash at the end of the term if a benchmark GPU rental index happens to drop below a strike level chosen by the client
Given this is a parametric product, payouts follow a formula agreed at inception, rather than a traditional claims process, meaning there is no sale, appraisal, condition test, or loss adjustment, and the owner keeps the hardware.
“GPUs are the most expensive asset in an AI build, and their value at the end of a loan, lease, or depreciation schedule is a risk that buyers, lessors, lenders, and guarantors all carry on their balance sheets. New chip generations now arrive every year, while many depreciation schedules run five years or longer. If the chips lose value faster than those schedules assume, the shortfall lands on whoever holds the residual: the owner, the lessor, the lender, or the guarantor,” Arbol explained
The company also confirmed that each structure is built around the client’s own exposure, with each client possessing the ability to select the strikes, tenor, and limit to match its loan, lease, or depreciation schedule, and the payout formula is fixed at inception.
Owners, including hyperscalers, neoclouds, AI labs, enterprises, and infrastructure funds, will be able use the parametric product to protect book value against faster depreciation.
Moreover, Arbol also emphasised how different market participants will be able to use GPU Residual Value Protection.
According to the firm, lessors can protect the residuals they book, while lenders and asset-backed investors can put a floor under GPU collateral, residual value guarantors will be able to hedge the guarantees they write, and borrowers can use it secure financing on better terms.
Because settlement depends solely on the index, every party sees the same trigger from the day the deal closes. The company stated that payouts do not hinge on a sale or a dispute over hardware condition, leaving owners free to keep running the chips, re-lease them or retire them on their own timetable.
Sid Jha, founder and CEO of Arbol, commented: “The AI buildout is creating risks that did not exist a few years ago, and the insurance market has been slow to give anyone a way to transfer them.
“What a GPU will be worth in three or four years is the question every owner, lessor, lender, and guarantor in this market is trying to answer. We pay on an index instead of a resale price or an appraisal, so anyone holding GPU risk has a number it can underwrite, book, or hedge against from the day the deal closes. This is the second product we have brought to AI and digital infrastructure risk this fall, and we expect to bring more.”
The launch follows on from Arbol’s recently introduced parametric Data Center Regulatory Moratorium Cover, which is designed to protect against the regulatory risk of government halting or suspending data center construction as a class, rather than site-specific permit denials.
Both of these products come out of Arbol’s AI and Digital Infrastructure practice area, which prices exposures created by the speed of AI infrastructure investment.
View all of our Artemis Live video interviews and subscribe to our podcast.
All of our Artemis Live insurance-linked securities (ILS), catastrophe bonds and reinsurance video content and video interviews can be accessed online.
Our Artemis Live podcast can be subscribed to using the typical podcast services providers, including Apple, Google, Spotify and more.





























