Nvidia wants to redistribute the risks associated with building AI infrastructure. What will this change?

Chipmaker Nvidia has approached insurance companies with a number of proposals that would shift some of the risks associated with capital-intensive semiconductor financing to insurers and other investors, the Financial Times reports, citing sources familiar with the negotiations. Among the ideas under discussion is insurance against losses on loans issued to small cloud computing companies. In this way, Nvidia CEO Jensen Huang aims to stimulate demand for semiconductors among customers outside the group of tech giants, the newspaper notes.

“Nvidia is trying to get the market on board, to show other capital providers that these [AI chips] are investment assets,” noted one of the publication’s sources. Jensen Huang himself said in August that semiconductors should be viewed precisely in this light—as an “asset class,” similar to other expensive, durable equipment—such as airplanes—for which complex financial structures are created to distribute risks and costs among users and investors— was discussed in August by Jensen Huang himself, the FT notes.

To this end, according to the newspaper’s sources, Nvidia held talks with insurance companies to have them assume the risks associated with loans secured by its chips. In the case of cloud-based startups (“neo-clouds”) — in the event of their insolvency — such insurance could cover losses if the Nvidia chips pledged as collateral for their debts cannot be resold at a price high enough to repay their obligations to creditors, the FT explains.

Such protection, the publication continues, has the potential to attract more capital to that segment of Nvidia's customer base that does not have the same financial resources as large technology companies.

However, the newspaper notes that Nvidia’s negotiations with insurers on this matter are still in the early stages and may not result in any agreements. At the same time, Nvidia has provided at least one insurance company with data on chip depreciation and the expected future value of computing power, one of the FT’s sources said. Another source told the publication that Nvidia is working with the broker Howden Re to develop a mechanism involving insurers. Howden declined to comment to the newspaper.

In addition, Nvidia is considering joining a consortium that would support such agreements alongside insurers, hedge funds, and asset managers, another source added.

Nvidia shares rose 0.8% in premarket trading on September 29; they are up 22.7% year-to-date.

The financial mechanisms Nvidia is considering are similar in principle to so-called residual value insurance—protection against the depreciation of technological equipment, the FT explains. There are already players in the market, such as Forward Compute and American Compute, offering such products. Forward Compute CEO Kanthen Saler noted that such insurance products could level the playing field for small cloud providers by eliminating the risk of their bankruptcy before a contract is fulfilled. “Large buyers of computing power don’t want to deal with the counterparty risk of a neo-cloud provider,” Saler explained. “But with insurance, small neo-cloud providers can compete with Amazon or Google, because the counterparty risk immediately becomes the same.”

The negotiations are taking place as insurers launch a whole series of products designed to support the development of artificial intelligence infrastructure. These include insurance against credit risks and declines in the value of chips, as well as coverage for losses resulting from breach of contract caused by power outages or cooling system malfunctions in data centers, the newspaper also notes.

An upcoming study by Barkr AI, a company that specializes, among other things, in evaluating AI chips, found that an Nvidia H100 server system with eight 2022-model chips currently costs about $320,000 — the same as when sales began, according to the FT. The authors predict that if supply catches up with demand, the system will retain two-thirds of its price in a year and will drop to about $30,000 in six years. “Nvidia needs valuation data to attract outside capital,” explained Thomas Galbraith, founder of Barkr AI, whose service helps fill the gap in secondary market data. “It’s important for lenders to understand how much revenue an asset will generate and how much it will sell for over time.”

This article was AI-translated and verified by a human editor