NVIDIA is in negotiations with several insurance companies to discuss how to share the risks associated with lending that are secured by their chips. Jensen Huang hopes to use this opportunity to open up demand for their chips beyond large technology corporations.

According to people familiar with the matter, this chip manufacturer has approached several insurance institutions to discuss various transaction structures, with the aim of transferring some of the risks associated with its capital-intensive semiconductor financing activities to insurance companies and other investors.

One of the schemes under discussion is a loan loss insurance for emerging cloud service providers ("new clouds"): in the event that such companies default on their debts, and the resale income from NVIDIA chips used as collateral is not sufficient to repay the principal and interest to the lenders, then the insurance will cover the loss.

Such safeguard tools can direct more capital towards a group of NVIDIA's customers—these enterprises do not possess the robust balance sheets of large technology conglomerates.

Currently, the related discussions are still in the early stages and may not necessarily lead to actual transactions. However, these negotiations indicate that this company, which has the highest market value globally, is expanding its reach beyond Wall Street and private capital to the insurance industry as well. It is experimenting with various financial structures in order to increase the range of customers who can purchase its chips.

Jensen Huang has proposed that chips should be regarded as a type of investable asset, analogous to expensive, long-lived equipment such as airplanes; the aircraft industry has developed complex financial instruments to spread risks and costs between users and investors.

A person familiar with NVIDIA's fundraising efforts said, "NVIDIA is working to engage the market and prove to other investors that chips are assets with investment value."

While the insurance negotiations are progressing, insurance companies are intensively launching various products tailored for the infrastructure construction of AI, covering credit risk protection, compensation for chip depreciation losses, as well as protection against contract breach risks caused by power outages and cooling system failures in data centers.

Last month, Jensen Huang's ambitious goals became more apparent. NVIDIA had previously promised to provide a guarantee for some financing transactions, hoping to leverage a total of $500 billion in funds from Wall Street institutions such as Goldman Sachs and Apollo. NVIDIA also provided a $105 billion lease guarantee for the giant data center project of OpenAI. NVIDIA told investors that it expects a quarter of next year's revenue to come from the AI laboratories, which are funded by the company's own balance sheet.

Negotiations with insurance companies represent a new front. The relevant framework can provide protection for chip lessors and institutions that offer loans secured by chips, transferring a portion of the risk to insurance service providers.

According to informed sources, NVIDIA has provided at least one insurance company with data regarding chip depreciation and forward prices for computing power. Another source mentioned that NVIDIA is collaborating with the reinsurance broker Hodder Reinsurance ( Howden Re ) to develop an insurance transaction framework, but Hodder has declined to comment on this.

According to the participants in the discussion, the project is led by Ingo Ranev, the person in charge of NVIDIA's financial solutions ( Ingemar Lanevi ). NVIDIA is considering leveraging insurance institutions to further reinsure the risks with hedge funds and other alternative investors; the reason being that the potential volume of these transactions is so large that even the balance sheets of large insurance companies are unable to bear them alone. There are also reports that NVIDIA has discussed forming a consortium with insurance companies, hedge funds, and asset management institutions to jointly support such agreements.

NVIDIA officially stated: “The AI infrastructure possesses the attributes of investable assets; it features high production value, durability, and versatility. Leading capital partners are helping NVIDIA and the entire ecosystem to expand the financing scale for the construction of AI factories, making it easier for native AI enterprises to access computing power.”

At the same time, NVIDIA is also making large-scale returns to its shareholders, announcing a record $150 billion stock repurchase plan on Monday.

The financial structure studied by NVIDIA is similar to "surplus value insurance," which is specifically designed to protect against the depreciation of technology equipment assets. Emerging companies such as Forward Compute and American Compute have already launched such niche products.

Forward Compute CEO Quentin Sallerer stated that insurance products can level the competitive disadvantages of smaller and medium-sized cloud vendors and mitigate the risks associated with companies going bankrupt before fulfilling their obligations.

"Larger buyers of computing power are reluctant to assume the counterparty risk from emerging cloud providers," said Saller. "But with the introduction of insurance, smaller new cloud service providers can compete on an equal footing with Amazon and Google, as the counterparty risk is mitigated by insurance."

Forward Compute indicates that the company is referring to the chip forward-looking valuations provided by institutions such as Barkr AI, Sil, and Data. These institutions have become important sources of data for this emerging market.

Barkr AI, which specializes in asset valuation for artworks, private jets, and AI chips, is about to release a research report: In 2022, the current value of a NVIDIA 8-card H100 system is approximately $320,000, which is basically equivalent to the initial purchase price. The report predicts that once the supply of computing power catches up with demand, this device will still retain about two-thirds of its value after one year; after six years, the residual value will drop to around $30,000.

Barkr AI Founder Thomas Galbraith stated: "The entry of external capital is inseparable from reliable valuation data. Lenders need to know how much revenue the assets can generate and their resale value over time." The organization is filling the data gap in second-hand chip transactions.

Responsible Editor: Chen Yujia