Key facts
- Wall Street is creating a financial ecosystem around Nvidia's AI chips, treating them as a new asset class.
- Banks are lending against GPUs, and insurers are protecting their value.
- Nvidia is acting as a guarantor for some deals.
- Companies like Amazon are placing GPUs into separate entities rather than owning them directly.
- AI cloud providers are hiring compute traders to manage GPU contracts and future prices.
- Nvidia has partnered with BlackRock, Apollo, and Goldman Sachs to raise over $500 billion for AI infrastructure.
Wall Street is actively constructing a new financial system centered on Nvidia's graphics processing units (GPUs), aiming to establish them as a distinct asset class akin to real estate or oil. This burgeoning ecosystem involves financial institutions offering loans backed by GPUs, insurers providing protection against their depreciation, and investors seeking ways to trade them. The immense demand for Nvidia's AI chips, driven by the AI boom, has outstripped the ability of many companies to purchase them outright. By enabling borrowing against the chips themselves, this new financial framework could unlock hundreds of billions of dollars for AI infrastructure development, aligning with Nvidia's goal of widespread compute availability.
This evolving financial playbook includes Nvidia acting as a guarantor for certain deals, insurers mitigating the risk of falling GPU values, and companies like Amazon structuring their GPU ownership through separate entities. The maturation of this market is also creating new roles, such as full-time compute traders employed by AI cloud providers to manage and speculate on future GPU prices, according to Ethan Vera, chief operating officer of Luxor Technology.
The success of this financial system hinges on key questions regarding the long-term value and resale potential of these chips, especially given Nvidia's rapid release of new generations that could devalue existing hardware. GPU financing gained traction in the early 2020s, with banks and private credit firms lending to AI cloud providers. A significant contract between CoreWeave and Microsoft in 2023 provided lenders with greater confidence in the repayment of GPU-backed loans. While customer contracts currently form the primary basis for lending, lenders are increasingly exploring the use of GPUs themselves as collateral, according to Bernie Margulies, CEO of American Compute. Nvidia has reportedly engaged BlackRock, Apollo, and Goldman Sachs to assist in raising over $500 billion for AI infrastructure financing.
Beyond financing, the next step involves creating a marketplace for locking in future GPU prices and enabling speculative trading. Companies like Silicon Data, Compute Desk, and Ornn are developing indexes to standardize chip rental prices, addressing the lack of a universal price for compute, which varies by chip model, location, and other factors. Regulators have expressed concerns about potential manipulation of these nascent GPU indexes. If these indexes gain acceptance, they could pave the way for a GPU futures market. Some forms of speculation are already occurring on prediction markets like Kalshi and Polymarket, where users are betting on future rental costs for Nvidia's B200 chips.
