Tech companies are increasingly relying on a novel financing mechanism to fuel their massive investments in artificial intelligence: loans backed by the very chips that power these systems. This trend, gaining momentum throughout 2025 and into , allows companies to access capital while keeping debt off their corporate balance sheets, but also introduces new complexities and risks for lenders.
The loans are structured around special-purpose vehicles (SPVs) that acquire high-performing graphics processing units (GPUs) and then lease them to tech businesses for AI model training. This approach has become popular as the AI “arms race” intensifies, with companies burning through hundreds of billions of dollars annually on these essential components. Investors are attracted by yields, typically ranging from the high single digits to the mid-teens, which are often higher than those offered by the tech companies themselves.
“Investors are very excited,” said David Ridenour, a partner specializing in finance and restructuring at law firm King & Spalding. “People are willing to dive into [GPU deals] on a take-it-or-leave-it basis.”
The practice was pioneered by cloud computing provider CoreWeave in late , and has since gained traction as demand for advanced chips soars. Citigroup estimates that GPUs and associated servers can account for 30 to 40 percent of total project costs for data centers, highlighting the significant capital expenditure required for AI infrastructure.
Recent deals demonstrate the scale of this financing trend. In January , Apollo announced a $3.5 billion financing package for a digital infrastructure fund managed by Valor Equity Partners, specifically to purchase Nvidia’s GB200 “AI superchips” and lease them to xAI, Elon Musk’s artificial intelligence company. IREN Limited, an AI cloud service provider, secured a $3.6 billion loan commitment from Goldman Sachs and JPMorgan earlier in to acquire chips for its AI contracts with Microsoft.
The speed at which these transactions are executed is notable. According to a lawyer familiar with GPU financing, lenders are often asked to act quickly and commit substantial capital. “A big player would basically ask, ‘would you like to participate in a deal that closes in two weeks and throw in a couple hundred million?’” the lawyer said.
This rising popularity reflects a broader investor appetite for asset-backed finance, where lenders seek debt secured by stable cash flows. GPU-backed loans typically include “hell or high water” clauses, designed to prevent tech companies from terminating the leases prematurely, mitigating the risk of obsolescence as AI technology rapidly evolves.
However, the nascent nature of this market introduces unique challenges. Moody’s, which has begun rating GPU-backed debt, withdraws credit ratings once the underlying leases expire. Dorina Yessios, US co-head of energy, infrastructure and natural resources at A&O Shearman, emphasized the need to carefully consider GPU lifespan. “That has to be factored into underwriting, just like any other equipment financing.”
Some investors express concerns about the economic life of GPUs and the potential for their value to decline faster than anticipated. The lack of a robust secondary market and price history for older AI chips further complicates valuation. Current valuations may also be inflated by temporary chip supply shortages, according to some market participants.
“We really want to ensure the GPUs’ useful life well exceeds the amortized period of our investment,” said Jen Marques, head of strategy and structuring for Oaktree’s structured credit strategy. The concern isn’t merely about technological advancement; one investor who declined multiple GPU financing pitches bluntly stated, “Those things won’t make it three years before they are antiquated. It’s a huge gamble.”
The investor further cautioned that reselling older GPUs in the event of a default would be difficult, likening it to “beating a dead horse.” This highlights the inherent risk associated with financing assets in a rapidly evolving technological landscape.
AMD is also participating in this trend, preparing to backstop a $300 million loan to data center operator Crusoe, as reported on . The loan, provided by Goldman Sachs, will be secured by AMD’s AI chips and related equipment. As part of the agreement, AMD has agreed to rent the chips from Crusoe if the company cannot find sufficient third-party customers, demonstrating a willingness to support customer financing to increase adoption of its Instinct accelerators in a market dominated by Nvidia.
This new financing model underscores the immense capital requirements of the AI boom and the innovative ways companies are seeking to fund it. While offering attractive yields for investors, GPU-backed loans also present unique risks that require careful consideration and robust underwriting practices.
