Nvidia is working with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to assemble $500 billion in financing aimed at turning computing hardware into a new asset class. Nvidia CEO Jensen Huang told CNBC that this marks the first time technology chips have become investable assets, describing them as revenue-generating, long-lived, fungible, and flexible. He compared the moment to the early days of mortgage-backed securities in the 1970s. BlackRock CEO Larry Fink echoed that sentiment, calling it the next frontier for financial engineering.
The plan, however, is not yet finalized. Nvidia has signed only memorandums of understanding, which are non-binding agreements. The source notes that a similar $100 billion memorandum between Nvidia and OpenAI last year never materialized. Still, the announcement signals a shift in how the company wants its chips valued, particularly as it faces questions about their useful life.
Huang’s current language about chip longevity contradicts what he said last year. When promoting Nvidia’s Blackwell architecture, he said that once Blackwell shipped in volume, you could not give away the older Hopper chips, adding that only a few circumstances made Hopper adequate. Now, he is touting the A100 chip, introduced in 2020, as remaining in active commercial use with an economic life approaching a decade. This matters for lenders because depreciation schedules determine loan terms. CoreWeave, a pioneer in GPU-backed loans and an Nvidia-associated client, can borrow less as its chips depreciate, according to its corporate filings.
The financing structure resembles a deal Broadcom arranged earlier this summer with Apollo and Blackstone, involving about a million chips as collateral for $35 billion in funding. In that arrangement, the lenders earn interest while Broadcom guarantees senior notes issued by a special purpose vehicle holding the chips. The goal was to boost demand for Broadcom hardware, and Nvidia appears to be pursuing the same strategy.
Rental prices for older chips have been rising, contrary to expectations of decline. A cloud service provider nearly doubled the price of Nvidia Blackwell B200 chips for one customer during a contract renewal. Silicon Data projects these price increases will continue through 2028. An AI industry analyst cited a major shortage of inference chips as the cause, reversing the usual trend of falling prices.
Skeptics point to risks in this model. A former hedge fund manager noted that mortgage-backed securities failed when mortgages were overproduced, and the AI industry is becoming saturated with data centers. Chinese open-source models require less compute while remaining powerful, potentially threatening the assumption of ever-growing chip demand. There is also the open question of whether frontier labs like Anthropic and OpenAI, which drive much of the current demand, can actually turn a profit. Huang’s framing of “compute” also excludes data center real estate, which has already attracted significant financing, including a platform valued at $185 billion built by Blackstone.
