Key Takeaways:
Wall Street is pricing Nvidia GPUs as assets that keep generating cash for years, and a new CoreWeave contract suggests the chips may hold their value far longer than bears assume.
Nvidia's A100, launched in 2020, is still drawing customer commitments into 2029 — roughly nine years after debut — as AI cloud provider CoreWeave signs contracts that challenge the assumption that GPU hardware loses its economic value within two or three years.
"We recently signed an A100 contract that extends into 2029 at an attractive price," CoreWeave Chief Financial Officer Nitin Agrawal told analysts Tuesday, adding that the company is "largely sold out" of older Nvidia chip generations.
The deal is fresh evidence in one of the biggest debates hanging over the AI boom: how long AI chips stay useful and keep generating revenue. Rental-market data backs up CoreWeave's comments. Silicon Data, which tracks GPU prices, said A100 rental rates have held up well after a strong rebound in 2026. "We are still learning when it comes to the question of economic lifespan of GPUs. It certainly doesn't appear to be 2-3 years as some seem to casually assume," Silicon Data wrote on X.
The stakes are enormous. Nvidia this week announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR aimed at mobilizing more than $500 billion for AI infrastructure. Those investments depend in part on AI hardware retaining its value for years — and an A100 still attracting customers into 2029 is a powerful test of that assumption.
The newest chips matter for building the most advanced AI models. Once those models are created, companies have many other computing jobs that don't require the latest hardware. Older GPUs can be repurposed for these less demanding tasks and continue generating revenue.
Erwan Menard, a senior vice president at AI infrastructure company Crusoe, said GPUs can move from one type of work to another as they age. Lambda executive Matt Rowe said their effective lives can stretch to seven or eight years.
GPUs inevitably break and have to be replaced, which cuts the useful life of a fleet. But observers worrying about depreciation often overlook warranty contracts, Rowe said. These warranties typically last five years, so if GPUs fail, they are replaced with new ones, extending the life of the overall fleet.
Critics and short sellers have argued that rapid advances from Nvidia could make older chips obsolete within two or three years. If that happened, companies spending billions on AI infrastructure might have to write down those investments much faster, hurting profits.
The A100 delivers 312 TFLOPS of FP16 compute, while Nvidia's newer H100 offers 990 TFLOPS — a roughly threefold jump that has fueled the obsolescence argument. Yet the rental market tells a different story, with A100 rates holding firm. CoreWeave shares have surged since the initial public offering and jumped 20 percent on Wednesday after the comments.
The question of GPU useful lives has grown more important as Wall Street pours money into AI infrastructure. Nvidia is also pushing the AI factory concept beyond the rack, partnering with financial institutions to treat compute as an infrastructure asset and rolling out an 800 VDC power roadmap to upgrade existing data centers without full redesigns.
For investors, the CoreWeave contract is an encouraging data point against the depreciation thesis. If older GPUs keep generating revenue for nearly a decade, the economics of AI infrastructure — and the valuations built on them — look more durable than the bears suggest. Nvidia shares, which have led the AI rally, stand to benefit most if the market continues pricing GPUs as long-lived cash-flow machines.
This article is for informational purposes only and does not constitute investment advice.
