Wall Street's AI conversation has become fixated on Nvidia Corp (NVDA)'s newest chips. CoreWeave, Inc. (CRWV), however, used its second quarter earnings call to make a different point: older GPUs may have far more earning power than investors think.
The cloud infrastructure provider revealed it recently signed a customer contract for Nvidia's A100 GPUs that extends through 2029—nearly a decade after the chip debuted.
More importantly, management suggested this isn't an exception but evidence that AI infrastructure can continue generating attractive returns long after its first deployment.
CoreWeave Sees Long-Term Value in Older Nvidia GPUs
The clearest indication came from CFO Nitin Navin, who highlighted the longevity of one of the company's latest deals. "We recently signed an A100 contract that extends into 2029 at an attractive price. As a reminder, this SKU was introduced in 2020," Navin said.
CEO Michael Intrator expanded on that point, arguing the contract offers a glimpse into how AI infrastructure could be monetized over a much longer period than many investors currently assume.
"The fact that we have been able to go ahead and sell a GPU whose architecture was from 2020 in a contract that was fully priced out to 2029 really provides some insight into what the future is going to look like," Intrator said.
Those comments challenge a common assumption in the AI infrastructure market—that each new generation of Nvidia chips quickly renders older hardware economically obsolete.
Instead, CoreWeave is signaling that mature GPUs can continue attracting customers if the workload and pricing remain attractive.
CoreWeave's GPU Strategy Goes Beyond the First Contract
The company also hinted that older hardware could generate revenue more than once.
Navin said every time an existing GPU is renewed or redeployed after its initial contract, the revenue comes on top of returns already earned during the original lease. "Every resale or renewal is incremental on top of the returns already earned within the initial term," he said.
Intrator added that managed inference—a business the company expects to surpass $250 million in annual recurring revenue by the end of 2026—offers another avenue for putting GPUs coming off contract back to work rather than leaving them idle.
For investors, the takeaway extends beyond a single A100 contract. If CoreWeave can consistently renew, redeploy or repurpose older GPUs into new customer agreements and inference workloads, the economic life of its infrastructure could prove much longer than many expect.
That would allow the company to generate additional returns from assets already on its balance sheet. A dynamic that could become increasingly important as the AI market moves beyond the race for the latest chips and toward maximizing the value of existing GPU fleets.
