CoreWeave, an AI cloud company, has signed a contract to rent Nvidia A100 GPUs through 2029, extending their operational lifespan to nearly nine years after their 2020 launch. The agreement, disclosed by CoreWeave’s Chief Financial Officer Nitin Agrawal on Tuesday, provides fresh evidence in the ongoing debate over how long AI chips retain economic value.
CoreWeave’s stock surged nearly 20% on Wednesday following its earnings call and subsequent CNBC appearance, where CEO Mike Intrator stated that older generations of GPUs may have longer useful lives than anticipated. The company also reported revenue exceeding expectations, narrower-than-expected losses, and an upward revision to its annual revenue outlook.
The deal follows CoreWeave’s broader commitment to older Nvidia chips, with Agrawal noting the company is “largely sold out” of these generations. Rental-market data from Silicon Data supports these claims, showing that A100 rental rates have remained stable after a rebound in 2026. In a post on X, Silicon Data wrote: “It certainly doesn't appear to be 2-3 years as some seem to casually assume.”
Market reactions reflected the optimism. Nvidia shares rose 3%, while suppliers like Corning (5.2%), GE Vernova (2.7%), and Micron (4.9%) also saw gains. The development counters a key bear case on the AI trade—that AI hardware becomes obsolete within two to three years—by demonstrating sustained demand for older chips in non-cutting-edge AI workloads.
Context: The AI Infrastructure Trade
The longevity of AI chips has been a critical factor in investor confidence, particularly for companies like Amazon, Microsoft, and other data center builders weighing multi-billion-dollar capital expenditures. CoreWeave’s contract suggests that the return on investment for current AI infrastructure spending may be more sustainable than previously feared.
Nvidia’s own initiatives, including a $500 billion financing plan announced Monday in partnership with major Wall Street firms, have also gained indirect validation from CoreWeave’s comments. The plan aims to accelerate AI infrastructure deployment, and CoreWeave’s data implies that such investments could yield longer-than-expected returns.
Why Older GPUs Remain Viable
Experts note that while the newest chips are essential for training advanced AI models, older generations remain highly effective for inference tasks—such as running established AI applications—where raw speed is less critical. This division of labor may extend the economic lifespan of GPUs well beyond initial expectations.
CoreWeave’s financial performance further underscored the trend. The company’s revenue topped projections, losses narrowed more than anticipated, and its full-year revenue outlook was raised. Analyst Jim Cramer described the results as “a true breakout quarter” during the CNBC appearance.
Broader Implications
The contract and CoreWeave’s statements suggest that the AI infrastructure cycle may be more resilient than skeptics have argued. For chipmakers, power suppliers, and data center operators, this could mean steadier revenue streams and reduced pressure to frequently replace hardware. However, the debate is not fully resolved, as some investors continue to question whether rapid advancements in AI chip technology could still render older models obsolete sooner than expected.