CoreWeave CEO Mike Intrator said older generations of NVIDIA GPUs will remain useful for up to a decade, exceeding previous market expectations.
This shift suggests that the rapid pace of AI hardware iteration may not render older chips obsolete as quickly as feared. For cloud providers and investors, extended hardware lifespans could reduce the frequency of costly capital expenditures while maintaining steady revenue streams.
Speaking Wednesday on CNBC’s “Squawk on the Street” during the company's Q2 earnings call, Intrator said that strong demand for AI infrastructure is driving this trend. He said that extended contracts for older GPUs indicate these assets remain valuable longer than anticipated [1, 2].
The market responded positively to the company's financial outlook. CoreWeave shares jumped nearly 16% in after-hours trading following the earnings release [1]. This surge comes as the company manages high demand for its computing capacity.
To address this demand, CoreWeave raised prices across its product lineup by roughly 25% in July [2]. The price hike reflects a market where near-term computing capacity remains sold out [2].
Despite the growth and hardware longevity, the company faces scrutiny in the credit markets. Credit default swap markets recently priced a 50% five-year default probability for CoreWeave’s debt [3]. This suggests a divide between the company's operational growth and the bond market's view of its long-term debt sustainability.
Intrator's comments highlight a critical tension in the AI sector. While NVIDIA continues to release more powerful chips, the foundational utility of older GPUs allows providers to scale services without immediate, total hardware replacement.
“Older generations of NVIDIA GPUs will have a longer useful life than anticipated”
The longevity of older GPUs suggests a stabilizing phase in AI infrastructure, where the 'compute crunch' is mitigated not just by new chips, but by the continued viability of existing hardware. However, the contrast between rising stock prices and a 50% credit default swap probability indicates that while the business model is generating revenue, lenders remain wary of the massive debt loads used to fund GPU acquisitions.
