This 2020 Nvidia chip is still going strong. That matters for the AI boom. AI Boom Debate: How Long Can AI Chips Stay Useful? The ongoing debate about the longevity of AI chips has just received new evidence, courtesy of AI cloud company CoreWeave. The company has signed a contract to rent out Nvidia A100 GPUs until 2029, sparking a fresh discussion about the economic value of AI hardware. The Significance of CoreWeave’s Deal CoreWeave’s Chief Financial Officer, Nitin Agrawal, revealed that the company has committed to using Nvidia A100 GPUs for nearly nine years after their initial launch in 2020. This development is crucial, as investors have been concerned about the rapid depreciation of AI hardware. Critics have argued that older chips could become obsolete within two or three years, necessitating significant write-downs and negatively impacting profits. Challenging the Status Quo However, CoreWeave’s deal suggests that the opposite may be true. The company’s ability to secure an attractive price for the A100 contract extending into 2029 indicates that older chips can still generate significant revenue. Furthermore, CoreWeave is “largely sold out” of older generations of Nvidia chips, according to Agrawal. Rental-Market Data Supports CoreWeave’s Claims Data from Silicon Data, which tracks GPU prices, reveals that A100 rental rates have held up well after a strong rebound in 2026. This evidence supports the notion that the useful life of GPUs may be longer than initially feared. As Silicon Data noted, “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.” The Repurposing of Older GPUs There are valid reasons why older GPUs can remain useful. While the newest chips are essential for building advanced AI models, companies can repurpose older GPUs for less demanding tasks, allowing them to continue generating revenue. This concept is supported by industry experts, such as Erwan Menard, a senior vice president at AI infrastructure company Crusoe, and Matt Rowe, an executive at Lambda. Extending the Life of GPU Fleets GPUs inevitably break and need to be replaced, which can cut the useful life of a fleet of GPUs. However, warranty contracts, which typically last five years, can extend the life of the overall GPU fleet. As Rowe pointed out, these warranties can mitigate the impact of depreciation, allowing companies to replace faulty GPUs with new ones. Implications for AI Infrastructure Investments The issue of GPUs’ useful lives has become increasingly important as Wall Street invests heavily in AI infrastructure. Nvidia’s recent partnerships with prominent companies, aimed at mobilizing over $500 billion, depend in part on AI hardware retaining its value for years. The fact that an A100 is still attracting customers into 2029 is a positive indication that AI hardware can maintain its economic value over an extended period. Source: WorldNewsAPI | Read original
