Nvidia reignites chip depreciation debate

Core Insights - The debate surrounding the economic lifespan of AI chips, particularly Nvidia's, has intensified, with Nvidia's CFO asserting that chips shipped six years ago are still fully utilized today [1][2] - However, there are concerns about the distinction between physical utilization and economic value creation, as older chips may not generate significant profits despite being operational [3][4] Nvidia's Chip Longevity - Nvidia's CFO Colette Crest emphasized that the company's chips have a longer useful life than commonly perceived, lasting up to six years with full utilization due to CUDA software [2] - Jensen Huang, Nvidia's CEO, acknowledged that while older GPUs can still function, their resale value diminishes significantly with the introduction of newer models, indicating a potential drop in economic value [4][5] Market Dynamics and Investor Concerns - The current tight supply of older chips allows for their continued use, but a shift in supply-demand dynamics could lead to a rapid decline in their value [5] - Investors face uncertainty regarding the economic viability of these chips, complicating the modeling of capital expenditures and profit margins in the AI sector [5] Competitive Landscape - Google is positioning itself as a competitor by developing its own Tensor Processing Units (TPUs) as a long-term alternative to Nvidia's GPUs, recently launching the Nano Banana Pro powered by Gemini 3 [6][7] - The tech trade is experiencing a shift, with Google shares rising nearly 1% amid broader market softness, reflecting investor interest in vertical integration strategies [7] Global Competition and Cost Efficiency - There are significant disparities in spending on GPU infrastructure between American and Chinese companies, with a reported 9:1 ratio, raising questions about how Chinese firms can produce comparable open-source models at a lower cost [9][10] - This cost efficiency and innovation from China contribute to ongoing concerns about the sustainability of the current market dynamics and the potential for a bubble in the AI chip sector [10]