This debate around how many years AI chips are actually useful, adding some fuel to those bubble fears. Nvidia CFO weighed in on that last night, saying that chips they shipped six years ago are still fully utilized today. It's a good topic for our dear DBosa in today's tech check.Hey D. >> Hey, good morning Carl. That's right.CFO Colette Crest, she tried to put the depreciation debate to bed, but she may have actually raised even bigger questions. She essentially said that Nvidia chips, they last longer th ...

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]