Investor Michael Burry, known for his 2008 financial crisis forecast, compared the current market sentiment surrounding Nvidia to the computer leasing bubble of the late 1960s. The debate was triggered by a September 27 Nvidia presentation, which argued that its AI accelerators retain economic value longer than the standard five-year accelerated depreciation schedule.
The Nvidia presentation, titled 'NVIDIA AI Infrastructure Retains Value Beyond Accelerated Depreciation Schedules,' compared the residual value of A100, H100, and B200 GPUs against standard depreciation models. The company suggested that businesses can write off AI infrastructure investments faster than the equipment actually loses its utility, countering claims that GPUs become obsolete too quickly.
Burry countered this by citing Adam Smith's 1968 book 'The Money Game,' which details how investors in the 1960s justified high valuations for computer leasing firms by citing infinite demand and long equipment lifespans. Burry argues that market participants today are repeating the same behavioral patterns, prioritizing growth narratives over the reality of rapid technological obsolescence.
The controversy also involves the rental market for AI hardware. Nvidia CEO Jensen Huang recently described the company's infrastructure as a long-lived, highly rentable asset. Investor Jim Chanos questioned this logic, asking why Nvidia does not lease the hardware directly if the assets are so profitable, especially given the rising rental rates for three-year-old H100 units.
The core of the disagreement lies in how to value AI hardware. While Nvidia emphasizes the sustained economic utility of its chips, critics like Burry point out that newer generations of GPUs can rapidly erode the market value of older models. The debate centers on the actual duration for which current AI infrastructure will remain a revenue-generating asset for its owners.
