Nvidia refutes 'Big Short' investor Michael Burry's bearish case. Here's what the company says

Nvidia pushed back on criticism that rapid generational gains in AI chips will render earlier hardware uneconomic, with CFO Colette Kress arguing on the earnings call that the company’s CUDA software extends the useful life and lowers total cost of ownership for its GPUs—citing that A100 units shipped six years ago remain at full utilization after software improvements. The defense aims to allay investor concerns that accelerating efficiency gains force customers into unprofitable, recurring capital cycles; analyst Ben Reitzes said this supports accurate depreciation schedules at large buyers. Michael Burry, who recently disclosed a sizeable bearish position in Nvidia and Palantir, reiterated the opposing view that much lower power consumption in newer GPUs makes older hardware uncompetitive and that firms may feel compelled to invest before those purchases are profitable, highlighting a fundamental tension with implications for corporate capex, asset lifecycles and Nvidia’s valuation.

Analysis

Nvidia used its earnings call to rebut concerns that rapid generational improvements in GPUs will render earlier hardware uneconomic, with CFO Colette Kress stating that the CUDA software stack extends useful life and Total Cost of Ownership; she cited that A100 GPUs shipped six years ago remain at full utilization after software improvements. The company argues CUDA compatibility across its installed base preserves economic life even as new generations deliver performance and efficiency gains, a point supported by analyst Ben Reitzes who said this helps validate depreciation schedules at large customers.

Michael Burry, who disclosed sizeable bearish positions in NVDA and PLTR, reiterated the opposing thesis that materially lower power consumption and efficiency of newer GPUs can make older hardware uncompetitive, meaning customers may feel compelled to invest in new AI hardware before those purchases are profitable. That presents a logical contradiction: Nvidia claims both meaningful superiority of new chips and extended value of old chips, so one defense may be overstated.

The dispute matters for capex pacing, revenue durability and valuation: if CUDA-driven life extension proves real, customer replacement cycles and depreciation risk are lower; if Burry is correct, accelerated churn could compress purchaser ROI and make AI capex cadence more cyclical. Key near-term indicators are customer utilization data, capex guidance and secondary-market pricing for legacy GPUs as leading signals of which thesis prevails.

Key Decisions for Investors

• Monitor corporate capex guidance and published utilization metrics for A100-era systems to validate Nvidia’s CUDA-driven life-extension claim, adjust NVDA exposure if evidence of rapid replacement cycles emerges
• Size or hedge positions in NVDA ahead of clear real-world TCO proof — use options or partial trims to protect against a downside scenario where newer GPUs force accelerated refreshes and compress customer ROI
• Track secondary-market pricing and resale velocity for older Nvidia GPUs and follow Burry’s disclosures (including his PLTR position) as sentiment/short-interest indicators that could amplify volatility