Six-year-old A100 chips from Nvidia Corp. (NASDAQ:NVDA) are still valued at nearly $5,000 each, complicating Michael Burry’s warning that Big Tech may be overstating profits by depreciating AI hardware over five or six years.
Silicon Data estimates the A100’s residual value at $4,956. Its value has remained roughly flat near $5,000 since late 2025 as rising rental income offset the decline that would normally come with aging.
The figure isn’t a second-hand sale price. Silicon Data estimates what a GPU is worth from the future rental income it can generate, factoring in utilization and operating costs.
Why Old Nvidia Chips Still Pay
Older Nvidia GPUs do not necessarily become useless when a newer generation arrives. Chips such as the A100 can shift from cutting-edge training into inference, fine-tuning and lower-cost workloads.
That can extend their economic life. CoreWeave Inc. (NASDAQ:CRWV) has signed a customer contract for A100 GPUs extending through 2029, nearly a decade after the architecture debuted.
That matters to Burry’s argument because his warning centers on whether Big Tech is assuming AI hardware will remain economically useful for too long.
Burry estimated last year that hyperscalers, or the biggest cloud providers, could understate depreciation by roughly $176 billion between 2026 and 2028 by stretching the useful lives of computing equipment.
Longer useful lives mean companies record less depreciation expense each year.
Meta (NASDAQ:META) said in early 2025 that extending the useful lives of certain servers and network assets to 5.5 years would cut that year’s depreciation expense by roughly $2.9 billion.
Burry has argued that continued demand for older GPUs does not prove those schedules are correct.
“A chip can rent and still depreciate very fast economically,” he wrote in July.
Why It Matters
How long Nvidia’s chips keep earning money is becoming increasingly important as Wall Street pours more capital into AI infrastructure.
Nvidia has partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on platforms aimed at mobilizing more than $500 billion for AI infrastructure. CEO Jensen Huang has said lenders would look at factors including utilization, cash flow and residual value.
Kalshi traders are also betting that A100 rental prices will hold up.
Kalshi’s year-end market currently gives A100 compute about a 74% chance of finishing above $1.39 per GPU-hour compared with a current rental price of roughly $1.29 on Silicon Data’s A100 index.
Kalshi Research said in July that its A100 forward curve remained broadly flat into 2027, suggesting traders expected rental prices for the 6-year-old chip to remain relatively stable.
Rental prices for Nvidia’s 4-year-old H100 chips have also risen sharply recently.
Image: Shutterstock
Kalshi and Benzinga have an existing data collaboration agreement.
Login to comment