Mostly True
The Nvidia A100, a data-center AI accelerator first released six years ago, is showing an unexpected reversal in its leasing market: rental prices for the aging GPU are rising again, driven by sustained demand for AI computing capacity.
The A100, launched in 2020, was long expected to fade from the market as Nvidia rolled out newer generations such as the H100. Instead, rental rates for the six-year-old chip have moved in the opposite direction of the usual depreciation curve, climbing back up rather than continuing to decline.
The reversal reflects tight overall supply of AI computing capacity, as demand for GPU rentals to train and run large language models (LLMs) continues to outstrip availability of newer hardware. With newer accelerators difficult to secure, capacity built around the older A100 has retained — and in some cases regained — pricing power.
The extent and duration of the price rebound remain unconfirmed in detail, and rental rates can vary significantly by cloud provider, region and contract terms. The finding is judged to be broadly accurate, though the precise scale of the increase could not be fully pinned down across all market segments.
The claim that the A100, now in its sixth year since release, is seeing rental rates move back up is assessed as Mostly True.
Verdict: Mostly True