News

Amazon Web Services maintains full fleet of Nvidia A100 GPUs amid high demand

Wednesday, February 4, 2026 at 08:48 AM

Amazon Web Services CEO Matt Garman stated that due to persistent high demand for AI infrastructure, the company has not yet retired any Nvidia A100 GPUs from its data centers.

Context

AWS CEO Matt Garman recently confirmed that the cloud giant has yet to retire a single Nvidia A100 GPU from its fleet, citing "insatiable" demand for AI compute. Despite the rollout of newer H100 and Blackwell chips, the A100 remains fully utilized for inference and mid-tier training. This highlights an unprecedented extension of the hardware's lifecycle, as customers prioritize immediate capacity over the latest raw performance. This sustained utility supports Amazon's aggressive infrastructure strategy, which includes doubling its total data center capacity to 8 GW by 2027. While AWS is scaling its proprietary Trainium3 silicon to offer more cost-effective options, the continued reliance on Nvidia’s legacy and cutting-edge chips underscores a persistent supply-demand imbalance. For investors, this signals high-margin revenue longevity for older GPU architectures, which are currently generating rental income far beyond their typical depreciation cycles.

Related Companies

Nvidia
Nvidia
NVDA
US
Amazon
Amazon
AMZN
US