CoreWeave's continued use of NVIDIA A100 GPUs underscores how much life remains in the company's older AI infrastructure. The accelerators are part of NVIDIA's Ampere generation, which launched in 2020, and the arrangement shows that legacy hardware can still support long-term workloads.
Jensen Huang said the A100 fleet is mission-capable from 2020 through 2029, stressing that NVIDIA computing is about more than chips. In the same post, he said CUDA gives developers and NVIDIA engineers a common platform to continually upgrade Ampere, Hopper and Blackwell throughout their useful lives.
The mighty A100 fleet is mission-capable from 2020 through 2029. NVIDIA computing is more than chips. CUDA gives developers and NVIDIA engineers a common platform to continually upgrade Ampere, Hopper and Blackwell throughout their useful lives.
The sources say CUDA upgrades make older compute platforms a viable "productive asset" that remains rentable, durable and financeable. CoreWeave also reportedly got an attractive price for the NVIDIA A100 chips, which helps explain why the older accelerators remain part of its rental lineup.
That timing matters because the coverage says AI chip shortages and price hikes have affected almost every segment of the tech world. Even with Blackwell and Rubin ahead, the A100 is still being treated as hardware with real value rather than something ready to be written off.
How long will older AI accelerators keep their place if newer chips keep arriving?
-
NVIDIA denies Rubin and Vera memory cut rumors amid DRAM shortage -
NVIDIA CMP 170HX Mining Cards Are Being Turned Into AI Hardware -
NVIDIA Adds Windows 11 ARM64 Support to RTX Neural Texture Compression SDK -
NVIDIA Raises RTX PRO 6000 Blackwell Price to $16,000, Nearing 87% Above Launch -
NVIDIA unveils Nemotron VoiceChat, an open voice assistant for developers