Rumor

Rumor: Major AI Customer Reportedly Planning a 100,000-Unit NVIDIA H100 GPU Cluster

Monday, December 8, 2025 at 06:16 PM

A major AI player, referred to as 'The Whale,' is rumored to be planning or building a supercomputer cluster consisting of 100,000 NVIDIA H100 GPUs.

Context

A significant rumor circulating suggests a major AI customer is planning a 100,000-unit NVIDIA H100 GPU cluster, potentially for Elon Musk's xAI venture. While the specific tweet remains unverified, the scale aligns with observed hyperscaler demand, as companies like Meta have disclosed plans for hundreds of thousands of H100 or equivalent devices. This development, if confirmed, would represent a substantial order for NVIDIA and further strain the AI supply chain. The estimated hardware-only cost for such a cluster is in the low single-digit billions, with forum discussions citing around $2 billion. However, infrastructure and operational expenses would add significantly to this figure. The project, sometimes referred to as the 'Colossus' or 'Memphis SuperCluster,' involves partnerships with companies like Supermicro and NVIDIA for a large liquid-cooled GPU deployment. Given the current high demand and supply constraints for high-end AI GPUs, a 100,000 H100 order would have considerable implications for NVIDIA's revenue and market position, as well as for the broader semiconductor and AI infrastructure sectors. While plausible, explicit confirmation from the buyer or vendor is needed before treating this as a definitive fact.

Related Companies

Nvidia
Nvidia
NVDA
US