Hugging Face and NVIDIA launch Training Cluster as a Service with DGX Cloud Lepton

AI Impact Summary

Hugging Face and NVIDIA announce Training Cluster as a Service to provision large GPU clusters (DGX Cloud Lepton) in regional data centers, using NVIDIA Hopper and GB200 GPUs with centralized scheduling and monitoring. The on-demand model lets 250,000 Hugging Face organizations request cluster size and duration, shifting training capacity from fixed on-prem or vendor-specific pools to a pay-per-use cloud pattern. For engineers, this creates new integration points with Hugging Face libraries and workflow tooling, and a need to manage costs, regional feasibility, and data residency within the DGX Cloud Lepton network.