CoreWeave (CRWV) said Wednesday it has brought multi-rack Nvidia (NVDA) Vera Rubin NVL72 systems to its cloud, allowing hundreds of Nvidia Rubin GPUs to operate as a single cluster for AI training and inference.
The company said the systems can support larger models, demanding inference workloads and reinforcement learning.
CoreWeave uses Nvidia Spectrum-X Ethernet networking to connect multiple racks and operate them as a single system, it added.