CoreWeave has become the first AI cloud provider to operationalize Nvidia’s Vera Rubin NVL72 system, combining hundreds of next-generation GPUs into a production cluster designed for the most demanding AI workloads. Each NVL72 rack packs 72 Rubin GPUs and 36 Vera CPUs, delivering up to 3.6 exaFLOPS of inference capability with 260 TB/s bandwidth via NVLink 6. CoreWeave’s stock (CRWV) surged roughly 14%, with trading volume nearly double its three-month average. The company claims the Rubin system delivers 10x inference throughput per megawatt compared to the Blackwell generation, reducing costs to roughly one-tenth per million tokens. Full production availability of Rubin clusters is expected later in 2026.
Source: Read the original article
