CoreWeave became the first company globally to power up Nvidia’s Vera Rubin NVL72 rack-scale system, delivering the most powerful commercially deployed AI inference machine ever built. Each rack features 72 Rubin GPUs and 36 Vera CPUs, capable of 3.6 exaFLOPS of inference power, connected by NVLink 6 fabric at 260 TB/s. The platform delivers up to 10 times the token throughput per megawatt compared to the previous Blackwell generation, particularly on demanding workloads like DeepSeek R1. Dell Technologies completed the integration in under 6.5 hours, reaching production readiness on June 1, 2026. CoreWeave’s stock surged approximately 14% following the announcement, and Nvidia expects full production of the Vera Rubin platform by July 2026.
Source: Read the original article
