๐ข ๐๐จ๐ฆ๐ง ๐๐ก: CoreWeave Brings Multi-Rack NVIDIA Vera Rubin NVL72 Clusters to Its AI Cloud - $Coreweave(CRWV.US) $NVIDIA(NVDA.US)
๐ ๐๐ฒ๐ ๐๐ถ๐ด๐ต๐น๐ถ๐ด๐ต๐๐:โค ๐๐ผ๐ฟ๐ฒ๐ช๐ฒ๐ฎ๐๐ฒ brings multi-rack ๐ก๐ฉ๐๐๐๐ ๐ฉ๐ฒ๐ฟ๐ฎ ๐ฅ๐๐ฏ๐ถ๐ป ๐ก๐ฉ๐๐ณ๐ฎ to its cloud.โค Clusters connect ๐ต๐๐ป๐ฑ๐ฟ๐ฒ๐ฑ๐ ๐ผ๐ณ ๐ฅ๐๐ฏ๐ถ๐ป ๐๐ฃ๐จ๐ for agentic AI workloads.โค Each NVL72 rack combines ๐ณ๐ฎ ๐ฅ๐๐ฏ๐ถ๐ป ๐๐ฃ๐จ๐ with ๐ฏ๐ฒ ๐ฉ๐ฒ๐ฟ๐ฎ ๐๐ฃ๐จ๐.โค ๐ก๐ฉ๐๐๐๐ ๐ฆ๐ฝ๐ฒ๐ฐ๐๐ฟ๐๐บ-๐ซ Ethernet connects multiple racks into scale-out clusters.โค Each Rubin GPU receives up to ๐ญ.๐ฒ ๐ง๐ฏ/๐ scale-out connectivity.โค ๐๐ผ๐ฟ๐ฒ๐ช๐ฒ๐ฎ๐๐ฒ ๐ ๐ถ๐๐๐ถ๐ผ๐ป ๐๐ผ๐ป๐๐ฟ๐ผ๐น automates rack setup, validation, power, and cooling.โค ๐๐ข๐ง๐ delivers local NVMe-speed reads with up to ๐ณ ๐๐/๐ per GPU.โค LOTA reduces latency by ๐ด๐ versus traditional storage-cluster reads.โค New ๐ฐ๐ฟ๐ผ๐๐-๐ฟ๐ฒ๐ด๐ถ๐ผ๐ป ๐๐ฟ๐ถ๐๐ฒ ๐ฎ๐ฐ๐ฐ๐ฒ๐น๐ฒ๐ฟ๐ฎ๐๐ถ๐ผ๐ป minimizes AI checkpoint delays.โค New ๐๐ฟ๐ฐ๐ต๐ถ๐๐ฒ tier targets lower-cost long-term AI data storage.๐ ๐ช๐ต๐ ๐ง๐ต๐ถ๐ ๐ ๐ฎ๐๐๐ฒ๐ฟ๐:โค Expands CoreWeave's capacity for large-scale ๐ฎ๐ด๐ฒ๐ป๐๐ถ๐ฐ ๐๐ training and inference.โค Faster storage access can reduce GPU idle time and improve utilization.โค Multi-rack Rubin enables larger models and ๐ฟ๐ฒ๐ถ๐ป๐ณ๐ผ๐ฟ๐ฐ๐ฒ๐บ๐ฒ๐ป๐ ๐น๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด at scale.โค Cross-region acceleration addresses a key bottleneck in distributed AI infrastructure.๐ ๐๐ ๐ฝ๐ฒ๐ฟ๐ ๐ฆ๐๐ฎ๐๐ฒ๐บ๐ฒ๐ป๐๐:๐๐ต๐ฒ๐ป ๐๐ผ๐น๐ฑ๐ฏ๐ฒ๐ฟ๐ด, Executive Vice President of Product & Engineering at CoreWeave:โCoreWeave was the first AI cloud provider to validate and bring up a Vera Rubin NVL72, demonstrating that this advanced rack-scale architecture could operate as a reliable, high-performance cloud service. With multi-rack Vera Rubin, we are connecting hundreds of Rubin GPUs as a single scale-out cluster. For customers building agentic AI, that means greater scale, faster iteration, and higher productivity as models and agents continuously learn and improve.โ๐๐ฒฬ๐ฐ๐ถ๐น๐ฒ ๐ฅ๐ผ๐ฏ๐ฒ๐ฟ๐-๐ ๐ถ๐ฐ๐ต๐ผ๐ป, Director of Internal Infrastructure at Cohere:โOur datasets span multiple regions, and we canโt afford to have our training schedule dictated by cross-region retrieval delays. CoreWeave AI Object Storage gives us a unified dataset footprint across regions with reads cached locally, so nothing waits on the network. Itโs the difference between planning around our data and simply training.โ





