CoreWeave (CRWV.US) AI cloud computing power upgraded again! First to launch multi-rack NVIDIA Corporation (NVDA.US) Vera Rubin NVL72 cluster.

CoreWeave (CRWV.US) announced that it has enabled a multi-rack NVIDIA Corporation (NVDA.US) Vera Rubin NVL72 cluster on CoreWeave Cloud, interconnecting hundreds of NVIDIA Corporation Rubin graphics processing units (GPUs) into a single system for agentic AI workloads. According to CoreWeave, this multi-rack architecture allows training and inference workloads to be scheduled and run across hundreds of Rubin GPUs, providing compute support for larger-scale models, higher-intensity inference, and large-scale reinforcement learning. A single NVIDIA Corporation Vera Rubin NVL72 rack integrates 72 Rubin GPUs and 36 Vera CPUs, paired with NVIDIA Corporation's networking and data processing technologies. CoreWeave said that multiple racks are interconnected via NVIDIA Corporation Spectrum-X Ethernet and operate together as a single scale-out cluster. CoreWeave's infrastructure can automate rack deployment, firmware upgrades, validation, and power and cooling, and conducts full-rack-level testing before racks are put into production. Chen Goldberg, Executive Vice President and Head of Product and Engineering at CoreWeave, said that to her knowledge, CoreWeave is the first AI cloud service provider to complete validation of Vera Rubin NVL72 and put it into operation. She said: "With multi-rack Vera Rubin, we are connecting hundreds of Rubin GPUs into a single scale-out cluster." She added that the system can provide customers developing agentic AI with greater scale and faster iteration speed. While expanding computing capacity, CoreWeave also launched cross-region write acceleration for its AI object storage platform. This feature allows AI workloads to write data in the region where they are running, while CoreWeave replicates the data to another region in the background. The company said that for customers running training workloads across multiple regions, this feature can reduce pause time. This AI cloud computing company also launched an Archive storage tier, providing a lower-cost storage option for customers who want to retain data for the long term. This storage tier does not charge data retrieval fees, early deletion fees, or fees for reading data from the Archive storage tier. Its local object transfer accelerator, or LOTA, can also cache data closer to AI workloads. CoreWeave said that compared with reading data from traditional storage clusters, this feature can reduce read latency by 8 times.