CoreWeave Connects Hundreds of Rubin GPUs in One Multi-Rack Cluster
CoreWeave launched the first multi-rack NVIDIA Vera Rubin NVL72 cluster among AI clouds, linking hundreds of Rubin GPUs into one system for agentic AI. It also added cross-region write acceleration and an archive tier to CoreWeave AI Object Storage.
- One NVL72 rack: 72 Rubin GPUs and 36 Vera CPUs with NVLink 6
- Each GPU delivers 1.6 Tbit/s via two ConnectX-9 SuperNICs
- LOTA speeds reads 8x and provides up to 7 GB/s per GPU
- Archive tier has no read, delete, or early deletion fees
Read next
Hardware