CoreWeave expands its next-generation AI infrastructure, connecting hundreds of Nvidia Rubin GPUs into a unified high-performance cluster.
CoreWeave ($CRWV) has brought a multi-rack Nvidia ($NVDA) Vera Rubin NVL72 cluster online, marking another major expansion of the AI cloud provider’s next-generation computing infrastructure.
The deployment connects hundreds of Rubin GPUs into a single scale-out cluster designed for demanding agentic AI workloads. Each Vera Rubin NVL72 rack combines 72 Rubin GPUs with 36 Vera CPUs, while Nvidia Spectrum-X Ethernet connects multiple racks into a coordinated computing system.
The architecture gives CoreWeave customers additional capacity to train larger AI models, handle increasingly complex inference workloads and perform reinforcement learning at scale. CoreWeave said the infrastructure is engineered across compute, networking, storage, cooling, power and software to operate as one system.
The company also introduced cross-region write acceleration for CoreWeave AI Object Storage, allowing workloads to write data locally while information is replicated to another region in the background. A new lower-cost Archive storage tier was also unveiled.
The deployment strengthens CoreWeave’s position in the rapidly expanding AI infrastructure market while further deepening its relationship with Nvidia as demand for advanced computing capacity continues to grow.
