CoreWeave (CRWV) Stock Surges 3% on Nvidia’s Vera Rubin Multi-Rack Deployment Milestone

Key Highlights

• Shares of CoreWeave advanced approximately 3% during Wednesday’s premarket session following the company’s announcement of a multi-rack Nvidia Vera Rubin NVL72 cluster deployment.
• The infrastructure links hundreds of Rubin GPUs into one unified scale-out system engineered for agentic AI applications.
• The company asserts it’s the inaugural AI cloud service provider to successfully validate and operationalize a Vera Rubin NVL72 at this scale.
• CoreWeave simultaneously unveiled two enhanced AI Object Storage capabilities: accelerated cross-region writing and an Archive storage tier.
• Shares of Nvidia (NVDA) increased roughly 1% following the announcement.

CoreWeave (CRWV) shares advanced approximately 3% in Wednesday’s premarket session after the infrastructure provider revealed it had successfully deployed a multi-rack Nvidia Vera Rubin NVL72 cluster within CoreWeave Cloud.

CoreWeave, Inc. Class A Common Stock, CRWV

This development builds upon CoreWeave’s previous achievement from June, when it first brought a standalone Vera Rubin NVL72 rack into production. Wednesday’s disclosure represents a substantial expansion beyond that initial deployment.

The newly implemented multi-rack configuration interconnects hundreds of Rubin GPUs to form a unified scale-out cluster. This represents a significant advancement in computational capacity for organizations operating extensive AI applications.

Each individual Vera Rubin NVL72 rack incorporates 72 Rubin GPUs paired with 36 Vera CPUs, alongside Nvidia NVLink 6 technology, ConnectX-9 SuperNICs, and BlueField-4 DPUs. The multi-rack configuration joins several such racks through Nvidia’s Spectrum-X Ethernet networking infrastructure.

The ConnectX-9 SuperNICs provide 1.6 Tb/s scale-out connectivity for each GPU through multiplane, multirail pathways. The architecture can accommodate approximately 128,000 GPUs per rail within a non-blocking fabric structure.

The modular architecture enables additional racks to be integrated without requiring fabric redesign for each expansion. This adaptability proves valuable when organizations need rapid scaling capabilities.

Chen Goldberg, EVP of product and engineering at CoreWeave, stated the multi-rack Vera Rubin cluster provides organizations developing agentic AI with “greater scale, faster iteration, and higher productivity as models and agents continuously learn and improve.”

The cluster supports both training and inference operations at substantial scale. Agentic AI applications, which demand continuous model learning and autonomous action, impose significant requirements on computational infrastructure.

Enhanced Storage Capabilities Unveiled

Along with the cluster announcement, CoreWeave rolled out two additional capabilities for its AI Object Storage platform.

The first feature enables cross-region write acceleration, allowing data to be written with local-level latency while background replication occurs to remote regions. The second introduces an Archive tier offering reduced-cost storage without retrieval charges, early deletion penalties, or reading fees.

CoreWeave’s Local Object Transport Accelerator (LOTA) enables read operations at speeds matching local NVMe performance. The provider claims it cuts latency by as much as 8x versus reading from conventional storage clusters.

Storage System Specifications

According to CoreWeave, LOTA delivers throughput reaching 7 GB/s per GPU. This performance level is achieved through managed caching implemented on each CoreWeave Kubernetes Service node.

The Archive tier serves as an economical solution for datasets requiring infrequent access while remaining integrated within CoreWeave’s infrastructure environment.

Nvidia shares rose approximately 1% Wednesday coinciding with CoreWeave’s announcement. The partnership between these companies attracts significant industry attention, with Nvidia hardware forming the foundation of CoreWeave’s cloud platform.

CoreWeave emphasized it remains the sole AI cloud provider to have successfully validated and deployed the Vera Rubin NVL72 architecture at multi-rack scale.