CoreWeave Deploys NVIDIA Vera Rubin NVL72 at Scale, Cognition Sees…

Event summary

CoreWeave launched NVIDIA Vera Rubin NVL72 availability, with Cognition as the first production customer.
Cognition measured a 4.8x increase in token throughput for SWE-2 inference workloads on Vera Rubin NVL72 vs. GB200 NVL72.
CoreWeave's platform supports seamless deployment across NVIDIA GPU generations, including Volta to Vera Rubin.
The deployment was announced during CoreWeave's Fully Connected AI cloud conference with 4,500+ attendees.

The big picture

CoreWeave's rapid deployment of NVIDIA's latest GPU architecture underscores its position as a critical infrastructure provider for AI development. The company's ability to support seamless transitions across GPU generations gives it a strategic advantage in serving AI innovators like Cognition, who require both cutting-edge performance and operational consistency. This deployment comes as AI workloads increasingly demand higher concurrency and longer contexts, positioning CoreWeave to capture growing market share in specialized AI cloud services.

What we're watching

Performance Scaling: How CoreWeave will maintain its performance lead as AI workloads grow in complexity.
Customer Adoption: Whether other AI labs will rapidly adopt Vera Rubin NVL72 following Cognition's benchmark results.
Competitive Dynamics: The pace at which competitors can match CoreWeave's multi-generation GPU deployment capabilities.