CoreWeave puts NVIDIA Vera Rubin into production on its cloud
Cognition is already using the new setup in production, and CoreWeave is tying live feedback into how models and agents are trained and evaluated.
CoreWeave said it is now bringing NVIDIA’s latest AI infrastructure into production on CoreWeave Cloud, with NVIDIA Vera Rubin NVL72 systems and Spectrum-X 102.4T Ethernet available to customers. Cognition, the team behind Devin, is the first customer already running production workloads on Vera Rubin, and CoreWeave says it stood up that cluster within days. CoreWeave also plans to offer NVIDIA Vera, the CPU aimed at agentic workloads, and it launched CoreWeave Forge to connect production feedback back into training and evaluation. In testing, Cognition reported up to 4.8x higher token throughput for SWE-2 inference on Vera Rubin NVL72 versus a GB200 NVL72 baseline, while CoreWeave said its Vera CPU tests produced more than 3x faster agent sandbox startups and a 1.7x gain on Terminal-Bench.
Why it matters
This moves NVIDIA’s newest infrastructure from test and positioning into day-to-day use for agentic AI workloads. For teams running production systems, the change is less about a new product label and more about having a faster path from deployment back into training and evaluation, which can tighten the loop between how agents behave and how they are improved.
Keep or strike?
Does this story matter, or is it hype? Mark it before you see what everyone else did.
Sources
- NVIDIA Blog