CoreWeave Vera Rubin NVL72 Debuts on Cloud
CoreWeave Vera Rubin NVL72 availability and company-reported 4.8× throughput and 10× tokens-per-megawatt results may refocus trader positioning and flow.
KEY TAKEAWAYS
- Announced availability of NVIDIA Vera Rubin NVL72 on CoreWeave cloud with Cognition as first production customer.
- Reported Cognition measured 4.8× total token throughput and 10× token throughput per megawatt versus GB200 NVL72.
- Launched CoreWeave Forge, a Partner Network, and plans to offer NVIDIA Vera CPUs.
HIGH POTENTIAL TRADES SENT DIRECTLY TO YOUR INBOX
Add your email to receive our free daily newsletter. No spam, unsubscribe anytime.
CoreWeave said on Sept. 30, 2026, that its CoreWeave Vera Rubin NVL72 is available on its cloud, with Cognition as the first customer running production workloads. The company also launched CoreWeave Forge, a Partner Network, and announced plans to offer NVIDIA Vera CPUs.
Vera Rubin NVL72 Availability and Performance
CoreWeave announced the availability of NVIDIA Vera Rubin NVL72 on its cloud, identifying Cognition as the first customer running production workloads on the system. Cognition began deploying a Vera Rubin NVL72 cluster in early September and ran an inference benchmark against an NVIDIA GB200 NVL72 baseline.
Cognition measured a 4.8-times increase in total token throughput for SWE-2 inference workloads compared with the GB200 NVL72 baseline. CoreWeave also reported a 10-times gain in token throughput per megawatt on the DeepSeek R1 reasoning model at matched interactivity. The company described this as an industry-first measured silicon-performance result.
The announcement was made at CoreWeave’s Fully Connected conference, which the company said attracted more than 4,500 customers, partners, developers, and AI leaders.
Platform Expansion and Partnerships
CoreWeave launched CoreWeave Forge, a development environment that covers the AI loop—run, observe, curate, improve, evaluate, and repeat—in a connected platform. The company said Forge is open across models, frameworks, and other clouds. MasterClass and Canva were named as early users.
The company also introduced a curated Partner Network with integrations validated on CoreWeave Cloud under production conditions. Named collaborators include CrowdStrike, VAST Data, Reflection, and ClickHouse.
CoreWeave announced plans to offer NVIDIA Vera, described as the first CPU built for AI agents, expanding its compute portfolio beyond GPUs.
Together, these product announcements extend CoreWeave’s offerings beyond GPU rental into development tooling, inference, agent workloads, and partner integrations. The company also highlighted software and management tools—including CoreWeave Kubernetes Service, SUNK, CoreWeave Mission Control, CoreWeave Sandboxes, and serverless inference—that customers can use to deploy and manage workloads across GPU generations.
