Overview
- CoreWeave announced it completed the industry’s first bring‑up and system validation of NVIDIA’s Vera Rubin NVL72 after a Dell‑shipped rack passed diagnostics and a roughly 147‑hour test suite.
- The Vera Rubin NVL72 is a liquid‑cooled rack with 72 Rubin GPUs and 36 Vera CPUs linked by a 260 TB/s NVLink fabric and is claimed by NVIDIA to deliver up to 10× better inference per watt and much lower token costs versus prior systems.
- To run the rack in production, CoreWeave deployed purpose‑built rack‑scale controls including Valvey for software‑defined per‑rack liquid cooling, Racky for unified rack control, and a Kubernetes‑native Rack Lifecycle Controller to manage the NVL72 as a single cloud resource.
- CoreWeave’s technical milestone comes as the company reports rapid commercial momentum with trailing‑twelve‑month revenue of $6.23 billion, Q1 2026 revenue of $2.08 billion, near‑$100 billion of contract backlog, and FY2026 guidance of $12–$13 billion.
- Broader customer rollouts and revenue gains will depend on two practical limits to watch: availability of data centers that can support liquid cooling and the company’s continued capital deployment and facility buildout as Vera Rubin shipments scale in H2 2026.