CoreWeave has completed the industry's first validation of the Nvidia Vera Rubin NVL72 platform within its cloud environment. This system is designed to support agentic AI workflows and offers a reduction in cost per million tokens to one-tenth of previous Nvidia releases.
The company reported a revenue backlog of approximately $104 billion as of June 30. This figure represents contracted revenue that has not yet been recognized, reflecting the firm's growth in infrastructure and services to meet rising demand for AI inference workloads.
CoreWeave is focusing on specialized infrastructure to address the gap between AI experimentation and production. The company has redesigned its facilities and software control plane, with Vera Rubin racks supporting power densities of up to 250 kilowatts per rack.
The firm is hosting the Fully Connected event in San Francisco from September 30 to October 1. The conference will feature discussions on the role of neoclouds in providing scalable deployment, observability, and security for AI applications throughout their lifecycle.
