NVIDIA has detailed its Vera Rubin NVL72, which features 72 GPUs, 36 CPUs, and significant scale-up bandwidth. The system is reportedly being deployed by OpenAI and will also be used in Japan's first national AI factory, with production and pricing details emerging, including a high estimated cost for the Rubin Ultra Rack. The specifics of production capacity and rollout timelines are less clear, with some outlets citing limitations and others discussing large-scale deployments.