NVIDIA VP Ian Buck stated that dynamically managing power consumption at the rack and firmware levels through software allows for the deployment of up to 40% more racks within the exact same power envelope, directly boosting throughput to 1.2 billion tokens per second. The NVIDIA Vera Rubin platform, through full-stack co-design, achieves a 30x increase in AI factory throughput for Agentic AI workloads, with some scenarios reaching up to 60x.