Nvidia says DSX fits 40% more GPUs per watt; Vera Rubin to cut token cost ~45x

VishalG · x · 2026-09-16

Nvidia is pitching "every megawatt matters": its DSX platform lets AI factories fit up to 40% more GPUs within the same power envelope. Paired with the upcoming Vera Rubin architecture, Nvidia claims 30x higher throughput per megawatt and 45x lower cost per million tokens versus GB300.

Related event: NVIDIA DSX Platform Boosts AI Factory Efficiency, 40% More GPUs per Megawatt(4 posts)→

Original post →

More from Infra

Infra channel →