Flexibility at gigawatt-scale | Reference architectures for the NVIDIA DSX Blueprint

Vertiv and NVIDIA accelerate time-to-first-token for generative AI

Hear from Vertiv executives about the impact of the gigawatt reference architecture

Engineered for gigawatt-scale AI architecture, enabled by Vertiv™ OneCore

Scalability on platforms including NVIDIA Vera Rubin

Full-scale digital twin of AI factory designs before construction begins

Optionality in deployment methods—from traditional stick-built and hybrid models to fully prefabricated solutions

SimReady 3D assets, available in NVIDIA Omniverse

Related resources

Key points in this article:

Vertiv™ OneCore reference architecture for the NVIDIA DSX Blueprint accelerates Time to First Token by up to 50%.
The entire facility is a single co-designed system — compute, power, cooling, and services holistically integrated from grid to chip.
High-density liquid cooling delivers capacities over 140 kW per rack at maximum resilience, scalable to NVIDIA Vera Rubin platforms.
Full-scale digital twin of AI factory designs via NVIDIA Omniverse enables simulation and optimization before construction begins.
Deployment optionality spans traditional stick-built, hybrid, and fully prefabricated models — reducing risk and accelerating deployment at gigawatt scale.

Spotlight: Flexibility at gigawatt-scale | Reference architectures for the NVIDIA DSX Blueprint