Vertiv Releases Gigawatt-Scale Architectures for NVIDIA Omniverse DSX

Vertiv announced in a press release the launch of its gigawatt-scale reference architectures for the NVIDIA Omniverse DSX Blueprint. The designs aim to reduce Time to First Token by up to 50% for large-scale generative AI systems and support deployment on platforms such as NVIDIA Vera Rubin.

The new architectures use the Vertiv OneCore platform, which integrates compute, power, cooling, and services into a unified system. This prefabricated approach allows deployment flexibility across stick-built, hybrid, and fully prefabricated models, enabling faster implementation of AI factory infrastructure.

Vertiv’s reference designs feature optimized power topologies and advanced liquid cooling systems to manage high thermal loads from accelerated computing. The integration of digital twin simulations through SimReady 3D assets allows customers to model and validate AI factory designs before construction.

According to Vertiv, these architectures are already being applied in the design of several large-scale AI factory projects, providing scalable, energy-efficient infrastructure for multi-generational AI platforms.