Vertiv unveils gigawatt AI data centre blueprints with NVIDIA

Vertiv has introduced its gigawatt-scale reference architectures for the NVIDIA Omniverse DSX Blueprint, offering enhanced deployment flexibility and reducing the time to first token in generative AI facilities by up to 50%.

The newly released architectures are designed to support large-scale artificial intelligence deployment using NVIDIA's Vera Rubin platform and other environments, as requirements for rapid, reliable, and scalable AI infrastructure continue to grow worldwide.

Reference architectures

Vertiv's reference architectures for the NVIDIA Omniverse DSX Blueprint aim to move beyond traditional, uniform approaches by offering options for traditional stick-built, hybrid, and fully prefabricated deployment strategies. This flexibility intends to help customers address a wide variety of needs regarding speed, customisation, and scale in AI data centre development.

By leveraging a library of SimReady 3D assets, Vertiv also enables the integration of virtual and physical systems, allowing clients to conduct full-scale digital twin simulations of AI factory designs before any physical construction begins. This approach aims to accelerate design cycles and enhance the accuracy of facility planning.

OneCore platform

The prefabricated variant of these gigawatt-scale reference architectures is built on Vertiv's OneCore platform. The platform is developed as a co-designed system integrating compute, power, cooling, and related services holistically across an entire facility. According to Vertiv, this approach can compress delivery schedules by up to 50% compared to conventional construction methods, offering space and performance optimisations suitable for the demands of advanced AI deployments.

Scott Armul, Executive Vice President, Global Portfolio and Business Units at Vertiv, said:

"The AI industrial revolution requires a complete reinvention of data centre infrastructure. Vertiv's collaboration with NVIDIA is foundational to our mission of equipping our customers with the most advanced and flexible solutions to navigate this transformation."

"The Vertiv reference architecture for NVIDIA Omniverse DSX is more than just a blueprint; it's a holistic approach that accelerates the deployment of AI at a scale previously unimaginable. Through our comprehensive grid-to-chip solutions and flexible deployment models like the Vertiv OneCore platform, combined with the power of digital twins within the NVIDIA Omniverse libraries, we are enabling the factories of the future, today."

The Vertiv reference architectures incorporate several technology advancements, including optimised power topologies and liquid cooling systems.

Technological features

The power infrastructure is based on Vertiv's "Grid-to-Chip" power train, which the company states improves upon the DSX Blueprint by providing a more efficient and compact system. This configuration is designed to deliver the resilience and scalability sought by large-scale, multi-generational AI platforms.

For thermal management, the architecture includes liquid cooling systems that address the high heat output of accelerated computing components. The solution is structured as a comprehensive chip-to-heat reuse thermal chain, designed to provide the necessary cooling performance for current and future compute platforms deployed at scale.

Vertiv is also positioning this reference architecture as the first to offer a validated design covering traditional, hybrid, and prefabricated construction methodologies, targeting reductions in the time required to operationalise AI infrastructure and enabling customers to pursue the most appropriate approach for their projects.

Industry perspectives

Dion Harris, Senior Director of HPC, Cloud, and Infrastructure Solutions at NVIDIA, said:

"Efficient, high-performance AI infrastructure is essential to the next industrial revolution - and building it requires a robust ecosystem of partners. NVIDIA and Vertiv are collaborating to develop flexible, rapidly-deployable reference architectures that deliver the critical power and cooling foundation needed to scale the next generation of gigawatt-class AI factories."

According to Vertiv, this suite of architectures is backed by a comprehensive portfolio of power and cooling technologies. It is supported by a global service organisation comprising more than 4,000 field engineers.

This support structure aims to facilitate the accelerated and resilient deployment and operation of large-scale, energy-efficient AI facilities. The company has stated that it is already engaged in several design projects for AI factories using the OneCore platform, which is developed to meet increasing gigawatt-scale demands.

Vertiv's announcement comes as organisations across various industries seek to rapidly scale their artificial intelligence capabilities and require infrastructure solutions that enhance deployment speed, energy efficiency, and operational reliability.