Vertiv Strengthens AI Infrastructure Leadership with New Gigawatt-Scale Reference Architectures for the NVIDIA DSX Blueprint, Demonstrating Deployment Flexibility with the Vertiv™ OneCore Prebuilt Platform and Reducing Time to First Token by Up to 50%
Vertiv (NYSE: VRT), a global leader in critical digital infrastructure, announces the release of gigawatt-scale reference architectures for Blueprint NVIDIA Omniverse DSX, designed to reduce the Time to First Token for large-scale generative AI, on platforms like NVIDIA Vera Rubin. As NVIDIA introduces its AI Factory research center in Virginia and the NVIDIA Omniverse DSX Blueprint for multigenerational, gigawatt-scale AI infrastructures, Vertiv plays a key role in enabling the next generation of AI innovations.
Going beyond the standardized approach, Vertiv reference architectures for the Blueprint NVIDIA Omniverse DSX offer a wide variety of deployment modes — from traditional models stick-built and hybrids, up to fully prefabricated solutions — to meet different customer needs for speed, flexibility and scalability. To further accelerate the design process, Vertiv enables the integration of virtual and physical systems thanks to its 3D asset library SimReady, allowing simulations digital twin large-scale AI factory projects before construction begins.
The prefabricated version of the architecture is based on the pre-validated Vertiv™ OneCore platform, which considers the entire structure as a single co-designed system, in which computing, power, cooling and services are holistically integrated. The Vertiv OneCore platform offers key benefits for gigawatt-scale deployments, including reducing lead times by up to 50% compared to traditional construction methods, while optimizing space efficiency and performance.
"The AI industrial revolution requires a complete reinvention of data center infrastructures. Vertiv's collaboration with NVIDIA is at the core of our mission to provide customers with advanced and flexible solutions to address this transformation," said Scott Armul, executive vice president of global portfolio and business units at Vertiv. “The Vertiv reference architecture for NVIDIA Omniverse DSX is more than just a blueprint – it is a holistic approach that accelerates the implementation of AI on a scale previously unthinkable. Through our comprehensive solutions grid-to-chip and flexible deployment models like the Vertiv™ OneCore platform, combined with the power of digital twin in the NVIDIA Omniverse libraries, we are making the factories of the future possible today.”
Vertiv reference architectures include several key innovations:
- Optimized power topologies: Vertiv's “Grid-to-Chip” power system further enhances the NVIDIA Omniverse DSX Blueprint, resulting in more efficiency and a smaller footprint. This approach ensures resilience and scalability for multi-generational AI platforms.
- Advanced Liquid Cooling Solutions: the architecture integrates Vertiv's industry-leading liquid cooling systems, designed to handle the extreme thermal demands of accelerated computing. This advanced thermal solution, combined with a complete heat reuse chain from chip to source, enables a perfectly tuned design for high performance in current and future computing generations.
- Deployment Flexibility: Recognizing that each AI deployment is unique, the Vertiv reference architecture is the first to offer a validated design for methodologies stick-built, hybrid and prefabricated. This flexibility can reduce the Time to First Token up to 50% compared to traditional constructions, allowing organizations to choose the best path for their objectives.
“Efficient, high-performance AI infrastructure is essential to the new industrial revolution, and building it requires an ecosystem of trusted partners,” said Dion Harris, senior director of HPC, Cloud, and Infrastructure Solutions at NVIDIA. “NVIDIA and Vertiv are collaborating to develop flexible, rapidly deployable reference architectures that provide the critical power and cooling foundations needed to scale the next generation of gigawatt-scale AI factories.”
This comprehensive suite of architectures is supported by the industry's broadest power and cooling portfolio and a global service organization of more than 4,000 field engineers. The suite provides the foundation to quickly deploy and manage energy-efficient, highly available, and scalable AI factories. Vertiv is currently in the design phase of several large AI factory projects that utilize the prefabricated Vertiv™ OneCore design, sized to meet gigawatt-scale needs.
