Vertiv presents gigawatt-scale reference architectures for a new era of AI factories

Vertiv announced the launch of a new generation of gigawatt-scale reference architectures created for the NVIDIA Omniverse DSX Blueprint. These solutions are aimed at accelerating the launch of large-scale generative AI systems, including on platforms such as NVIDIA Vera Rubin. Against the backdrop of NVIDIA opening its own Research Center in Virginia and the growing demand for giant AI factories, Vertiv plays a key role in shaping the infrastructure of the new stage of AI development.

A new approach to building large-scale AI centers

Unlike a one-size-fits-all approach, Vertiv offers customers a choice of three deployment formats: traditional construction, hybrid format, or fully prefabricated modules. This allows organizations to adapt infrastructure to their own pace, budget, and scaling requirements. The company also integrates the SimReady library of 3D objects, which makes it possible to create digital twins and test future AI factories even before construction begins.

The OneCore platform – the key to the rapid launch of large AI infrastructures

The prefabricated architecture is based on the Vertiv OneCore platform. Scalable without limitations in size, it offers a complete, comprehensive data center infrastructure designed to meet the unprecedented demands arising from the use of artificial intelligence and new technologies. Its preassembled design simplifies transportation and installation, reducing labor needs and on-site risks. This ensures reliability, cost efficiency, and clarity of schedules for modern enterprises. It represents a hybrid, preassembled solution that transforms traditional deployment for data centers with a capacity of more than 5 MW.

Key innovations presented in Vertiv architectures:

● Optimized energy topologies: The “Grid-to-Chip” system provides high efficiency and reduces infrastructure footprint while maintaining stability and scalability for new generations of AI platforms.
● Advanced liquid cooling: Vertiv integrates high-performance liquid cooling systems designed for the extreme thermal loads of modern accelerators. The full “thermal chain” from the chip to the heat recovery system ensures optimal operating conditions even for future generations of AI computing.
● Full deployment freedom: Vertiv solutions are the first on the market to have validation for all three approaches to building data centers: stick-built, hybrid, and prefabricated. This allows Time to First Token to be reduced by 50% and gives companies the freedom to choose the optimal path for implementation.

Key advantages presented in Vertiv architectures:

● Secure integration: Provides a single point of contact for optimizing multi-vendor management and coordinating complex data center deployment to ensure turnkey efficiency.
● Scalable and modular design: It uses preinstalled components integrated into the steel building supplied by Vertiv, offering data center modules that are customizable and building sizes that are flexible without limitations.
● Ideal for colocation, remote spaces, and turnkey deployment: Offers a universal, ready-to-deploy infrastructure for any data center needs.
● Accelerated installation: Thanks to precisely designed components that undergo strict quality control, the on-site assembly process becomes accelerated and efficient.

Representatives of NVIDIA confirm that it is thanks to the partnership with Vertiv that the creation of rapidly deployable solutions with a powerful energy and cooling foundation necessary for gigawatt-class AI factories becomes possible. Vertiv is already working on several large AI factories where Vertiv OneCore is used to achieve gigawatt scale.

Alpha Grissin, which is the official distributor of Vertiv in Ukraine, specializes in comprehensive uninterruptible power systems, modern cooling solutions, modular server rooms, and integrated projects of any complexity. Thanks to many years of experience, in-house engineering, and constant support, Alpha Grissin helps companies create reliable, energy-efficient, and scalable infrastructures ready for the challenges of the new AI era.