Dassault Systèmes and Vertiv design AI factories

Dassault Systèmes and Vertiv Holdings Co. have entered into a partnership concerning the configuration, simulation and verification of AI factory infrastructure before construction begins. The companies will combine the Dassault Systèmes 3DEXPERIENCE agentic platform and virtual twin technology with Vertiv's expertise in critical digital infrastructure. The joint solution is intended to reduce the complexity of on-site work and shorten the time from the start of design to the achievement of productive capacity by AI factories using NVIDIA DSX technology.

The partnership uses the capabilities of the configurable Vertiv SmartRun virtual twin while extending the model-based approach to other Vertiv infrastructure solutions for AI factories, including Vertiv OneCore. The 3DEXPERIENCE platform provides tools for systems engineering, simulation and virtual twins. These are intended to enable the configuration and validation of Vertiv's pre-engineered SimReady power and cooling reference designs for AI factories, based on the NVIDIA Vera Rubin DSX AI Factory reference design and NVIDIA Omniverse DSX Blueprint.

Integrated infrastructure design

AI factories are tightly coupled systems in which power, cooling, networking, control and computing capacity must be designed together. A change in one of these areas can have a significant impact on the others. Engineers must therefore understand and verify how the complete infrastructure system will operate before it is built and deployed.

Vertiv OneCore is a hybrid data centre solution that integrates critical power, thermal management and supporting infrastructure. The model-based systems engineering, simulation and virtual twin functions of the 3DEXPERIENCE platform make it possible to represent Vertiv infrastructure as an integrated digital system. This is intended to provide traceability and insight into the effects of design changes.

Through the 3DEXPERIENCE platform, Vertiv SimReady power and cooling assets become configurable system models. They provide a starting point for developing customer-specific AI factory configurations based on the NVIDIA Vera Rubin DSX AI Factory reference design and NVIDIA Omniverse DSX Blueprint. Instead of preparing a fixed design, engineers can assess configurations, introduce changes, perform system-level validation and generate engineering outputs within a shared environment.

This approach is intended to support the development of configurable AI factory infrastructure that is continuously co-designed and validated. Design intent is to be maintained from the early configuration stage through construction and commissioning to future technology upgrades.

Claimed reduction in deployment time

According to the companies, the expected customer benefits include up to a tenfold acceleration of the process from defining requirements to obtaining a configured and verified AI factory solution. The specified infrastructure deployment time could be reduced by up to 50%.

Based on the assumptions of Vertiv's model, the combined approach could enable productive capacity to be brought online around 10 weeks earlier than in the modelled conventional baseline scenario. With the same installed productive capacity, this is expected to provide up to 30% more cumulative output within 12 months of the start of the project. Vertiv notes that this benefit results solely from earlier site readiness and does not reflect improvements in operational efficiency or measured customer results.

"The winners in AI will be those who can turn computing capacity into productive capacity the fastest," said Scott Armul, Chief Product and Technology Officer at Vertiv. "This requires the entire AI factory to operate as a single system, covering computing infrastructure, power, cooling and physical infrastructure. By combining Vertiv's expertise in pre-engineered infrastructure solutions with the model-based capabilities of Dassault Systèmes, we can configure and verify this system before construction begins, reduce on-site complexity and help customers bring AI computing capacity online faster."

"AI factories are among the most complex engineering systems being built today. This complexity requires a systems approach, including modelling, simulation, coordination and optimisation within a virtual twin before the entire factory is built," said Elisa Prisner, Executive Vice President, Strategy, Industries, Marketing and Transformation at Dassault Systèmes. "Together with Vertiv, we are transforming standard infrastructure knowledge into configurable virtual twins for NVIDIA DSX AI factories. Our 3DEXPERIENCE platform allows customers to explore different configurations, understand the consequences of design decisions and virtually verify the performance of the complete system before any infrastructure is physically built."

Presentation at NVIDIA GTC Berlin

During the NVIDIA GTC Berlin conference, held from 20 to 22 October, the companies presented elements of the design and deployment approach. They demonstrated how facility requirements influence infrastructure configuration, how teams can virtually assess power and cooling constraints, and how Vertiv uses pre-engineered factory infrastructure to accelerate physical deployment.

The presentation combined the NVIDIA Vera Rubin DSX AI Factory reference design and NVIDIA Omniverse DSX Blueprint, the Dassault Systèmes 3DEXPERIENCE platform with model-based systems engineering capabilities, and Vertiv SimReady power and cooling solutions.