nvidia unveiled Omniverse DSX during the GTC Washington, D.C., keynote as a comprehensive, open blueprint for designing and operating gigawatt-scale Artificial Intelligence factories. the blueprint was validated at the AI Factory Research Center at Digital Realty’s site in Manassas, Virginia, and ties Omniverse libraries and OpenUSD into a repeatable recipe intended to co-design buildings, power and cooling with NVIDIA’s infrastructure stack.
the design workflow centers on digital twins and SimReady OpenUSD assets. engineering partner Jacobs aggregates equipment models from Siemens, Schneider Electric, Trane Technologies and Vertiv into PTC’s product lifecycle system so cross-functional teams can collaborate on photorealistic 3D models. Cadence’s Reality Digital Twin platform is built on the DSX blueprint and Omniverse libraries, accelerated by NVIDIA CUDA, enabling high-fidelity thermal and electrical simulation to test designs, predict failures and optimize operations before construction.
physical build-out is driven by prefabricated modules from partners such as Bechtel and Vertiv, which are factory built and ready to plug in for modular scalability and faster time to revenue. once live, the digital twin becomes an operating system to monitor and optimize the facility. partner AI agents from Phaidra and Emerald AI are trained in the twin to continuously optimize power, cooling and workloads, improving grid flexibility and energy efficiency. the DSX offering is organized around three pillars: DSX Flex for dynamic grid collaboration, DSX Boost for performance-per-watt optimization that can deliver up to 30% higher GPU throughput within the same power envelope, and DSX Exchange for unified IT/OT integration. the ecosystem of energy and equipment specialists – Eaton, GE Vernova, Hitachi, Schneider Electric, Siemens Energy, Trane Technologies and Vertiv – plus software partners including Cadence, Emerald AI, Phaidra, PTC and Switch, aims to deliver scalable, validated infrastructure from 100 megawatts to multi-gigawatt capacity optimized for NVIDIA Grace Blackwell and Vera Rubin platforms.
