Nvidia has unveiled Omniverse DSX at its recent GTC event held in Washington, D.C. This initiative serves as a blueprint for the design and operation of gigawatt-scale AI datacenters, leveraging digital twin technology. Alongside this announcement, the company confirmed plans to establish an AI Factory Research Center at Digital Realty“s facility in Manassas, Virginia. This research center aims to develop a platform that merges Nvidia”s Omniverse simulation environment with the open-source Universal Scene Description (OpenUSD) technology.
During his keynote address, Nvidia CEO Jensen Huang stated, “Nvidia started out by designing chips, and then we started to design systems, and we designed AI supercomputers. Now we”re designing entire AI factories.” The Omniverse DSX platform creates a digital twin of the physical datacenter, encompassing all aspects such as the building structure, power, and cooling systems. Once operational, this digital twin will function as an operating system, facilitating continuous optimization for performance, energy efficiency, and sustainability.
As part of this project, Schneider Electric has joined forces with Nvidia, having introduced its own Omniverse-based digital twin platform at the GTC event earlier in the year. Huang emphasized the unique nature of these GPU-intensive facilities, distinguishing them from conventional datacenters. “It”s an AI factory because this factory produces one thing,” he elaborated. “It runs AI. And its purpose is designed to produce tokens that are as valuable as possible, meaning they have to be smart. And you want to produce these tokens at incredible rates because when you ask an AI for something, you would like it to respond.”
Huang also mentioned that this AI factory is being constructed for the Vera Rubin project, highlighting the technology”s capability to allow all partners to integrate digitally into the factory”s framework. He previously asserted that AI factories would become “the bedrock of modern economies” across the globe.
Additionally, Nvidia introduced BlueField-4 during the GTC event, which represents their next-generation data processing unit (DPU) designed to enhance infrastructure operations within AI datacenters. This DPU is engineered to offload network and storage functions and combines a Grace CPU featuring 64 Arm Neoverse V2 cores with ConnectX-9 networking, supporting Ethernet and InfiniBand at speeds up to 800 Gbps per port. Nvidia claims that the new DPU offers six times the compute power of its predecessor, BlueField-3. This silicon is specifically designed as a comprehensive engine for a new class of AI storage platform, serving as the backbone for high-performance data pipelines necessary for AI factories. The initial availability of BlueField-4 is anticipated to coincide with the launch of the Vera Rubin platform in 2026.
