News
NVIDIA says companies are using DSX to model power, cooling and compute together as grid constraints and interconnection delays slow AI factory projects.
NVIDIA said companies are using its DSX AI Factory Platform to plan, simulate and run grid infrastructure and AI factories as one system, a response to what it described as a growing power bottleneck in data-center expansion. In a blog post published Oct. 22, 2026, NVIDIA said more than $300 billion in equipment is backlogged and more than 200 gigawatts of projects are waiting in U.S. interconnection queues.
Evidence and context
The company said the problem stems in part from siloed planning, with the grid, facility and compute stacks traditionally designed separately. NVIDIA said DSX is intended to bring those layers together, using DSX Sim and SimReady assets to model substations, generators, switchgear, power systems and cooling before construction starts. It also said DSX Flex is designed to align AI-factory demand with grid conditions and onsite generation, while DSX MaxLPS and DSX Exchange are meant to improve compute output per watt and share operational signals across compute, networking, energy, cooling and operations.
NVIDIA also pointed to a set of partner integrations. GE Vernova is working with NVIDIA to pair grid-to-facility and rack-power expertise with DSX, according to the post. Hitachi Energy’s battery energy storage system has qualified as NVIDIA DSX Ready, which NVIDIA said is intended to help customers identify infrastructure that matches applicable reference-architecture requirements for resilient and scalable AI factories. Schneider Electric, with AVEVA and ETAP, is advancing validated NVIDIA Vera Rubin power and cooling reference architectures and an Omniverse-powered digital twin. Siemens Energy is developing a reference design for grid-aware power infrastructure aligned with the Vera Rubin AI factory design.
NVIDIA said several companies are using DSX to improve efficiency or grid flexibility. It said Phaidra integrated DSX Flex, DSX Max-Q and DSX Exchange into its self-learning agents and that the setup is designed to optimize cooling and power allocation, delivering about 10% more compute within the same power envelope. NVIDIA also said Emerald AI integrated its Conductor platform into DSX Flex to help utilities tap underused grid capacity, with a model it said could unlock up to 100 gigawatts on the existing U.S. grid. That figure is a vendor-stated modeling claim, not an independently verified result.
The announcement follows earlier NVIDIA DSX updates in recent weeks, including DSX POD AI factories for Europe on Oct. 21, 2026, open sourcing of Run:ai on Oct. 13, 2026, and the launch of DSX Ready for power and cooling products on Sept. 21, 2026. NVIDIA’s broader DSX documentation describes the platform as a set of reference designs, APIs, software libraries and partner technologies aimed at faster deployment, lower token cost, improved energy efficiency and operational resilience.
Be the first to comment