NVIDIA and Emerald AI pitch flexible AI factories as grid assets for faster power access

NVIDIA and Emerald AI said on March 23, 2026 at CERAWeek 2026 that they are working with AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra on a new model for AI infrastructure. Instead of treating large AI data centers as passive power loads, the companies want to build flexible AI factories that can connect to the grid faster, generate AI tokens and intelligence, and also operate as grid-supporting energy assets.

The announcement ties that approach to the NVIDIA Vera Rubin DSX AI Factory reference design and the DSX Flex software library. According to NVIDIA, DSX Flex is designed to connect AI factories to power-grid services, while Emerald AI's Conductor platform coordinates compute flexibility with onsite generation, batteries, and other behind-the-meter resources. The goal is to protect quality of service for AI tenants while giving operators a more precise way to respond to grid conditions and interconnection constraints.

The companies argue that this matters because gigawatt-scale AI projects are increasingly running into slow grid interconnection timelines. One workaround is to use co-located generation and storage as bridge power, but permanently isolating those assets can leave them underused and raise the long-term cost of AI output. NVIDIA and Emerald AI are proposing hybrid AI factories that use those resources to get online faster and then later supply flexible capacity back to the grid. They also said the reference architecture can support flexible AI factories that are grid-connected from the outset, even when there is no co-located power resource.

NVIDIA said power-flexible AI factories could help unlock up to 100 gigawatts of capacity across the U.S. power system by combining smarter infrastructure design with more efficient use of existing assets and targeted new generation. The participating energy companies echoed that point, describing flexible operations as a way to speed time to power, reduce pressure on peak demand, and support broader grid reliability while AI demand grows.

NVIDIA also said DSX Flex is expected to reach commercial-scale deployment later this year at the NVIDIA AI Factory Research Center in Virginia, which is planned as one of the world's first power-flexible AI factories built on Vera Rubin infrastructure. The significance of the announcement is broader than one partnership: it shows how AI data center planning is moving toward a model in which compute, storage, generation, and grid services are designed together rather than treated as separate layers.