NVIDIA and Emerald AI have introduced a grid-integrated AI factory architecture designed to accelerate data center interconnection and improve power system utilization, working in partnership with AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra.
The release centers on the NVIDIA Vera Rubin DSX AI Factory reference design and DSX Flex software, which enable AI facilities to operate as flexible energy assets, balancing high-performance compute workloads with real-time grid responsiveness.
The announcement is primarily framed through an energy infrastructure lens, positioning AI factories as dispatchable, grid-supporting resources rather than static power loads. The DSX architecture integrates compute, energy, networking, and cooling into a unified system, while DSX Flex connects AI operations directly to grid services. Emerald AI’s Conductor platform orchestrates compute demand alongside on-site generation, storage, and behind-the-meter assets, allowing operators to dynamically adjust power usage without compromising workload performance.
This approach addresses a core bottleneck in AI infrastructure deployment: slow grid interconnection timelines. By enabling hybrid configurations with co-located generation and storage as bridge power, AI factories can come online faster while maintaining a pathway to full grid integration. The same infrastructure can later supply excess capacity back to the grid, improving asset utilization and reducing peak-driven system costs.
Participating energy providers will evaluate generation strategies aligned to this architecture, including hybrid and fully grid-connected models. The objective is to meet surging AI-driven electricity demand while improving reliability and deferring costly grid expansion.
Operationally, the model reframes AI data centers as active participants in energy markets. By aligning compute flexibility with grid conditions, these facilities can reduce peak demand pressure, unlock latent capacity across existing infrastructure, and support faster scaling of gigawatt-level AI deployments. NVIDIA and Emerald AI report that DSX Flex has already been tested in commercial environments, with planned deployment at the NVIDIA AI Factory Research Center in Virginia.