NVIDIA and Emerald AI collaborate with U.S. energy firms to develop adaptable AI infrastructure for grid management
AI-generated image: synthetic visual, not an actual depiction of events, people, or locations.
Key Takeaways
- NVIDIA and Emerald AI are collaborating with AES, Constellation, Invenergy, NextEra Energy, Nscale, and Vistra to develop grid-flexible AI factories.
- The project utilizes the NVIDIA Vera Rubin DSX reference design and DSX Flex software to integrate AI compute with power-grid services.
- Co-located generation and storage will serve as bridge power to accelerate the deployment of AI infrastructure ahead of standard interconnection timelines.
- Emerald AI's Conductor platform will orchestrate compute flexibility and onsite energy resources to provide grid-responsive power management.
The initiative utilizes the NVIDIA Vera Rubin DSX AI Factory reference design, which incorporates the DSX Flex software library. This architecture is designed to connect AI facilities directly to power-grid services, allowing them to function as responsive energy assets rather than passive loads.
To expedite deployment, the factories are designed to use co-located energy generation and storage as bridge power. This approach allows AI infrastructure to come online before traditional grid interconnections are finalized. Once connected, these resources can be used to supply the broader power system or adjust consumption during periods of grid stress.
Emerald AI's Conductor platform will provide the orchestration for these facilities. The software manages computational flexibility alongside behind-the-meter resources, such as onsite batteries and generation, to meet specific grid-responsive power targets while maintaining service quality for AI compute tenants.
The participating energy companies will evaluate optimized generation applications specifically for these AI factories. By pairing large-scale compute loads with intelligent controls and onsite energy resources, the collaboration aims to improve grid reliability and reduce the need for infrastructure expansion sized around peak demand.
