EPRI and witnesses say data centers can provide grid flexibility while nuclear supplies baseload power
Get AI-powered insights, summaries, and transcripts
SubscribeSummary
EPRI described consortiums and pilots to make data centers more flexible grid partners; witnesses said combining data-center flexibility with nuclear baseload could ease integration of AI infrastructure.
Experts from EPRI and industry witnesses said data centers can be both large new loads and potential flexible resources to support grid reliability, and they described research initiatives to exploit that flexibility.
Dr. Jeremy Renshaw, executive director of AI and Quantum at EPRI, described two EPRI efforts: the Open Power AI consortium, a collaboration of more than 100 organizations to build more energy-efficient AI models and improve interconnection processes; and the DC Flex initiative, which will test how data centers can provide grid services by shifting workloads or using backup generation as dispatchable resources. Renshaw said some AI-specific data centers can consume up to five times more energy than traditional data centers and that they may have highly variable loads with spikes, creating both challenges and opportunities for the grid.
Committee members and witnesses discussed examples of flexibility: shifting training workloads across locations, using backup generators as dispatchable resources during peak hours, and leveraging colocated sites to avoid new long-distance transmission. Renshaw said experimental test sites for DC Flex were announced to evaluate data-center flexibility in practice.
Witnesses cautioned that not all workloads are deferrable and that market and contractual incentives will be necessary for data centers to provide flexibility without compromising service. Constellation and EPRI speakers also noted AI and other efficiency improvements could blunt demand growth over time.
