Key facts
- Data center electricity use in the US could nearly double by 2030, reaching 383 to 793 TWh.
- Data centers surveyed reported peak power reduction potential of 10% to 30%.
- Greater flexibility from data centers could save $40 billion to $150 billion in capital investments over the next decade.
- OpenAI agreed to cut electricity draw from a planned 3.2-gigawatt facility in Georgia by up to 1 gigawatt during grid stress.
Technology companies and utilities are increasingly exploring ways to make data centers more flexible in their electricity consumption, a strategy known as "demand response," to help manage the growing strain on US power grids driven by AI demand.
This approach involves data centers temporarily reducing or shifting their electricity use during periods of peak demand or grid stress. Experts believe this can reduce the need for costly grid upgrades and new power generation. According to the Electric Power Research Institute (EPRI), US data center electricity use could rise from approximately 177 to 192 terawatt-hours in 2024 to between 383 and 793 TWh by 2030. EPRI surveys indicate that data centers reported peak power reduction potential ranging from 10% to 30%, with some hyperscalers achieving even higher reductions.
A study by Duke University's Nicholas Institute estimated that enhanced flexibility from data centers could lead to savings of $40 billion to $150 billion in capital investments over the next decade. This flexibility could also expedite grid connections for facilities willing to curb electricity use during high-demand periods, as regulators and grid operators develop faster pathways.
While demand response has largely been implemented through pilot projects and one-off agreements, such as OpenAI's recent deal to cut electricity draw from a planned 3.2-gigawatt facility in Georgia by up to 1 gigawatt during grid stress, the effort is evolving into a broader initiative. In June, federal regulators directed grid operators to consider new rules for connecting large power users, including expedited pathways for facilities offering flexibility. Last month, Alphabet's Google, NVIDIA, and Emerald AI established the AI Energy Management Alliance to advance the deployment of flexible data centers.
Experts highlight that scaling this strategy requires mutual adjustments. Data centers must find ways to adjust power consumption without disrupting customers, while utilities and grid operators need to develop appropriate tariffs, market incentives, and faster interconnection processes to reward this flexibility. Alexander Kheder, an analyst with BMI, noted that while demand response can significantly mitigate peak impacts, expanding curtailment agreements to hundreds of new facilities will necessitate substantial capital expenditure and coordinated policy frameworks.