Virtual Power Plants to Accelerate U.S. AI Data Center Growth

RMI publishes an analytical brief recommending rapid deployment and novel commercial models for virtual power plants (VPPs) to speed interconnection and reliably supply growing AI/data-center loads in the United States.

• Main announcement/action: RMI proposes three commercial models (Pass-Through Funding for Utility-Managed VPPs; VPP Capacity Transfer; VPP as Reliability Reinforcement) to enable large loads to sponsor VPP capacity in exchange for expedited interconnection. Key figures and timelines include VPPs could scale to meet over 20% of US peak demand by 2030, Brattle’s finding that 400 MW of VPP resource adequacy costs ~$2 million annually versus ~$43 million for equivalent new gas plants and grid upgrades, and examples of program scale such as California’s DSGS enrolling over 750 MW (including a 500 MW increase during Jan–Oct 2025). Utility and grid timing constraints cited include gas turbine backlogs through at least 2028, average interconnection timelines >5 years, and localized waits (e.g., Dominion warns up to 7 years in Northern Virginia; some DFW data center deliveries delayed to 2027 or later).

• Background and implementation details: The brief documents operational examples (National Grid Connected Solutions; Green Mountain Power battery programs; Ontario 90 MW residential VPP that enrolled 100,000 homes in six months) and outlines policy and market prerequisites: changes to interconnection policy (e.g., Oregon, Nevada, CAISO, SPP experiments), stronger integrated planning and data access (capacity accreditation, Green Button Connect), and customer protection measures (transparent tariffs, up-front payments/long-term contracts, rate-design evaluations). It emphasizes measurement & verification, transferable capacity credits, and that models shift different financial and delivery risks among large loads, VPP aggregators, and utilities.