Get a Load of This

This report by RMI examines the challenges of forecasting large electricity loads—such as data centers and advanced manufacturing—in the United States. It argues that traditional forecasting methods are insufficient for the rapid growth and unique operational characteristics of these loads, which can lead to risks in grid reliability and customer affordability. The document provides a framework for understanding load characteristics, outlines best practices for utilities, and offers specific regulatory actions to ensure forecasts are thorough, up-to-date, and validated.

• Electricity demand in the United States is experiencing a significant upward trend, with utility integrated resource plans (IRPs) covering 48% of sales projecting a 20% load growth from 2023 through 2035, a substantial increase from the 7% growth projected in January 2021.
• Historically, US utilities have systematically over-forecasted electricity demand. Between 2006 and 2023, 5-year forecasts were over-forecasted by an average of 8%, and 10-year forecasts by 17%. More recent data from 2012 to 2023 shows an even higher average over-forecast of 23%.
• Large load drivers possess unique characteristics that impact grid planning differently than traditional loads. These include load shape (timing of demand), forecast uncertainty (confidence in projections), flexibility potential (ability to change consumption based on signals), and flight risk (likelihood of shifting locations due to economic or grid conditions).
• Data centers present high forecast uncertainty and significant geographic concentration, with 15 states accounting for 80% of the national data center load. In Virginia, data centers were estimated to comprise 25.6% of the electric load in 2023, with projections suggesting this could reach 46% by 2030.
• Industrial and manufacturing loads are returning to the US, driven by the clean energy transition and Inflation Reduction Act incentives. These loads typically have lower forecasting uncertainty than data centers because siting and permitting require substantial time and specific infrastructure.
• RMI recommends several forecasting best practices: using scenario-based or stochastic methods instead of single deterministic forecasts, integrating end-use forecasting with econometric models to avoid double counting, and ensuring consistent data use across all planning processes.
• Regulators can improve forecasting by pursuing three goals: making forecasts 'Thorough' (reflecting unique load characteristics), 'Up-to-date' (increasing update frequency, such as quarterly), and 'Validated' (increasing transparency and accountability).
• To manage uncertainty, regulators are encouraged to prioritize 'least-regrets' capital investments—those that are fast, affordable, and flexible (e.g., energy efficiency, VPPs, and grid-enhancing technologies)—and to use tariff designs that shift financial risk to the large customers driving the growth.
• Case studies show emerging practices in the US: Georgia Power now files quarterly large load updates; Duke Energy in North Carolina uses an economic development adjustment to avoid double counting projects in econometric models; and Dominion Energy in Virginia uses a mix of statistical methods and historical metered data to create high, medium, and low load scenarios.

APA

Sward, J., Shwisberg, L., Stephan, K., & Becker, J. (2025). Get a Load of This. RMI. https://rmi.org/app/uploads/dlm_uploads/2025/03/Get_a_load_of_this_Load_Forecasting.pdf

Chicago

Sward, Jeffrey, Lauren Shwisberg, Katerina Stephan, and Jacob Becker. Get a Load of This. RMI, 2025. https://rmi.org/app/uploads/dlm_uploads/2025/03/Get_a_load_of_this_Load_Forecasting.pdf.

Wikipedia

{{cite report |last1=Sward |first1=Jeffrey |last2=Shwisberg |first2=Lauren |last3=Stephan |first3=Katerina |last4=Becker |first4=Jacob |title=Get a Load of This |publisher=RMI |date=February 2025 |url=https://rmi.org/app/uploads/dlm_uploads/2025/03/Get_a_load_of_this_Load_Forecasting.pdf |access-date=17 August 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{sward2025get, author = {Sward, Jeffrey and Shwisberg, Lauren and Stephan, Katerina and Becker, Jacob}, title = {{Get a Load of This}}, institution = {RMI}, year = {2025}, month = feb, url = {https://rmi.org/app/uploads/dlm_uploads/2025/03/Get_a_load_of_this_Load_Forecasting.pdf}, urldate = {2026-08-17}, note = {Indexed by Climate Insights Directory} }