U.S. electricity demand is rising, and data centers are a large part of the reason. They used about 5% of U.S. electricity in 2025 and could use 9% to 17% by 2030. That growth comes as federal clean energy policy shifts: the 2025 budget law changed eligibility and timelines for many of the Inflation Reduction Act's tax credits. For companies, these changes affect what power costs, how clean it is, and how to account for it.
Our research uses energy-economic models and statistical methods to test common assumptions about the power system, from whether data centers raise household bills to how federal policy changes the pace of clean energy investment. You can see a summary of related research and publications below. We'll add to this brief as new research becomes available.
Watershed research on energy and electricity
Policy: How federal law shapes the clean energy buildout
Impacts of the Inflation Reduction Act and One Big Beautiful Bill Act on the US energy system Nature Reviews Clean Technology, 2026. A review of how the Inflation Reduction Act affected clean energy deployment through mid-2025, and what modeling studies project for emissions and clean energy investment under both the original law and the 2025 budget law that changed many of its tax credits. Led by John Bistline with 13 co-authors, including researchers at the University of Maryland's Center for Global Sustainability.
Prices: What data centers mean for electricity bills
Have data centers raised your electric bill? Causal evidence from the United States arXiv preprint, June 2026. Tests whether data center growth caused household electricity prices to rise, using state-level data from 2015 to 2024 and a method that separates cause from correlation. It finds that data centers lowered average residential rates slightly over that period, and explains the conditions that could reverse the result. Written with researchers at the Electric Power Research Institute (EPRI).
Measurement: Accounting for the electricity behind AI
Estimating GHG emissions from AI use: Framework for corporate-level measurement Watershed white paper, July 2026. An open framework for estimating the electricity use and emissions of a company's AI use. On the electricity side, it documents how fast data center demand is growing, why on-site power plants at data centers fall outside standard grid emission factors, and how clean energy purchases can open a wide gap between location-based and market-based results. Developed with contributors from Stanford University and Tsinghua University. Also featured in our AI emissions research brief.
Key findings
Across this work, one pattern holds: electricity demand is rising quickly, and whether that growth raises or lowers prices and emissions depends on how new supply gets built and who pays for it.
Data centers and electricity prices
- Data centers have lowered household rates slightly, so far. From 2015 to 2024, U.S. data center capacity grew by 21.3 gigawatts. A 10% increase in a state's data center capacity lowered average residential electricity prices by about 0.4%. (Watten, Bistline & Blanford, 2026)
- The reason is scale. Power systems carry large fixed costs for transmission, distribution, and generation. Steady new demand spreads those costs over more kilowatt-hours, which lowers the average price. (Watten, Bistline & Blanford, 2026)
- Costs were not shifted onto households. Effects on commercial and industrial customers were smaller than for residential customers, and the study found no evidence of systematic cost-shifting between customer classes. (Watten, Bistline & Blanford, 2026)
- The result could reverse. Shortages of turbines and transformers and slow permitting raise the cost of new capacity. If utilities build for data center demand that never arrives, fixed costs get spread over fewer kilowatt-hours. Kansas, Michigan, and Delaware have used minimum bills or long-term contracts for large new loads to limit that risk. (Watten, Bistline & Blanford, 2026)
How data centers are powered
- Data centers' share of U.S. electricity could double or triple by 2030. Data centers are expected to use 9% to 17% of U.S. electricity by 2030, up from about 5% in 2025. (Bistline et al., 2026b, citing EPRI, 2026)
- Many new data centers plan to make their own power. About 30% of planned U.S. data center capacity plans to generate electricity on site, nearly three-quarters of it from natural gas. Standard grid emission factors don't capture this power, and how to report it under the GHG Protocol is unsettled. (Bistline et al., 2026b)
- Clean energy purchases widen the reporting gap. Power purchase agreements and renewable energy certificates can lower market-based emissions even where the local grid is carbon-intensive, which can leave location-based and market-based figures far apart. (Bistline et al., 2026b)
Federal policy and clean energy investment
- The Inflation Reduction Act was expected to speed up existing trends. Studies projected it would roughly double clean electricity investment over the next decade, across renewables, storage, nuclear, and plants with carbon capture, with near-term air quality and health benefits. (Bistline et al., 2026a)
- Households were expected to save. Projections showed energy spending falling by about $68 to $430 per household per year by 2035, compared with a case without the law. (Bistline et al., 2026a)
- The 2025 budget law moves projections back toward a no-IRA path. With credits narrowed or phased out sooner, modeled emissions and power-sector investment shift toward a scenario without the original law. (Bistline et al., 2026a)
- Investment stays high even without the credits. Because solar, batteries, and wind are cost-competitive and electricity demand is rising, total investment is still projected near the high end of its historical range. (Bistline et al., 2026a)
Note: The Nature Reviews Clean Technology article reviews and summarizes many modeling studies; its figures are projections, not measurements. The data center price study is a preprint and has not yet completed peer review. Figures from different studies use different methods and time periods and should not be compared directly.
What this means for sustainability teams
Your electricity emissions can shift for reasons outside your operations. Data centers could use 9% to 17% of U.S. electricity by 2030, and about 30% of planned data center capacity plans to make its own power on site, mostly from natural gas. Grid-average emission factors don't capture that power, so estimates for cloud and AI services that rely on them may not reflect how those services are actually powered. Reporting electricity use (in kWh) alongside emissions, on both a location-based and a market-based basis, shows which changes came from your own choices and which came from the grid. It's also worth asking cloud and AI providers whether any of the power serving your workloads comes from on-site generation.
About the researchers
Dr. John Bistline: Head of Science at Watershed. His research uses large-scale energy models to inform corporate decarbonization, power system planning, and climate policy, including the energy and emissions impacts of AI and data center growth. He previously led energy systems and climate analysis research at EPRI, is a Research Affiliate at MIT and an Associate Editor of The Electricity Journal, and is a contributing author to the IPCC Seventh Assessment Report. He holds a Ph.D. in Management Science and Engineering from Stanford University.
Dr. Mo Li: Climate Science Lead at Watershed, leading environmental modeling systems, the CEDA emissions database, and the Cornerstone open data collaboration. Her early research developed marginal emission factors that account for renewable generation, and their implications for energy storage and grid decarbonization. She holds a Ph.D. in Natural Resources Science and Management from the University of Minnesota.
Asa Watten: Researcher in the Energy Systems and Climate Analysis Group at Electric Power Research Institute (EPRI). He is an applied micro-economist and expert on consumer adoption of energy technologies and technology policy.
Geoffrey Blanford: Dr. Blanford is a Technical Executive and Program Manager for Energy and Climate Policy Analysis with the Electric Power Research Institute (EPRI) in Palo Alto, CA, where he has worked since 2006. He was a lead author for the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report and serves as co-director of the International Energy Workshop (IEW).
