Powering the AI Revolution: Energy Infrastructure in the United States

Part I

Global Economics Research Team

By: Patrick Gorman, The Quinnipiac University Global Economics Research Team

Key Points

• U.S. data center electricity consumption grew from approximately 108 TWh in 2020 to 183 TWh in 2024, a 69% increase, driven predominantly by the rapid scaling of AI infrastructure.

• The ChatGPT launch in November 2022 marked a structural inflection point, with Amazon, Microsoft, Google, and Meta collectively exceeding $200 billion in capital expenditure in 2024 alone.

• Nuclear energy has emerged as the preferred baseload power source for AI data centers, catalyzing landmark Power Purchase Agreements and first-of-its-kind co-location deals, including the high-profile restart of Three Mile Island.

The AI Energy Demand Surge: 2022–2026

For most of the past decade, growth in U.S. data center electricity consumption was modest and largely offset by efficiency improvements. That pattern broke decisively in late 2022. The commercial release of OpenAI's ChatGPT triggered a structural shift in power demand driven by the exponentially higher energy requirements of generative AI workloads. According to the International Energy Agency (IEA), U.S. data center electricity consumption grew from approximately 108 TWh (Terawatt per hour) in 2020 to 183 TWh in 2024, roughly 4% of national consumption.^[1] The Lawrence Berkeley National Laboratory (LBNL), in a 2024 report commissioned by the U.S. Department of Energy, projected further growth to between 325 and 580 TWh by 2028.^[2] Forecasts published by utilities for five-year summer peak demand growth jumped from 38 GW in 2023 to 128 GW in 2024, reflecting the degree to which AI's energy appetite caught grid planners off guard.^{[9, 10]}

A primary driver of this surge was a sharp escalation in capital expenditure by the largest technology companies. In 2024, Amazon, Microsoft, Google, and Meta collectively spent over $200 billion in capital expenditures, a 62% year-over-year increase, with each firm posting an all-time high.^[9] Analysts at Harvard's Belfer Center described data center demand as outpacing available capacity in key markets, forcing firms to contract power directly from private producers.^[3] The energy bottleneck, rather than computing hardware or capital, has become the defining constraint on AI infrastructure expansion.

The Nuclear Sector: Key Developments

Nuclear power has emerged as the preferred energy source for large-scale AI data centers due to its ability to operate continuously at high-capacity factors, typically above 90%, while producing no direct carbon emissions. In September 2024, Constellation Energy and Microsoft announced a 20-year Power Purchase Agreement (PPA) under which Constellation committed $1.6 billion to restart the Unit 1 reactor at Three Mile Island in Pennsylvania, renamed the Crane Clean Energy Center.^[4] The plant had been shut down in 2019 and is now targeting a return to service in 2027. The Department of Energy provided a $1 billion federal loan to support the restart in November 2025.^[5] Constellation followed with a second major deal in June 2025, a 20-year PPA with Meta Platforms for 1,100 MW from its Clinton, Illinois, plant.^[4] In a structurally different model, Amazon Web Services paid $650 million in March 2024 to acquire a data center campus co-located directly adjacent to Talen Energy's Susquehanna nuclear plant in Pennsylvania, drawing power through a behind-the-meter connection that bypasses the broader transmission grid.^[3]

Looking further ahead, several technology companies have begun laying the groundwork for Small Modular Reactors (SMRs), a next-generation nuclear technology designed for modular construction in increments of 50–300 MW. Google signed an agreement with Kairos Power in 2024 targeting SMR deployment by approximately 2030, while Microsoft has backed TerraPower, which broke ground on a demonstration reactor in Wyoming in June 2024.^[3] No SMRs are yet commercially operational in the United States, and their inclusion in current plans reflects strategic optionality rather than confirmed deployment. Should SMR technology prove cost-effective at scale, it could reshape how data center operators approach long-term energy sourcing in the 2030s and beyond.

The Electric Utility Sector: Grid Stress and Response

Virginia has emerged as the most striking example of grid strain, hosting over 600 data centers that alone account for roughly 39% of the state's total electricity consumption.^[6] Dominion Energy, the primary regulated utility serving the state, reported to Virginia's State Corporation Commission in February 2026 that total power requests from data centers in its territory had reached 70,000 MW, nearly three times its all-time peak load of 24,678 MW.^[7] Its contracted pipeline grew from 21 GW in mid-2024 to 47.1 GW by late 2025. To accommodate this trajectory, Dominion's 2024 Integrated Resource Plan projects $50.1 billion in capital investment through 2029, with 80% of new capacity targeted from carbon-free sources.^[6] The strain has reached electricity markets, the PJM regional capacity auction clearing price for 2025–2026 rose 833% from the prior year, and in November 2025 Virginia approved a new dedicated rate class for large data center customers, effective January 2027.^[8]

Beyond Virginia, NextEra Energy, the world’s largest wind and solar producer, announced plans in October 2024 to restart Iowa’s Duane Arnold nuclear plant through a 25-year PPA with Google, adding to a portfolio spanning renewables, nuclear restarts, and SMR development. These developments signal that AI-driven energy demand is no longer a regional story confined to the Mid-Atlantic corridor, but a national infrastructure challenge taking shape across multiple grid systems simultaneously.^[3] In the South, Entergy, serving Arkansas, Louisiana, Mississippi, and Texas, disclosed a $10 billion Meta Platforms AI data center project in Northeast Louisiana, committing to construct 1.5 GW of new gas generation to meet demand it projects will grow at 13–14% annually.^[6]

Key U.S. Companies in the Market

Company	Sector	Role in AI Energy
Constellation Energy	Nuclear Utility	Largest U.S. nuclear fleet operator; 20-year PPAs with Microsoft (Three Mile Island restart, 835 MW) and Meta (Clinton plant, 1,100 MW)
Dominion Energy	Regulated Utility	Primary utility in Virginia; managing 47+ GW in data center pipeline; $50.1B capex plan through 2029
NextEra Energy	Renewable/Nuclear	World's largest wind and solar producer; restarting Duane Arnold nuclear plant (Iowa) via 25-year PPA with Google
Talen Energy	Ind. Power Producer	First-of-its-kind nuclear co-location deal with Amazon AWS at Susquehanna plant in Pennsylvania
Entergy	Regulated Utility	Supporting Meta's $10B Louisiana AI data center; constructing 1.5 GW of new gas generation

Sources

[1] International Energy Agency. 'Energy and AI — Energy Demand from AI.' 2025. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai

[2] U.S. DOE / Lawrence Berkeley National Laboratory. '2024 Report on U.S. Data Center Energy Use.' https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers

[3] Belfer Center, Harvard Kennedy School. 'AI, Data Centers, and the U.S. Electric Grid.' 2025. https://www.belfercenter.org/research-analysis/ai-data-centers-us-electric-grid

[4] Constellation Energy. 'Constellation to Launch Crane Clean Energy Center.' September 2024. https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html

[5] CNBC. 'Trump administration backs Three Mile Island nuclear restart with $1 billion loan.' November 2025. https://www.cnbc.com/2025/11/18/trump-nuclear-three-mile-island-crane-loan-constellation-ceg.html

[6] American Action Forum. 'Meeting AI Data Center Power Demand in Virginia.' December 2025. https://www.americanactionforum.org/insight/meeting-ai-data-center-power-demand-in-virginia-challenges-posed-by-carbon-free-electricity-regulations/

[7] Virginia Business. 'Dominion prepares for 70,000 MW in data center demand.' February 2026. https://virginiabusiness.com/dominion-data-center-power-demand-virginia-scc/

[8] American Action Forum. 'Virginia's New Data Center Electricity Rate Class.' January 2026. https://www.americanactionforum.org/insight/virginias-new-data-center-electricity-rate-class/

[9] Pew Research Center. 'What we know about energy use at U.S. data centers amid the AI boom.' October 2025. https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom/

[10] World Resources Institute. 'Powering the US Data Center Boom: The Challenge of Forecasting Electricity Needs.' https://www.wri.org/insights/us-data-centers-electricity-demand