English
In early 2026, the collision between artificial intelligence expansion and grid constraints has become the defining bottleneck for tech growth. AI data centers are projected to consume over 1,000 terawatt-hours (TWh) of electricity globally by 2026—more than the entire annual consumption of most countries—triggering a structural energy crisis that is fundamentally redefining the relationship between Big Tech and global power markets. With grid interconnection delays exceeding three years and the United States facing a 49 GW generation shortfall through 2028, technology giants including Microsoft, Google, and Amazon are pivoting from public grid reliance to direct investments in nuclear restart projects and on-site generation.
The Scale of the Crisis
According to the International Energy Agency's Electricity 2024 report, AI data center power demands are on track to double to 1,000 TWh by 2026, representing approximately 3% of global electricity use. Morgan Stanley forecasts 126 GW of new data center-related power demand through 2028—nearly matching Canada's entire generating capacity. In the United States alone, PJM Interconnection, the largest grid operator serving 65 million people across 13 states, faces a projected generation shortfall of up to 49 GW by 2028 as AI load growth outpaces new capacity additions.
The grid interconnection crisis has become the single greatest barrier to deploying new energy and data center capacity. The national interconnection queue has ballooned to 2,600 GW, with median wait times approaching five years and project withdrawal rates nearing 80%. Critical electrical components like transformers and switchgear face lead times of 36 to 48 months, with many sourced from China under tariff disruptions. Nearly half of planned U.S. AI data centers for 2026 have been delayed or canceled, creating a 7 GW capacity shortfall.
Big Tech's Nuclear Pivot
Facing these constraints, hyperscalers are taking unprecedented steps to secure dedicated power supplies. The most symbolic project is the restart of Three Mile Island Unit 1 in Pennsylvania, which shut down in 2019. In September 2024, Constellation Energy announced plans to invest $1.6 billion to bring the 835 MW facility back online, renamed the Crane Clean Energy Center, with operations expected to resume in 2027. Microsoft has entered into a 20-year power purchase agreement to buy the plant's entire output to power its AI data centers. The Trump administration backed the project with a $1 billion Department of Energy loan, with the first advance expected in Q1 2026.
Amazon has signed a deal to power its AWS cloud and AI servers using 1.92 GW from Talen Energy's Susquehanna nuclear plant in Pennsylvania, with energy secured through 2042. The company is also exploring small modular reactors (SMRs) with X-energy, planning 300 MW of capacity in the Pacific Northwest and Virginia. Google and Meta have similarly announced investments in nuclear energy, though no commercial SMR is yet operational. Analysts at Deloitte predict nuclear energy could meet up to 10% of data center electricity demand by 2035.
This nuclear renaissance for AI marks a dramatic shift. Three Mile Island Unit 1 is one of three shuttered U.S. nuclear plants aiming to restart this decade, alongside Palisades in Michigan and Duane Arnold in Iowa. The trend reflects a broader strategic calculation: nuclear provides reliable, carbon-free baseload power that can operate 24/7, unlike intermittent renewables.
Community Backlash and Rising Costs
The AI power crunch is not just an infrastructure problem—it is a political and social flashpoint. U.S. residential electricity prices have risen 42% since 2019, far outpacing the 29% increase in the Consumer Price Index over the same period. In 2025 alone, utilities requested a record $31 billion in rate hikes—more than double the near-record from 2024—driven primarily by data center demand. PJM capacity market prices have spiked nearly tenfold, from $28.92 per megawatt-day in 2024/2025 to $329.17/MW-day for the 2026/2027 delivery year. Data centers accounted for 63% of the price increase, adding $9.3 billion in costs to ratepayers.
Communities are pushing back. In Northern Virginia—the world's largest data center market—Dominion Energy Virginia expects data center load to increase 3.5 times by 2038, coinciding with rising utility disconnections. AEP Ohio has paused new interconnections, and multiple states are demanding tech companies fund their own power infrastructure. Virginia has introduced legislation targeting large-load additions and barring utilities from passing costs to residential consumers. The Brookings Institution warns that without transparency and regulatory reform, public backlash could lead to stricter regulations on hyperscalers.
Charles Hua of PowerLines, a non-profit advocating for utility customers, argues: 'The root cause is how utilities are financially rewarded for building infrastructure rather than efficiency. Ratepayers are being asked to subsidize the power infrastructure needs of trillion-dollar tech companies.' The ratepayer subsidy debate has become a central policy question, with some lawmakers proposing tariffs or fees specifically targeting data center energy usage.
Geopolitical and Strategic Implications
The AI power crunch is reshaping not just energy markets but geopolitics. The World Economic Forum argues that AI infrastructure must now be treated as critical national infrastructure, on par with electricity grids and oil pipelines. This shift was underscored in March 2026 when Iranian drones struck Amazon Web Services data centers in the UAE and Bahrain, marking the first time hyperscale data centers became explicit kinetic targets in modern conflict.
The competition for power is also driving a rethinking of energy investment flows. Investor-owned utilities are projected to spend $1.1 trillion on power grid expansion between 2025 and 2029, with AI and data centers as key drivers. Battery storage is emerging as a critical solution—lithium-ion batteries offer 90% round-trip efficiency with costs dropping to $132/kWh, and the U.S. Inflation Reduction Act targets 100 GW of battery storage deployment by 2030. However, supply chain bottlenecks and interconnection delays continue to hamper progress.
Globally, over 3,000 GW of generation and storage projects are stuck in interconnection queues worldwide, representing a structural failure to align transmission investment with energy transition goals. The geopolitics of AI energy is becoming a defining issue for international relations, as countries grapple with data sovereignty, critical infrastructure security, and the strategic implications of tech companies becoming their own power utilities.
Expert Perspectives
Industry analysts are divided on the path forward. A SemiAnalysis report argues that market design and policy decisions play a greater role in price increases than AI infrastructure alone. Microsoft has outlined a five-point plan to cover additional electricity costs and pursue renewable energy, and President Trump hosted executives to affirm a Ratepayer Protection Pledge. However, experts question these commitments as the industry struggles with profitability.
Dr. Cathy Kunkel of the Institute for Energy Economics and Financial Analysis notes: 'Inflated demand forecasts, infrastructure build constraints, and clean energy goals may limit future data center expansion while ratepayers continue subsidizing the required grid infrastructure.' The PJM independent market monitor has recommended requiring new data centers to supply their own generation instead of tapping into existing power supplies.
FAQ
How much electricity will AI data centers consume by 2026?
AI data centers are projected to consume over 1,000 TWh globally by 2026, equivalent to about 3% of total global electricity use, according to the International Energy Agency.
Why are tech companies investing in nuclear power?
Tech companies are investing in nuclear power because it provides reliable, carbon-free baseload electricity that can operate 24/7, unlike intermittent renewables. Grid interconnection delays and rising electricity costs have made dedicated power sources essential for AI data center expansion.
How much have electricity prices risen due to data centers?
U.S. residential electricity prices have risen 42% since 2019, with data centers driving a significant portion of the increase. PJM capacity prices have increased nearly tenfold, adding $9.3 billion in costs to ratepayers.
What is the Three Mile Island restart project?
Constellation Energy is investing $1.6 billion to restart Three Mile Island Unit 1 (835 MW) in Pennsylvania, renamed the Crane Clean Energy Center. Microsoft has signed a 20-year power purchase agreement to buy the plant's entire output for its AI data centers. Operations are expected to resume in 2027.
Are data centers causing utility rate hikes?
Yes. Utilities requested a record $31 billion in rate hikes in 2025, more than double the previous year, driven primarily by data center demand. Communities and regulators are pushing back, with several states proposing legislation to ensure tech companies bear more of the financial burden.
Conclusion and Future Outlook
The AI power crunch of 2026 represents a structural shift in the relationship between technology and energy. As data center electricity consumption is projected to nearly triple by 2028, the competition for power will only intensify. The defining question is whether ratepayers should subsidize the infrastructure needs of trillion-dollar tech companies, or whether hyperscalers must become their own power utilities. The answer will shape not only energy markets but the future of AI itself. With nuclear restart projects moving forward, utilities requesting unprecedented rate hikes, and governments beginning to wrestle with the strategic implications, the collision between AI and energy has become the defining infrastructure story of the decade.
Follow Discussion