Overview
Since late 2022, U.S. regulators and utilities have warned that a new class of digital infrastructure—AI-optimized data centers—could reshape national power demand, ending an era of flat electricity consumption and forcing a rapid buildout of generation and transmission. Regional grid operators like PJM now attribute a large share of their unexpectedly steep load growth to hyperscale data centers clustered in places like northern Virginia, while federal studies and consulting reports show national load forecasts rising sixfold versus pre-AI expectations.
In this context, NextEra Energy’s December 8, 2025 expansion of its partnership with Google Cloud—to co-develop multiple gigawatt-scale U.S. data center campuses, bundled with new power plants and grid capacity, and to launch AI tools for predicting grid equipment failures—marks a pivotal moment in how Big Tech and major utilities are jointly re-planning the U.S. grid. At the same time, NextEra’s 2.5 GW of clean-energy deals with Meta and decades-long nuclear supply contracts into the 2050s underscore a broader shift toward dedicated, long-term power arrangements for AI infrastructure, even as environmental groups push for moratoria on new data centers and watchdogs warn that large loads could compromise grid reliability if not carefully integrated.
Key Indicators
People Involved
Organizations Involved
NextEra Energy is a Fortune 200 energy company headquartered in Juno Beach, Florida, and the world’s largest electric utility holding company by market capitalization, with a diverse portfolio spanning renewables, natural gas and nuclear.
Google Cloud is Alphabet’s enterprise cloud and AI division, operating a global network of data centers and providing infrastructure, platform and AI services to businesses and governments.
Meta is a U.S.-based technology conglomerate that operates Facebook, Instagram, WhatsApp and other platforms, and runs a large global network of data centers increasingly dedicated to AI workloads.
PJM is the regional transmission organization coordinating wholesale electricity markets and grid operations for all or parts of 13 U.S. states and D.C., serving about 65 million people and operating 185 GW of generating capacity.
WPPI Energy is a member-owned, not-for-profit power company serving 51 locally owned utilities across Wisconsin, Iowa and Michigan.
The IEA is an autonomous intergovernmental organization that produces analysis and policy advice on global energy markets, now including AI and data centres.
Timeline
-
Environmental groups call for national moratorium on new U.S. data centers
AdvocacyOver 230 environmental organizations, including Greenpeace and Friends of the Earth, demand a nationwide moratorium on new data centers, citing their heavy electricity and water use, contributions to climate change, and impacts on consumer bills. They note that local opposition has already stalled or canceled an estimated $64 billion in projects, and that data centres could add 44 million tons of CO₂ emissions by 2030, even as AI companies argue they are decarbonizing supply.
-
NextEra and Google Cloud expand partnership for gigawatt-scale data center campuses
Corporate DealNextEra Energy and Google Cloud announce a major expansion of their long-standing collaboration: they will jointly develop multiple new gigawatt-scale data center campuses across the U.S., each paired with new generation and grid capacity, and release an AI-driven product by mid‑2026 to predict equipment failures and improve grid reliability. NextEra simultaneously reveals that it has reached roughly 2.5 GW of clean-energy contracts with Meta and extended nuclear supply to WPPI Energy into the 2050s, and raises its 2025–26 earnings guidance on the back of data center demand.
-
Point Beach nuclear license extended into the 2050s, enabling new long-term contracts
RegulatoryThe NRC approves subsequent license renewal for NextEra’s Point Beach Nuclear Plant, allowing its units to operate through 2050 and 2053. The decision supports new long-term contracts such as WPPI Energy’s agreement to continue taking 168 MW into the 2050s, partly to serve growing regional loads.
-
Pennsylvania emerges as a major gas- and nuclear-adjacent data center hub
InfrastructureReporting highlights that Pennsylvania added about 2.4 GW of data center capacity from March 2024 to March 2025, leveraging abundant natural gas and proximity to nuclear plants. Amazon alone plans a $20 billion investment in two campuses, including one next to a nuclear station, while PJM’s role as the largest grid operator makes the state central to AI-era power planning.
-
EIA forecasts double-digit retail load growth in ERCOT and strong growth in PJM from data centers
ForecastThe U.S. EIA’s Short-Term Energy Outlook projects retail electricity sales in ERCOT growing at an average 11% in 2025–26 and PJM by 4%, compared with 2.2% nationwide. Analysts attribute much of this growth to a wave of new data centres, particularly AI-focused facilities.
-
IEA’s "Energy and AI" report projects data centre demand will more than double by 2030
ReportThe IEA publishes its Energy and AI special report, estimating that global data centre electricity use will rise from about 415 TWh in 2024 to around 945 TWh in 2030, with AI as the largest driver. It warns that in the U.S., data centres could account for nearly half of demand growth this decade, demanding massive investment in grids and generation.
-
Corporate clean energy procurement hits record 68 GW, led by data centers
AnalysisS&P Global Commodity Insights reports that corporations contracted a record 68 GW of clean energy in 2024, a 29% annual increase, with data centers leading procurement at over 17 GW. Microsoft, Amazon, Google and Meta top the list of buyers, underscoring how hyperscalers are reshaping power markets.
-
PJM 2025 Long-Term Load Forecast warns of capacity shortfall amid exponential data center growth
ForecastPJM’s 2025 forecast extends its horizon to 20 years, projecting summer peaks rising about 70 GW to 220,000 MW and winter peaks reaching ~210,000 MW by 2039. The operator explicitly cites exponential data center growth, building and vehicle electrification, and new manufacturing as core drivers, and warns of potential capacity shortages as early as the 2026/27 delivery year.
-
Amazon announces SMR and nuclear agreements to power future data centers
Corporate DealAmazon reveals three agreements to support development of nuclear energy projects, including advanced small modular reactors in Washington state, a potential SMR near Dominion’s North Anna plant in Virginia, and a data center co-located with a Pennsylvania nuclear facility, signaling tech’s growing role in underwriting firm carbon-free power.
-
Microsoft signs 20-year PPA to revive Three Mile Island Unit 1 for AI data centers
Corporate DealConstellation Energy and Microsoft agree to a 20-year power purchase agreement taking 100% of output from a revived ~835–837 MW Three Mile Island nuclear unit starting in 2028, with power matched to Microsoft’s AI data center loads across several PJM states. The deal becomes a flagship example of tech-driven nuclear restarts.
-
Grid Strategies’ "The Era of Flat Power Demand is Over" highlights surging load forecasts
AnalysisConsultancy Grid Strategies releases a widely cited report showing that U.S. utility load forecasts, after decades of flat expectations, have turned sharply upward due to data centers, manufacturing and electrification, marking a conceptual shift in how planners view future demand.
-
PJM’s 2023 Load Forecast flags data centers as a major new driver of demand
ForecastPJM publishes its 2023 Long-Term Load Forecast, using improved modeling to project higher growth rates and explicitly citing rising energy demand from data centers—mostly in Virginia and Maryland—as a chief factor behind increased load expectations.
-
New York enacts first-in-the-nation moratorium on certain fossil-fueled cryptomining
PolicyNew York Gov. Kathy Hochul signs a two-year moratorium on new or renewed permits for proof-of-work cryptocurrency mining operations at fossil-fueled power plants. The law reflects early concern that energy-intensive digital loads can undermine climate goals and strain local grids, foreshadowing similar debates now emerging around AI data centers.
Scenarios
Coordinated AI–Energy Buildout: Data Center Boom Continues, Grid Holds
Discussed by: IEA, Grid Strategies, S&P Global, industry analysts and participating utilities
In this scenario, Big Tech, utilities and regulators roughly keep pace with AI-driven load growth. Multi-gigawatt utility–cloud partnerships like NextEra–Google become standard, bundling data center campuses with on-site or contracted renewables, nuclear and gas, plus major transmission upgrades. IEA’s projections of data centre demand more than doubling by 2030 prove accurate but not catastrophic, because U.S. regions invest heavily in new capacity and grid reinforcements, and AI-based tools improve forecasting and outage prevention. Corporate clean-energy procurement continues to grow, maintaining pressure for low-carbon supply. Electricity prices rise but remain politically manageable, and outright moratoria on AI data centers remain limited or temporary.
Grid Bottlenecks and Local Backlash: AI Data Center Expansion Slows
Discussed by: Environmental groups, local governments in Virginia and Pennsylvania, investigative and climate-focused media
Here, local grid and land-use constraints, combined with rising retail bills, lead to growing resistance against large data center projects. The wave of opposition seen in Virginia, New Jersey, Georgia and Pennsylvania spreads, and the sort of moratorium environmental groups are now demanding gains traction at state or regional levels. Grid operators like PJM and ERCOT become more selective about interconnecting large loads, potentially deferring or denying some projects. Watchdogs’ complaints that data center approvals threaten reliability prompt FERC or state commissions to impose stricter criteria, slowing the pipeline and driving some AI investments overseas or into regions with more accommodating policies and excess power (for example, certain hydropower-heavy regions).
Reliability Crisis Trigger: Major Outage Linked to Data Center Clusters
Discussed by: NERC reliability reports, some grid planners, critical trade press and policy think tanks
In a more adverse outcome, a major region experiences a widespread blackout or near-miss event that investigators tie partly to the sudden disconnection or tripping of large data center loads, akin to the 1,500 MW load loss in a 2024 incident cited by NERC. Under stress from extreme weather and insufficient firm capacity, grid operators find that the concentration of AI data centers amplifies disturbances. The episode triggers emergency reliability standards: stricter ride-through and backup requirements for data centers, mandatory demand-response participation, and maybe hard caps on large-load interconnections in constrained zones. Project timelines stretch, and financing costs rise as investors reassess risk.
Nuclear and Firm Clean Power Renaissance Driven by AI
Discussed by: Corporate energy teams at Microsoft, Amazon, Meta, nuclear industry groups, some policymakers
In this path, AI data centers become the political and financial engine for a large-scale revival of nuclear and other firm clean power (advanced geothermal, long-duration storage). Deals like Microsoft’s Three Mile Island PPA, Meta’s Clinton contract, Amazon’s SMR agreements and NextEra’s nuclear license renewals proliferate, giving developers long-term revenue certainty. SMRs begin to come online in the early 2030s, often sited near data center clusters, and dormant or at-risk reactors are restarted with tech-company backing. Emissions impacts of AI are effectively offset, but public debate over nuclear waste, safety, and cost intensifies, and environmental groups fracture between those prioritizing climate goals and those opposing nuclear expansion.
Efficiency and Flexibility Tame the Beast: AI Demand Grows, Peaks Stabilize
Discussed by: Some IEA scenarios, chipmakers, hardware and software efficiency researchers
In a more optimistic scenario, rapid improvements in AI hardware efficiency, model architectures and workload scheduling drastically reduce per-inference and per-training energy costs. Coupled with more sophisticated demand response—data centers shifting non-urgent compute to times and locations with surplus renewable generation—the share of data centres in peak demand stabilizes even as total computing grows. IEA’s “High Efficiency” and similar scenarios materialize, limiting data centre demand’s share of global electricity to around 3–4% and easing grid-strain fears, though not eliminating local hotspot issues.
Historical Context
Cryptomining Moratoria and Grid Pushback in New York
2021–2024What Happened
Amid a boom in energy-intensive bitcoin mining, New York legislators passed Assembly Bill A7389B and Gov. Kathy Hochul signed a 2022 law imposing a two-year moratorium on new or renewed air permits for proof-of-work cryptomining at fossil-fueled power plants. The law required a statewide environmental impact study of cryptomining’s emissions and grid impacts. Other states and federal agencies also began scrutinizing mining’s electricity use, with DOE attempting an emergency survey and environmental groups urging stricter controls.
Outcome
Short term: Several fossil-fueled mining projects were effectively frozen, investors shifted some operations to other states, and New York became a test case for using moratoria to manage high-energy digital industries.
Long term: The experience created a policy template—temporary moratoria plus impact studies—that environmental advocates now propose applying to AI data centres, and it showed that local resistance can meaningfully redirect digital infrastructure siting.
Why It's Relevant
The cryptomining precedent demonstrates that U.S. policymakers are willing to pause digital infrastructure when they judge its energy use incompatible with climate and grid goals. Current calls for a nationwide AI data center moratorium echo this approach, suggesting that similar tools could be used if AI-driven power demand triggers political or reliability crises.
The California Electricity Crisis and Silicon Valley Power Cuts
2000–2001What Happened
California’s partial deregulation of electricity markets in the late 1990s, combined with tight supply and market manipulation, led to soaring wholesale prices and rolling blackouts in 2000–2001. Silicon Valley, then in the midst of the dot‑com boom, experienced costly outages that forced companies like Apple, AMD and Oracle to temporarily shut down operations or invest in backup systems. The crisis exposed the vulnerability of high-tech industries to grid instability and underinvestment in generation and transmission.
Outcome
Short term: Blackouts, bankruptcies (notably PG&E), emergency state interventions, and reputational damage to deregulation led to higher retail prices and a wave of investments in generation and reliability.
Long term: California tightened oversight, diversified supply, and aggressively pursued efficiency and renewables, while tech companies increasingly invested in onsite backup and, later, long-term PPAs to hedge grid risk.
Why It's Relevant
The California crisis shows how concentrated high-tech loads in regions with stressed power systems can amplify the economic cost of reliability failures. As AI data centers cluster in PJM, ERCOT and other regions already facing weather and capacity stress, the California experience warns that market design, reserve margins and transmission planning must anticipate these new loads to avoid similar crises.
Mid‑Century Aluminum Smelters and the Hydropower Buildout
1940s–1970sWhat Happened
During and after World War II, the aluminum industry—highly electricity-intensive—clustered around new federal hydropower projects in the Pacific Northwest and Tennessee Valley. Agencies like Bonneville Power Administration (BPA) and TVA built dams and transmission partly to serve smelters whose loads matched that of mid-sized cities. At one point, about 40% of U.S. aluminum smelting capacity and a large share of BPA’s sales came from these plants, which received low-cost power in exchange for providing steady baseload demand and, at times, interruptible load.
Outcome
Short term: The hydropower–aluminum bargain accelerated regional industrialization, supported war production, and enabled low retail rates for households across the Northwest and TVA regions.
Long term: Over time, rising power prices, environmental constraints and global competition reduced smelter viability, leaving BPA and others to rebalance their portfolios. The episode showed both the benefits and risks of designing public power systems around a few large industrial customers.
Why It's Relevant
AI data centers play a role today similar to mid‑century aluminum smelters: large, often geographically concentrated loads that can justify major generation and transmission investments but also risk distorting power systems if policy and contracts are misaligned. The hydropower–aluminum history suggests that dedicating substantial public or quasi‑public infrastructure to single industries requires careful long-term planning, including exit options and protections for residential and small-business customers.