For decades, the U.S. regulated utility sector, often tracked by the Utilities Select Sector SPDR Fund (XLU), maintained a reputation as a defensive investment. These entities, providing essential services like water, gas, and electricity, were characterized by stable, regulated revenues, low volatility, and predictable dividends, making them a preferred "bond-proxy" during periods of economic uncertainty and favorable interest rate environments.
However, the rapid and unprecedented infrastructure buildout required to support Artificial Intelligence (AI) has structurally undermined this defensive profile. The AI imperative has catalyzed a fundamental paradigm shift, re-rating the sector from a low-growth, defensive holding to a high-CapEx, high-growth infrastructure play, drawing historical comparisons to foundational buildouts such as the transcontinental railways and the internet backbone.
This transformation is evident in recent market performance. The S&P 500 Utilities Index delivered a year-to-date return of 16.2%, narrowly surpassing the broader S&P 500’s return of 15.8% as the year draws to a close. More dramatically, the sector achieved a 60% total return from its October 2023 low through June 2025, significantly outpacing the Morningstar US Index.
This rally introduces critical valuation pressures. Market participation has driven the sector to trade at a premium, registering 7% above its fair value with a price-to-fair value ratio of 1.07, according to Morningstar data. Investors appear to have chased utilities as a cheaper, second-derivative play on AI expansion, betting on sustained earnings acceleration. Critically, this upward valuation adjustment is occurring despite rising interest rates, which traditionally pressure utility performance as dividend yields become less attractive relative to Treasury benchmarks. This suggests the market is prioritizing the potential for long-term, structural earnings growth driven by CapEx over historical interest rate sensitivity.
The primary risk introduced by this high-growth cycle is execution risk. The sector faces massive challenges related to securing timely regulatory approval for soaring capital expenditures (CapEx), managing lengthy interconnection queues, and navigating intense political pressure regarding consumer affordability and cost allocation.
The Structural Demand Drivers: Quantifying the AI Imperative
The core of the utility sector’s re-rating is the verifiable, unprecedented acceleration in U.S. electricity demand, driven overwhelmingly by the power requirements of AI data centers.
Data Center Load Forecasts: Synthesis of IEA, DOE, and Industry Projections
Consensus estimates from government agencies and investment banks indicate a profound and immediate surge in power demand. The U.S. Department of Energy (DOE), through Lawrence Berkeley National Laboratory (LBNL), estimates that U.S. data center electricity demand could double or triple by 2028. The sector’s share of total U.S. electricity consumption, which stood at approximately 4.4% in 2023 (176 TWh), is projected to rise dramatically, reaching between 6.7% and 12% by 2028.
Leading financial firms concur with this aggressive trajectory. Goldman Sachs (GS) Research projects that overall data center power demand will surge by 165% to 175% compared to 2023 levels by the year 2030, an increase equivalent to adding a country that ranks among the top ten global power consumers. Long-term forecasts extend this growth, with Deloitte projecting U.S. peak electricity demand to grow by roughly 26% by 2035, driven significantly by data centers, which could alone reach 176 gigawatts (GW) of demand capacity by that year—a more than fivefold jump from 2024 levels. Meeting the near-term requirements of this AI infrastructure expansion necessitates over 50 GW of incremental capacity by 2028.
The following table summarizes the key projections underpinning this growth thesis:
Table 1: U.S. Data Center Electricity Demand Forecasts
| Source | Metric | 2023 Baseline | 2028/2030 Projection | Implied Growth Rate |
|---|---|---|---|---|
| DOE/LBNL | Share of U.S. Electricity Use | 4.4% (176 TWh) | 6.7% - 12.0% (by 2028) | Doubling to Tripling |
| Goldman Sachs | Change in Data Center Power Demand | N/A | 165% - 175% (by 2030) | Substantial Surge |
| BCG | U.S. Data Center Share | N/A | 7.5% (by 2030) | Accelerated Growth |
| Deloitte | Peak Demand Capacity (GW) | 25 GW (2024) | 176 GW (by 2035) | Fivefold+ Jump |
Intersecting Growth Vectors and Geographical Concentration
The AI surge is not isolated; it is compounded by other secular demand drivers that began accelerating load forecasts in 2025. These include widespread electrification in transportation and industry, industrial reshoring initiatives, and the broader deployment of electric vehicles and heat pumps. Together, these vectors are ushering in the biggest utility growth cycle in decades. The intensity of this demand is reflected in capital markets; data center construction spending has recently surpassed construction spending on all other U.S. commercial buildings, confirming the massive scale of this infrastructure arms race.
A crucial element often overlooked is the geographic concentration of this demand. The grid challenge is highly localized and acute, primarily focused in regions such as the Mid-Atlantic and Texas. For example, the load forecast from the Electric Reliability Council of Texas (ERCOT) indicates that if all data centers currently seeking connection were built, they would consume 78 GW of power during peak hours—nearly equivalent to the entire ERCOT grid's consumption today. This regional intensity necessitates immediate, targeted CapEx within specific utility service territories, creating concentrated risk and reward profiles for operators in these regions.
The Bullish Counterpoint: AI-Driven Efficiencies and Demand Mitigation
While the scale of power demand is daunting, a counterpoint exists regarding potential long-term mitigation. The International Energy Agency (IEA) and analyses by firms like Morgan Stanley (MS) suggest that technological innovation and efficiency improvements may help curb future consumption growth.
Specifically, efficiency gains in hardware and smarter model designs, coupled with more effective cooling systems, are expected to mitigate the overall strain. Notably, AI itself is proving to be a powerful tool for energy management. One leading IT company utilized AI to reduce the energy required to cool its data centers by approximately 40% simply by optimizing temperature and airflow. Furthermore, efficiencies unlocked by AI in non-data center use cases—such as building energy management systems, which represent nearly 40% of U.S. energy usage—could potentially offset much of the computational power consumed by AI itself. While these future efficiencies are a long-term mitigating factor, the near-term strain on grid capacity remains undeniable and mandates immediate CapEx.
Financial and Infrastructure Implications: The CapEx Supercycle and Reliability Crisis
The structural demand acceleration translates directly into the largest CapEx cycle the utility sector has experienced in recent history, but capitalizing on this opportunity requires overcoming severe infrastructure and financing constraints.
Modeling the CapEx Requirements and Rate Base Growth
The unprecedented demand accelerates the need for new generation, transmission, and distribution upgrades, positioning regulated utilities for massive rate base expansion. This anticipated CapEx growth is the primary rationale for the market’s positive re-rating. Analysts model sustainable earnings growth, with Morningstar projecting at least 5% annual earnings growth for many regulated operators through 2030. This sustained growth trajectory is attainable, provided utilities secure timely and comprehensive cost recovery through the regulatory process.
Grid Constraint Deep Dive: The Interconnection Queue Bottleneck
The supply side of the equation lags significantly behind the accelerating demand. Over 2.6 terawatts (TW) of mainly clean energy projects—solar, wind, and battery storage—are currently gridlocked in interconnection queues across the U.S., delaying the urgently needed addition of new supply. This bottleneck demonstrates that high load demand does not automatically translate into high operational supply due to administrative and infrastructural constraints in the transmission system. Utilities and developers must undergo lengthy impact studies to secure connection, a process that often results in project withdrawal or substantial delays, highlighting a major infrastructure vulnerability that impedes the realization of projected CapEx returns.
The Reliability Imperative and the Nuclear Advantage
The AI infrastructure requires not just vast power, but firm, reliable, round-the-clock baseload capacity. Data centers are highly sensitive to grid disruptions; a transmission line fault in Virginia’s "Data Center Alley" in 2024 caused 1,500 MW of data center load—equivalent to three large power plants—to suddenly disconnect.
The current capacity deficit is most pronounced in firm baseload power. While 209 GW of new capacity is projected by 2030, only 10% of those additions will be firm, widening the reliability gap.
Consequently, there is a premium on generation sources that can deliver reliability and high volume. Nuclear power provides the ideal solution: clean, reliable, scalable, and dispatchable. Merchant generators like Constellation Energy (CEG) have aggressively capitalized on this demand, securing long-term Power Purchase Agreements (PPAs) with hyperscalers. High-profile agreements include Meta Platforms (META) signing a 20-year nuclear PPA to extend the operations of the Clinton Clean Energy Center, and Microsoft Corp (MSFT) entering a 20-year PPA with Constellation to restart the 835 MW Three Mile Island Unit 1 nuclear reactor by 2028.
The reliance on nuclear PPAs by hyperscalers underscores that "clean" is often secondary to "firm" in the near-term power procurement strategy for energy-intensive AI workloads. The intermittency of renewable energy, despite making up 93% of new capacity additions through mid-2025, does not meet the 24/7 reliability mandate of high-density AI computing. This places a strategic value on existing, dispatchable generation assets and utilities positioned to leverage them.
Cost of Capital (WACC) Sensitivity and Interest Rate Headwinds
The massive CapEx associated with the AI buildout requires substantial debt financing, making the sector acutely vulnerable to the cost of capital. Higher interest rates increase debt costs, which compress allowable returns (Return on Equity, ROE), and increase the overall Weighted Average Cost of Capital (WACC).
The fundamental financial execution challenge for regulated utilities is securing timely rate cases that allow them to recover soaring capital expenditures and ensure regulators permit necessary adjustments to the allowed ROE. Failure to adjust for higher WACC through the regulatory process results in regulatory lag and potential cost disallowance, which directly threaten the attainment of bullish earnings forecasts.
Regulatory Risk and Cost Allocation Frameworks: The Political Friction
While demand is a structural bull case, the primary financial risk facing regulated utilities remains regulatory execution—specifically, securing cost recovery while addressing public concerns over consumer affordability.
The Ratepayer vs. Hyperscaler Subsidy Debate
The concurrent rise in retail electricity prices—projected to be approximately 4.5% higher in 2025 compared to 2024—is directly tied to the infrastructure investments needed to serve data centers. This has fueled political and regulatory scrutiny over who bears the burden of these costs.
Historically, cost allocation methodologies have often spread infrastructure costs across the entire customer base, leading to the socialization of costs. This means residential and other non-industrial customers effectively subsidize the substantial new infrastructure required by wealthy tech companies, a practice drawing fire from consumer protection groups. The emerging political consensus in the U.S. is that large-load customers should pay for the infrastructure built specifically to serve them, a shift that requires radical changes to current regulatory compacts.
Case Study: Dominion Energy (D) and Virginia's Data Center Alley
Dominion Energy, the utility serving the largest concentration of data centers globally in Northern Virginia, exemplifies the acute challenges utilities face. Dominion reported a massive pipeline of over 40,000 MW of data center capacity in various stages of contracting—the equivalent of powering 10 million homes. The energy demand of data centers in this region had already expanded eightfold between 2013 and 2024.
In response to public pressure and growing evidence that data centers were receiving a favorable deal, the Virginia State Corporation Commission (SCC) issued a final order in Dominion’s rate case. This order, while approving a substantial rate increase, mandated significant changes aimed at mitigating burdens on residential customers. Critically, the SCC approved a new rate class for large energy users with much stronger financial requirements and required Dominion to shift away from its current cost allocation methodology. This regulatory intervention establishes a precedent for other jurisdictions grappling with concentrated demand.
Mitigating Execution and Financial Risk
To protect ratepayers and financial integrity, states are rapidly evolving their regulatory frameworks. States like Pennsylvania are pursuing legislation to mandate Cost-Responsibility Contracts (CRCs). These contracts require large-load facilities to sign long-term service agreements (no less than 10 years) to ensure they cover the full marginal infrastructure costs they generate. The Pennsylvania Public Utility Commission (PA PUC) is also being empowered to set cost-allocation rules, monitor compliance, and protect non-industrial customers.
For utilities, this shift introduces complex financial counterparty risks. Utilities must secure robust mechanisms, such as letters of credit or cash deposits, to cover Capital Contributions in Aid of Construction (CIAC) amounts, along with potential tax gross-up liabilities to safeguard against cost socialization. Hyperscalers often prefer Parent Company Guarantees, which place an additional administrative burden on utilities to conduct sophisticated credit risk assessments of the parent entity, especially considering commitments made across multiple jurisdictions. The success of future regulated EPS growth models hinges not just on construction execution, but on the utility’s ability to transition into a sophisticated financial manager capable of assessing and securing the long-term creditworthiness of its largest customers.
Investment Thesis and Peer Group Analysis
The AI-driven structural shifts necessitate a segmentation of the utility investment landscape based on regulatory exposure, generation mix, and geographic concentration.
High-Beta Exposure: Concentration and Regulatory Test Case (Dominion Energy)
Dominion Energy (D) offers the most direct and highest-beta exposure to the AI theme. The company has aggressively contracted capacity, nearly doubling its data center power capacity under contract in the second half of 2024. This sheer scale of committed load positions D for massive rate base expansion.
However, this concentration makes D highly susceptible to regulatory headwinds. The utility serves as the test case for the success of the Virginia SCC's new cost allocation regime. The high-risk/high-reward profile of D is directly tied to its ability to manage regulatory lag, avoid cost disallowance, and successfully implement financial security mechanisms for the 40 GW contracted pipeline. Other regulated utilities with strong positions in rapidly growing load regions include Entergy (ETR) and Southern Co. (SO).
The Pure-Play Firm Power Provider (Constellation Energy)
Constellation Energy (CEG) is strategically positioned as a merchant generator capitalizing on the critical demand for firm, clean baseload power. By leveraging its existing fleet of conventional nuclear capacity, CEG has bypassed the often-sluggish interconnection queue bottlenecks and complex rate recovery hurdles inherent to regulated utilities.
CEG's strategy focuses on securing long-term PPAs with high-credit-quality counterparties like Microsoft and Meta. This differentiated model offers a cleaner investment thesis on AI energy demand, transferring infrastructure development and financing risk largely to the hyperscalers while securing guaranteed, long-term revenue streams for existing assets.
Diversified Growth and Renewable Scale (NextEra Energy & Duke Energy)
Companies like NextEra Energy (NEE), a top holding in XLU, and Duke Energy (DUK) offer exposure that is diversified across both regulated operations (Florida Power & Light for NEE) and large-scale renewable energy and transmission development.
Their thesis is built on capturing the benefits of broad structural electrification, industrial reshoring, and AI demand spread across larger geographic footprints. This diversification helps mitigate the specific, localized regulatory risks seen in highly concentrated areas like Northern Virginia, offering a lower-beta approach to the infrastructure supercycle.
Table 2: Key Utility Exposure and Strategic Response to AI Demand
| Utility | Primary Exposure Area | Key Strategic Asset | Key Risk |
|---|---|---|---|
| Dominion Energy (D) | Northern Virginia (Data Center Alley) | 40 GW contracted capacity pipeline | Severe regulatory lag & cost disallowance risk (SCC scrutiny) |
| Constellation Energy (CEG) | Merchant Nuclear Generation | Long-term PPAs (Meta, MSFT) for clean, firm baseload power | Exposure to competitive merchant power markets and contract renewal risk |
| Southern Company (SO) | Southeast Load Growth (GA, AL) | Strong regulated service territories; diversified industrial load base | High CapEx requirements; regulatory pressure on affordability |
| NextEra Energy (NEE) | Florida Power & Light (FPL) & NEER | Scale in renewables and transmission infrastructure development | Interconnection complexity; potential for regulatory lag in regulated FPL segment |
Conclusion and Forward-Looking Recommendations
The U.S. utility sector is no longer defined by its defensive characteristics. The structural demand introduced by AI has permanently shifted its risk/reward profile, replacing historical stability with accelerated growth potential and corresponding execution risk. The sector is now firmly established in its largest growth cycle in decades.
The market’s current valuation premium (7% above fair value) is justified by the visibility into sustained, above-average EPS growth through 2030, driven by mandated CapEx spending. However, this growth trajectory is fraught with operational and political complexity.
Future alpha generation will be determined not solely by geographic load exposure, but by management’s superior skill in two critical areas:
- Regulatory Navigation and Cost Recovery: Investment favorability will shift toward utilities operating in regulatory environments that proactively approve Cost-Responsibility Contracts and allow for timely adjustments to the allowed ROE, thereby minimizing regulatory lag and avoiding cost socialization.
- Infrastructure Execution and Reliability: Priority must be given to utilities that possess or can rapidly deploy firm generation capacity—especially nuclear or natural gas—to bypass the existing multi-year interconnection queue bottleneck. Utilities must also demonstrate the financial sophistication required to manage the elevated credit and counterparty risk associated with massive hyperscaler contracts.
The long-term viability of the AI infrastructure thesis hinges on the successful transition of regulated utilities from operators focused on maintenance to high-volume capital executors who can successfully de-risk the massive financial undertaking while satisfying both hyperscaler demand and regulatory affordability mandates.
Source
- J.P. Morgan Wealth Management (via Chase) - What to consider when investing in utilities - Updated Mar 10, 2025
- Congress.gov (CRS) - Report: U.S. Data Center Energy Use Projections - Context: Data center energy use projections through 2028
- Deloitte - Power and Utilities Industry Outlook - Context: 2026 Outlook (referencing 2025 trends)
- U.S. Department of Energy (DOE) - Domestic Energy Usage from Data Centers Expected to Double or Triple by 2028 - December 20, 2024
- Department of Energy (EERE) - Queued Up: Characteristics of Power Plants Seeking Transmission Interconnection - As of the End of 2021 (Report)
- Fidelity - AI outlook (Long-term infrastructure buildout) - Context: Fidelity Asset Allocation Research Team estimates
- Goldman Sachs Research - AI to drive 165% increase in data center power demand by 2030 - Context: GS SUSTAIN Report Summary
- International Energy Agency (IEA) - Energy and AI Report - Context: Global data center energy demand forecast through 2035
- Morgan Stanley - Can AI decrease energy usage? (AI Might Not Fry the Grid) - Context: Analysis on efficiency gains and demand mitigation
- Morningstar / MarketWatch - AI Needs Power Desperately: Here's How to Invest in Companies Profiting from the Pain - December 3, 2025
- Morningstar - 4 Utility Stocks to Benefit from Data Center Growth - May 2, 2024
- NRG Energy - Comments on Virginia Commission Data Center Technical Conference (Filing Excerpt) - Context: Regarding Dominion/Rappahannock load issues
- Pennsylvania Senate (Co-Sponsorship Memo) - Legislation Mandating Cost-Responsibility Contracts for Data Centers - Context: Memo on proposed PA legislation
- Pennsylvania Public Utility Commission (PA PUC) - Tax Gross-up and Security Standards for Large-Load Interconnections (EAP Recommendation) - Context: Commission Filing
- Charles Schwab - Utilities Lose Defensive Touch as AI Ignites Rally - Context: Analysis of 2025 market rally
- State Street Global Advisors (SSGA) - Powering the AI Economy: Utilities Enter Their Biggest Growth Cycle in Decades - November 26, 2025
- World Economic Forum (WEF) - How Data Centres Challenge the Electricity Regulatory Model - July 2025
