AI Data Centers Are Driving a Historic Surge in U.S. Electricity Demand

I have spent more than 30 years analyzing energy markets from Darien, Connecticut. In that time, I have watched oil price shocks reshape geopolitics, lived through the shale revolution, tracked the rise of renewables, and followed the slow grind of the energy transition from fossil fuels toward cleaner sources. None of those shifts changed the fundamental dynamics of U.S. electricity demand very much. For two decades, power consumption in this country was essentially flat, growing at well under one percent per year. That era is over.

Artificial intelligence and the data centers built to run it have put electricity demand on a trajectory I have not seen in my career. The U.S. Energy Information Administration projects total U.S. electricity consumption rising from roughly 4,110 billion kilowatt-hours in 2024 to more than 4,260 billion kilowatt-hours in 2026, with record-breaking demand now the baseline expectation rather than a tail risk. For anyone who analyzes energy equities or fixed income, this changes a lot of assumptions that have been baked into valuations for years.

What the Numbers Actually Show

The data center story is real, but it requires careful reading. Utility power provided to data centers rose roughly 22 percent in 2025, reaching about 61.8 gigawatts, according to 451 Research. Projections from the same source put that figure at 134.4 gigawatts by 2030. Deloitte’s 2026 Power and Utilities Outlook goes further, estimating that data center demand alone could reach 176 gigawatts by 2035, a fivefold increase from 2024 levels.

At the same time, it is worth being precise about what these numbers mean in practice. New data center deals actually fell more than 40 percent between the third and fourth quarters of 2025. Only about one-third of the 240 gigawatts of announced construction is actively being built. The headline numbers from hyperscaler announcements are often aspirational. The investment case for utility equities is not about taking every announced project at face value. It is about understanding the structural floor of demand growth, which remains substantial even under conservative assumptions.

That is the kind of nuance I try to bring to my work at E3 Research Associates. Anyone can read a headline about Amazon spending $85.8 billion in capital expenditures last year, up 78 percent year over year, or Microsoft committing over $80 billion in 2025 for AI infrastructure. The harder analytical job is tracing that spending through the power system and figuring out where real earnings growth lands.

What It Means for Utility Equities

Utilities have been a slow, predictable corner of the equity market for most of my career. Regulated earnings, predictable dividends, low beta. The sector attracted income-oriented investors and largely stayed out of the growth narrative. That characterization is shifting quickly.

Electric and gas utility capital expenditures are projected to jump 22 percent year over year to $212 billion in 2025, according to Deloitte, well above the 7.6 percent compound annual growth rate of the past decade. The electric power sector needs more than $1.4 trillion in cumulative investment through 2030 to meet rising demand while managing the ongoing transition away from retiring coal and gas capacity. That capex cycle creates earnings visibility for utilities that have locked in data center interconnection agreements, and it creates credit stories for investors willing to look at the fixed income side of the capital structure.

The utilities best positioned are those with large service territories in states that have become data center hubs. Virginia and Texas lead in current data center load, with Virginia projected at roughly 12.1 gigawatts of data center demand in 2025, up from 9.3 gigawatts in 2024. But the geographic footprint is expanding. Georgia, Ohio, Arizona, Oregon, and Illinois are all seeing meaningful growth. Investors who look beyond the mega-cap technology names that dominate the AI narrative will find that some of the most durable beneficiaries are the regulated utilities providing the power those companies depend on.

The Grid Constraint Problem

Demand growth is only half the equation. The U.S. grid was not built for what is being asked of it right now. There are currently two terawatts of generation capacity stuck in interconnection queues, which is roughly equal to the entire currently installed capacity of the U.S. power system. Getting new supply online fast enough to meet demand is the central operational challenge for utilities, regulators, and grid operators through the end of this decade.

This matters for investors in two ways. First, the companies that can execute on grid expansion and interconnection faster than their peers will earn premium valuations over the next several years. Speed of execution has rarely differentiated utility valuations the way it may going forward. Second, transmission infrastructure spending is becoming its own investment category. Grid enhancement technologies, advanced conductors, and digital monitoring systems are attracting capital because they allow more power to move through existing lines without requiring decades-long permitting for new corridors.

The IEA’s Energy and AI report provides the clearest global framework I have seen for thinking through the long-range demand picture. It is worth reading carefully if you want to understand both the upside scenarios and the downside cases, including what happens if AI adoption is slower or more energy-efficient than current projections assume.

The Critical Minerals Connection

One piece of this story that does not get enough attention in the utilities conversation is what grid expansion requires on the supply chain side. Transformers, cables, conductors, and storage systems all depend on the same minerals that the clean energy build-out has been competing for. I wrote earlier about the unseen bottlenecks in the critical minerals supply chain and those constraints apply directly here. A utility that needs to triple its transformer procurement to meet data center load growth is operating in a market where transformer lead times have stretched to two or three years in some cases. Supply chain risk is a real factor in modeling utility earnings growth, and it is underweighted in most sell-side models I have reviewed.

The energy transition and the AI demand surge are not separate stories. They are competing for the same infrastructure, the same materials, and the same regulatory bandwidth. Navigating the energy transition as an investor now means understanding how AI demand interacts with decarbonization commitments, because utilities are trying to build clean capacity and firm capacity at the same time, on compressed timelines, with supply chains that were not designed for this pace.

Frequently Asked Questions