The future of artificial intelligence is often presented as a contest of algorithms, advanced chips and corporate ambition. Yet behind every breakthrough sits something less glamorous: an electricity connection. Before a data center can train a powerful model or serve millions of users, it needs dependable power, delivered where and when its operators require it.
That requirement is becoming a constraint on America’s technology expansion. A February 2026 policy brief from Harvard’s Belfer Center and engineering school found that, in some regions, AI-related electricity demand is already outpacing available capacity. Companies have responded by delaying projects, contracting directly with power producers or installing on-site natural gas generators. The challenge is no longer simply building more computing infrastructure. It is making sure that infrastructure can run.
A Digital Industry With a Physical Appetite
The scale of the shift is substantial. The International Energy Agency’s Energy and AI report projects that global electricity consumption by data centers will reach approximately 945 terawatt-hours by 2030, more than twice its estimated 2024 level. AI is a major force behind that increase, although the total also includes conventional computing, storage and other digital services.
Inside these facilities, electricity does more than power the processors performing calculations. It also supports networking equipment, storage systems and the cooling needed to keep hardware operating safely. According to the IEA, servers account for about 60% of electricity demand in modern data centers on average, while cooling’s share varies considerably with facility design and efficiency.
For the United States, the implications extend beyond individual corporate campuses. The IEA projects that data centers will account for almost half of the country’s electricity demand growth through 2030. That does not mean they will consume half of America’s electricity. It means they represent an unusually large share of the additional power the nation will need, placing technology companies increasingly at the center of energy planning.
Two Industries Running on Different Clocks
The tension begins with timing. A data center can become operational in two to three years, the IEA notes, while the energy infrastructure needed to support it generally requires longer planning and construction schedules. Technology companies can move quickly to secure equipment and develop campuses. Electricity networks cannot necessarily expand on the same timetable, particularly when projects require extensive investment and coordination.
Location adds another complication. National electricity totals can obscure the pressure created when large facilities cluster within a relatively small area. The IEA identifies this geographic concentration as a particular integration challenge. A country may have sufficient generation in aggregate while a specific region still struggles to accommodate a sudden increase in demand. For developers, access to usable local capacity matters more than an encouraging national statistic.
Northern Virginia illustrates how closely digital growth and grid operations have become intertwined. The Harvard brief describes a July 2024 voltage fluctuation that prompted 60 data centers to disconnect simultaneously, creating a 1,500-megawatt surplus and requiring emergency adjustments. The episode shows that reliability involves more than supplying enough electricity. Grid operators must also manage how large computing facilities respond when electrical conditions change.
The Question Behind Every New Connection
Expanding the grid raises a question that reaches far beyond Silicon Valley: who pays? Utilities may need substantial infrastructure investment to accommodate new computing demand. But Harvard’s researchers warn that if projected demand fails to materialize, utilities and their customers could face costs associated with infrastructure that is no longer fully needed. Building too little creates constraints; building too much creates financial exposure.
Forecasting is particularly difficult because future consumption depends on several moving targets. The IEA’s scenarios account for differences in AI adoption, hardware and software efficiency, and energy-sector bottlenecks. Its higher-efficiency scenario demonstrates how improvements could allow growing digital services to consume less electricity than they otherwise would. Demand projections therefore describe possible pathways, not guaranteed outcomes that utilities can safely treat as settled.
Regulators are beginning to address that uncertainty. Texas enacted Senate Bill 6 in 2025, introducing reforms covering large-load connections, cost sharing, transparency and emergency operations. As summarized in the Harvard brief, the law seeks clearer disclosures and a reasonable allocation of upgrade costs to new large users. Such measures reflect an evolving debate over how to support investment without leaving other electricity customers carrying disproportionate risk.
A Different Kind of Competitive Advantage
The response will require more than adding generating capacity. The IEA calls for faster investment in electricity networks alongside better data center efficiency, greater operational flexibility and closer cooperation between technology companies, energy providers and policymakers. Renewables and natural gas are expected to play leading roles in supplying growing demand, but additional generation must still be integrated into functioning electricity systems.
Flexibility could become especially valuable. Harvard researchers are examining how computing-job scheduling might help facilities adjust electricity consumption, including during periods of grid stress. The opportunity is promising, but practical limits matter. Not every workload can be delayed, and incentives must reflect what operators can reliably deliver rather than assuming all data center demand is equally adjustable.
The emerging competitive advantage is therefore not computing power alone, but coordination. The IEA’s analysis makes clear that AI expansion depends partly on aligning fast-moving digital investment with slower-moving energy infrastructure. For America’s technology industry, the next chapter will be shaped not just by what its machines can accomplish, but by whether the surrounding electricity system can support them.
