The Invisible Load: How AI Is Rewriting America’s Energy Future
For most of the past two decades, US electricity demand was essentially flat. Improvements in energy efficiency — better appliances, LED lighting, more efficient industrial processes — roughly offset the growth in population and economic output, leaving total electricity consumption in 2020 close to levels seen in 2007. Grid operators, utilities, and regulators had built an entire planning framework around this assumption of stagnant demand.
That assumption is now obsolete. The explosive growth of artificial intelligence infrastructure — primarily in the form of large-scale data centres housing thousands of energy-intensive graphics processing units (GPUs) — has fundamentally changed the electricity demand outlook for the United States. The numbers are startling in their scale and in the speed with which they have materialised.
The Data: From Flat to Surging
According to analysis from S&P Global, US data centre demand for grid power rose by 22% in 2025 alone. By 2026, total US data centre power demand — encompassing IT equipment, cooling systems, lighting, and power conversion losses — is projected to reach 75.8 gigawatts (GW). To put this in perspective, 75.8 GW is roughly equivalent to the entire generating capacity of France.
Looking further ahead, the trajectory is even more striking. BloombergNEF projects US data centre power demand could reach 106 GW by 2035. S&P Global projects demand nearly tripling from 2025 levels by 2030. The total amount of five-year summer peak demand growth forecasts published by US utilities jumped from 38 GW in 2023 to 128 GW in 2024 — a more than threefold increase in projected demand growth in a single year, driven almost entirely by data centre and AI workload projections.
The US Energy Information Administration (EIA) reported that US retail electricity prices rose by almost 7% in 2025 compared to 2024 — double the rate of overall inflation. While this increase reflects multiple factors including fuel costs and grid investment, the data centre demand surge is increasingly recognised as a structural driver of long-term electricity price pressure.
Where the Demand Is Concentrated
Data centre demand is not evenly distributed across the United States. Two states — Virginia and Texas — account for the largest concentrations of data centre load in the country. Virginia’s data centre demand for grid power was forecast to reach approximately 12.1 GW in 2025, driven by the Northern Virginia corridor near Dulles Airport, which has become the largest data centre market in the world. Texas utility power demand from data centres was hitting approximately 9.7 GW over the same period.
Other significant concentration points include Georgia (particularly the Atlanta metropolitan area), Oregon and Washington state (home to major hyperscaler facilities from Amazon, Microsoft, and Google), and Arizona and Nevada (favoured for their land costs, power availability, and climate for cooling).
This geographic concentration creates significant localised grid stress. The PJM Interconnection — the grid operator serving the mid-Atlantic states including Virginia, Maryland, Pennsylvania, New Jersey, Delaware, and parts of the Midwest — has warned that it could face 31 GW of new data centre load over the next five years, approximately 3 GW more than projected capacity additions from new generation in its territory. This creates a structural supply-demand imbalance that threatens grid reliability.
Grid Reliability: The Warning Signs
The North American Electric Reliability Corporation (NERC) — the body responsible for monitoring and enforcing grid reliability standards across North America — has issued increasingly urgent warnings. In its most recent assessment, NERC warned of “elevated risk” of summer electricity shortfalls in 2026 and beyond in major grid regions. The concern is not that blackouts are inevitable, but that the margin of safety — the buffer of spare generating capacity that grid operators rely on to handle unexpected demand spikes or generator failures — is eroding faster than new capacity can be built.
Over 8.9 GW of new data centre projects across 105 developments are targeting operation by the end of 2026, with 47 already under construction. Each of these facilities will require reliable, around-the-clock power — data centres typically operate at high capacity factors of 90% or above, compared to renewable generation sources that operate intermittently. This creates particular challenges for grid operators who must ensure baseload power availability at all hours.
The irony of the data centre boom is that it is partly undermining the economics of the renewable energy transition. Because data centres require firm, constant power and cannot run on intermittent solar or wind alone (without massive battery storage backing), their growth has prompted utilities to extend the operational life of coal and natural gas power plants that might otherwise have retired. Analysis suggests that data centre growth could drive US electricity prices up by as much as 25% in some markets by 2030 — a significant headwind for household and business energy costs.
For comprehensive coverage of electricity prices and market developments across the US and globally, visit our electricity section.
The Supply Response: Record Capacity Additions
The power generation industry is responding to the demand signal. The EIA projects that US power plant developers and operators plan to add 86 GW of new utility-scale electric generating capacity to the US power grid in 2026 — a record if realised, and more than double the additions of just five years ago.
The composition of these additions tells its own story about the energy transition. Solar power accounts for 51% of planned 2026 capacity additions, followed by battery storage at 28% and wind at 14%. Essentially all — more than 99% — of new US capacity additions in 2026 will come from solar, wind, and storage, with natural gas and other fossil fuel sources playing almost no role in the construction pipeline.
However, capacity alone does not solve the data centre reliability challenge. Solar generates power only during daylight hours; wind is variable. Battery storage can shift solar generation into evening hours but currently operates economically at four- to eight-hour durations. Meeting the 24/7 firm power requirements of large data centres requires either very large battery storage systems, demand flexibility programmes, or the continued operation of dispatchable generation sources including natural gas.
The Tech Giants’ Energy Commitments
The major technology companies driving data centre demand — Microsoft, Google, Amazon, Meta, Apple — have all made significant public commitments to power their operations with 100% renewable energy. In practice, this typically means purchasing renewable energy certificates (RECs) or entering into power purchase agreements (PPAs) with renewable generators, rather than running exclusively on renewable power at every hour of the day.
Some companies are going further. Microsoft has signed agreements with nuclear energy developers and is exploring advanced fission and fusion technologies as potential sources of firm, low-carbon power for data centres. Google has partnered with nuclear developers to explore small modular reactor (SMR) options. The interest in nuclear power from the technology sector — once predominantly seen as a renewable energy champion — reflects the practical reality that meeting 24/7 clean power requirements for massive data centres is extremely difficult with solar and wind alone.
What This Means for Your Electricity Bill
For ordinary US households, the data centre electricity boom has complex implications. In the near term, the surge in utility capital expenditure to build grid infrastructure for data centres — new transmission lines, substations, transformers, and generation — is being socialised across all ratepayers, contributing to higher electricity bills even for consumers who will never directly benefit from AI services.
In the longer term, the additional demand could help spread fixed grid costs across a larger consumer base, potentially moderating per-unit price increases. Large data centre operators also generate significant economic activity and tax revenues that benefit local communities. The net effect on electricity prices will depend critically on how efficiently the grid expands to meet the new demand and whether competitive pressure in power markets is maintained.
For practical guidance on managing electricity costs in a rising-price environment, visit our how to save on energy section. And for the latest energy news including data centre developments, nuclear power agreements, and grid investment, visit our energy news section.
The US electricity system is undergoing its most significant structural transformation in generations. The question is not whether AI will reshape power markets — it already is — but whether grid infrastructure, regulatory frameworks, and clean energy supply can keep pace with one of the fastest demand growth cycles the sector has ever seen.
