US electricity demand hits record highs as AI data centers surge
Power consumption is projected to climb through 2027, with data centers now consuming nearly 2% of global electricity generation.
The United States is experiencing unprecedented growth in electricity consumption, driven primarily by the rapid expansion of artificial intelligence infrastructure. Federal forecasts now project power demand will continue breaking records through 2027, creating urgent challenges for grid operators nationwide.
According to the US Energy Information Administration, total electricity use will climb from 4,195 billion kilowatt-hours in 2025 to 4,268 billion kWh in 2026 and 4,391 billion kWh in 2027. The surge marks the third consecutive year of record-breaking consumption, with AI-focused data centers and broader electrification trends as the primary drivers.
PJM Interconnection, the largest US grid operator serving roughly 20% of the population, is now developing proposals to address a widening gap between available electricity supply and accelerating demand. Data centers represent the fastest-growing load on the system.
Global data center consumption accelerates
The pressure extends well beyond US borders. A new report from the International Data Center Authority estimates that data centers worldwide now consume 67.7 gigawatts of electricity—approximately 1.9% of global generation—representing a 17% increase over the previous year. AI-specific facilities saw consumption jump 50% in 2025 alone.
The authority warns that current grid constraints "are only the beginning, as AI fundamentally alters the trajectory of digital infrastructure power consumption." Without new grid construction, data centers could consume 12% of US power by 2030, up from 6% in 2025.
Why it matters
The collision between AI's explosive growth and aging electricity infrastructure represents one of the most significant infrastructure challenges facing developed economies. Grid operators must now balance rising demand against limited transmission capacity, while policymakers face difficult decisions about prioritizing industrial loads, managing costs, and accelerating infrastructure investment. The timeline for grid expansion typically spans years or decades, while AI deployment is measured in months—a mismatch that could constrain both economic growth and technological advancement.
The dual role of AI in energy systems
Yet the relationship between AI and energy isn't purely extractive. Research from the World Economic Forum and Accenture describes an "AI-energy paradox": while AI increases electricity demand, it simultaneously offers tools to optimize energy systems, support renewable integration, and strengthen grid resilience.
Separate Forum research highlights how industrial competitiveness increasingly depends on data collaboration. Around 80% of industrial data in the European Union remains unused, but shared data frameworks could enable system-level optimization. At China's Ordos-Envision Net-Zero Industrial Park, data sharing across more than 50 enterprises has achieved approximately 80% on-site green electricity consumption while reducing overall energy costs by roughly 10%.
Heat waves compound grid stress
Extreme heat is adding further pressure to electricity systems. Romania recently began disconnecting its only operational nuclear reactor amid record-low Danube water levels, while Hungary and France also curtailed nuclear generation during recent heat waves. European spot power prices rose more than 20% during the most recent extreme heat period, though increased solar output helped meet higher demand.
These details were first reported by the World Economic Forum's Centre for Energy and Materials in its energy sector roundup.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
