POWER USE · AI · UPDATED SEPTEMBER 13, 2026

AI and data-center electricity: what the numbers show

Electricity use by AI-focused data centers grew 50% in 2025, far faster than the 17% growth of all data centers, according to the International Energy Agency. In its projections, AI-focused consumption triples between 2025 and 2030.

• +50%Growth in electricity use of AI-focused data centers, 2025IEA
• 3×Projected growth in AI-focused data-center electricity, 2025–2030IEA
• 100,000Households’ worth of electricity used by a typical AI-focused data centerIEA
• 11×Increase in AI server power density, 2020–2025IEA

What is actually measured

AI electricity use is hard to isolate. The IEA says there are “no comprehensive statistics on the frequency and depth of AI usage around the world.” What exists are estimates of data-center consumption, company-reported totals and tracking of new facilities. This guide sticks to those published figures.

Bigger, denser buildings

Power density is climbing fast. AI server power density rose 11 times between 2020 and 2025, and the IEA expects a further fourfold increase by 2027. By then, a single server rack “only the size of a large refrigerator” could have peak power demand equal to 65 households.

The load is also unusually spiky. The IEA reports that AI data centers see repeated swings in server load of more than 50% of rated capacity within a second. That makes energy storage critical, and the IEA estimates that 20–25 GW of battery storage could be installed in data centers by 2030.

Capacity is growing quickly. The IEA’s satellite-based tracking shows that “AI factories”, data centers built specifically for AI, more than tripled in capacity in the past 18 months, according to its 2026 report.

Efficiency per task is improving

Consumption is rising even as individual tasks get cheaper. The IEA says energy use per AI task has dropped by at least an order of magnitude a year recently, and that simple text queries now typically use less electricity than running a television for the same amount of time. Total demand still rises because use is growing: major model providers reported a threefold increase in active users and a fivefold increase in revenue over the past year, according to the IEA.

• 10×+Annual drop in energy per AI task in recent yearsIEA
• 30% a yearProjected growth in accelerated-server electricity useIEA
• ~50%Accelerated servers’ share of the projected net increase in data-center electricityIEA

Money, supply chains and power

Five large technology companies had capital expenditure of more than USD 400 billion in 2025, expected to rise by another 75% in 2026. The IEA reports supply bottlenecks: a 70% surge in gas turbine orders in 2025, and a shortage of high-bandwidth memory, used in AI chips, expected to last until at least the end of 2027.

• ~40%Tech sector’s share of corporate renewable PPAs signed in 2025IEA
• 45 GWConditional offtake agreements with small modular reactor projects (25 GW at end-2024)IEA
• 15–27 GWOnsite natural gas that may power data centers by 2030IEA
• ~350 MtProjected data-center emissions in 2035, about 2% of electricity-sector emissionsIEA

Conditional offtake agreements are not operating reactors. Onsite gas and battery figures are ranges from IEA projections, not installed capacity.