Global data center electricity consumption will reach 565 TWh in 2026, up 26% from 447 TWh in 2025, according to Gartner’s June 2026 forecast. AI-optimized servers are the reason: their power draw is set to jump 84% this year. This page compiles AI energy consumption statistics verified against the IEA, Gartner, the U.S. Department of Energy, and Google’s own disclosures.
AI Energy Consumption Statistics: Key Notes
- Global data center electricity consumption is forecast at 565 TWh in 2026, up 26% from 447 TWh in 2025, per Gartner (June 2026).
- AI-optimized servers will draw 175 TWh in 2026 and surpass conventional servers by 2027, per Gartner (June 2026).
- The IEA projects data center electricity use of around 945 TWh by 2030, just under 3% of global consumption, in its Energy and AI base case (2025).
- A median Gemini Apps text prompt used 0.24 Wh of electricity, 0.26 ml of water, and emitted 0.03 gCO2e as of May 2025, per Google’s technical paper (August 2025).
- U.S. data centers could take 6.7% to 12% of national electricity by 2028, up from 4.4% in 2023, per the DOE’s LBNL report (December 2024).
The pattern across every dataset is the same: energy per AI task keeps falling while total AI energy consumption keeps rising, because usage grows faster than efficiency.
How Much Electricity Do AI Data Centers Use in 2026?
Gartner’s June 10, 2026 forecast puts worldwide data center electricity consumption at 565 TWh for the year, a 26% rise from 447 TWh in 2025.
Peak power demand climbs 27% to 132 GW in 2026, from 104 GW in 2025, and Gartner expects 290 GW by 2030. The firm also projects consumption above 1,200 TWh by 2030 and names power availability, not chips, as the constraint on AI capacity.
| Metric | 2025 | 2026 (forecast) |
|---|---|---|
| Data center electricity consumption | 447 TWh | 565 TWh |
| Data center power demand | 104 GW | 132 GW |
| AI-optimized server share of data center power | ~20% | 31% |
Source: Gartner press release, June 2026
Within the 2026 total, the United States accounts for about 204 TWh, or 36% of worldwide consumption, Gartner analyst Linglan Wang said. Dedicated AI data centers take roughly 68 TWh of that U.S. figure, one-third of the national total.
AI Servers vs Conventional Servers: Electricity Consumption
AI-optimized servers consumed 95 TWh in 2025 and are forecast to reach 175 TWh in 2026 and 258 TWh in 2027, per Gartner. Conventional servers barely move over the same window: 193 TWh, then about 195 TWh, then 200 TWh.
The crossover lands in 2027, when AI hardware out-consumes conventional hardware for the first time.
Cooling adds to the load. Gartner forecasts electricity used by cooling systems to climb 22.6% in 2026 to 195 TWh, a result of denser AI racks.
AI Energy Consumption Statistics: IEA Projections to 2030
The IEA’s Energy and AI report estimates data centers consumed around 415 TWh in 2024, about 1.5% of global electricity. Its base case sees that figure roughly doubling to around 945 TWh by 2030, just under 3% of global consumption.
From 2024 to 2030, the IEA expects data center electricity consumption to grow about 15% per year, more than four times faster than total consumption from all other sectors. Electricity use in accelerated servers, the category driven by AI adoption, grows 30% annually in the base case, against 9% for conventional servers.
| Period | Global data center electricity use | Share of global electricity |
|---|---|---|
| 2024 | ~415 TWh | ~1.5% |
| 2030 (base case) | ~945 TWh | Just under 3% |
| 2035 (base case) | ~1,200 TWh | Not stated |
Source: IEA, Energy and AI report, 2025
Note the model gap: the IEA’s base case reaches around 945 TWh in 2030, while Gartner’s June 2026 forecast puts 2030 consumption above 1,200 TWh. The two organizations use different scopes and assumptions, so both figures stand as published.
Which Countries Use the Most Data Center Electricity?
The United States held the largest share of global data center electricity consumption in 2024 at 45%, followed by China at 25% and Europe at 15%, per the IEA. Nearly half of U.S. data center capacity sits in five regional clusters.
US AI Data Center Energy Consumption Statistics
The DOE-commissioned report from Lawrence Berkeley National Laboratory, published December 2024, found U.S. data center electricity use tripled from 58 TWh in 2014 to 176 TWh in 2023. That 2023 figure equals 4.4% of total U.S. electricity consumption.
LBNL’s scenarios put 2028 consumption between 325 and 580 TWh, or 6.7% to 12.0% of forecast national electricity. Growth accelerates from roughly 7% a year in 2014-2018 to 18% in 2018-2023, then a projected 13% to 27% annually through 2028.
| Year | U.S. data center electricity use | Share of U.S. electricity |
|---|---|---|
| 2014 | 58 TWh | Not stated |
| 2018 | ~76 TWh | 1.9% |
| 2023 | 176 TWh | 4.4% |
| 2028 (scenario range) | 325-580 TWh | 6.7-12.0% |
Source: U.S. Department of Energy / Lawrence Berkeley National Laboratory, December 2024
How Much Energy Does One AI Prompt Use?
Google’s August 2025 technical paper was the first per-prompt disclosure from a major AI company. The median Gemini Apps text prompt in May 2025 consumed 0.24 Wh of electricity, emitted 0.03 gCO2e, and used 0.26 ml of water, about five drops.
Google compares the energy figure to watching television for less than nine seconds. Over the 12 months before May 2025, energy per median text prompt fell 33x and total carbon per prompt fell 44x.
| Metric (median Gemini Apps text prompt, May 2025) | Figure |
|---|---|
| Electricity | 0.24 Wh |
| Water | 0.26 ml |
| Emissions | 0.03 gCO2e |
| Energy reduction over prior 12 months | 33x |
| Carbon reduction over prior 12 months | 44x |
Source: Google technical paper and blog, August 2025
The active AI accelerator accounts for only 58% of that per-prompt energy; the rest goes to host CPUs and memory, idle backup capacity, and data center overhead, per the paper. The figures cover text prompts only, not image or video generation.
OpenAI CEO Sam Altman wrote in a June 2025 blog post that the average ChatGPT query uses about 0.34 Wh. Google reports a median and OpenAI an average, so the two numbers measure different things and are not directly comparable.
These prompt-level figures apply regardless of the device sending the request, since the computation runs in the cloud. That is part of why Gemini tools built into Chromebook Plus work well on modest local hardware, and why AI features inside the Chrome browser on ChromeOS carry the same per-prompt footprint as any other client.
What Do the AI Energy Consumption Statistics Add Up To?
Per-prompt energy fell 33x in a year, per Google, yet global data center consumption still grows 26% in 2026, per Gartner. Usage and workload scale are outrunning efficiency.
For an individual, one text prompt at 0.24 Wh is trivial. At industry scale, the IEA’s base case still doubles data center electricity use between 2024 and 2030.
FAQs
How much electricity will AI data centers use in 2026?
Gartner forecasts global data center electricity consumption of 565 TWh in 2026, up 26% from 447 TWh in 2025, with AI-optimized servers taking 31% of that power.
How much energy does a single AI prompt use?
Google measured the median Gemini Apps text prompt at 0.24 Wh, 0.26 ml of water, and 0.03 gCO2e as of May 2025, equal to under nine seconds of TV viewing.
Are the AI energy figures shared on Reddit accurate?
Reddit threads often cite older estimates far above Google’s measured 0.24 Wh per median Gemini text prompt (May 2025). Check the source and date, since per-prompt energy fell 33x in one year.
Why do Reddit users say AI energy use is falling and rising at once?
Both are true. Energy per median Gemini prompt dropped 33x in a year per Google, while total data center consumption grows 26% in 2026 per Gartner, because usage outpaces efficiency.
What share of U.S. electricity do data centers use?
U.S. data centers used 176 TWh in 2023, or 4.4% of national electricity, and could reach 325-580 TWh (6.7-12%) by 2028, per the DOE’s LBNL report.
