Gartner said global data centre electricity consumption is projected to reach 565 terawatt-hours (TWh) in 2026, a 26% increase over 2025 (447 TWh). The surge is being driven by intensive AI workloads, and Gartner warned that power availability is becoming the new bottleneck for scaling compute infrastructure.
Gartner highlighted several key facts for 2026:
- Electricity use: 565 TWh (up 26% from 2025).
- Power capacity: peak demand reaching 132 gigawatts (GW) worldwide – a 27% rise from 104 GW in 2025.
- AI server share: AI-optimised servers will account for 31% of data centre power in 2026, up sharply from 20% last year, and their power draw is expected to exceed conventional servers by 2027.
Gartner’s Linglan Wang, a director analyst, commented that “surging demand for compute-intensive AI workloads is driving unprecedented data centre power growth, while AI capacity is now constrained by power availability”. In short, the electricity needed to run AI is outpacing efforts to build more grid capacity or generators.
[Also Read: India’s Power Grid Faces Nighttime Strain Amid Solar Growth: Citi ]
In 2026, Gartner forecasts that nearly one-third of all data centre electricity will go to AI-optimised servers. The firm’s data show conventional server power demand growing only modestly (2–3% per year), while AI server demand roughly doubles each year through 2027 (see table). By 2027, AI servers are expected to consume more power than all other servers combined.
Gartner calls this a “power-hungry” phase of the AI cycle. The analysis notes that by 2030 data centre demand could top 1,200 TWh annually – nearly three times today’s level – unless operators prioritise energy efficiency and new power sources. The report explicitly recommends investing in high-efficiency cooling, edge computing and grid resiliency to “ensure sustainable, scalable growth” as AI demand grows.
As AI scales, the power problem scales with it
Gartner’s findings echo other energy data pointing to a looming crunch. For example, the U.S. Energy Information Administration (EIA) now expects total U.S. electricity use to hit record highs in 2026–2027, driven in part by AI data centres. The EIA forecasts U.S. power demand rising from about 4,195 billion kWh in 2025 to 4,271 billion in 2026 and 4,397 billion in 2027. It explicitly cites AI and crypto mining data hubs as a major factor behind the surge.
Another study also shares a similar story: United Nations researchers say that data centres consumed roughly 448 TWh worldwide in 2025 – more electricity than all of Saudi Arabia uses in a year. They warn power use could double to 945 TWh by 2030, with AI accounting for about 40% of the total. The International Energy Agency (IEA) predicts data centres will drive about 20% of electricity demand growth in advanced economies by 2030. In the EU, officials are already planning minimum efficiency standards for data halls because capacity is projected to more than double (to ~28 GW) by 2030.
448 TWh was used by data centres globally in 2025 – roughly the same as the entire electricity consumption of Saudi Arabia that year. Hitting 565 TWh in 2026 would put the industry’s demand on par with a mid-size country’s use. In the U.S., a recent study found data centres could consume up to 9% of all electricity by 2030, more than double their share today.
[Also Read: AI Is Growing Faster Than the World’s Data Centers Can Handle ]
What it means
Gartner’s forecast signals a shift in planning priorities. It’s no longer enough to buy the fastest servers – organisations must also ensure they have guaranteed power and energy budgets.
Companies building new AI clusters will increasingly vet data centre sites for grid capacity and renewable energy. Those that can marry compute with sustainable power sources will have a competitive edge.
