Global data centers consumed approximately 460 terawatt-hours of electricity in 2022, a figure the International Energy Agency projects will reach between 620 and 1,050 terawatt-hours annually by 2026, depending on AI adoption rates. At the upper bound, global data center electricity consumption would rival the total electricity consumption of Japan, the world's third-largest economy.

The United States hosts the largest data center stock in the world and accounts for roughly 38% of global data center electricity consumption. The Northern Virginia corridor is the single largest data center market by capacity, followed by Dallas, Chicago, Atlanta, and Silicon Valley.

AI is the primary demand driver accelerating the projections. Training a single large language model can consume as much electricity as several hundred US homes use in a year. Inference workloads, where AI models answer queries at scale, are growing faster than training and are expected to become the dominant electricity load category by 2027.

Cooling and power distribution infrastructure account for approximately 40% of total data center electricity consumption. Power usage effectiveness, a ratio of total power used to IT equipment power, has improved from an industry average above 2.0 in 2010 to approximately 1.5 today, with hyperscale operators reporting PUEs below 1.2 at newer facilities.

The IEA report identifies colocation and hyperscale data centers as the two fastest-growing facility types, with enterprise data centers declining as workloads migrate to cloud infrastructure. The shift concentrates electricity demand in fewer, larger facilities, making grid planning more predictable but also increasing the stakes of any single large campus connection.

Source: International Energy Agency -- https://www.iea.org/reports/data-centres-and-data-transmission-networks