Blab Tech
Data Centers to Match Japans Electricity Use by 2030, IEA Warns
By 2030, the global data‑center sector could consume as much electricity as Japan does today. That startling comparison comes from the International Energy Agency’s (IEA) April 2025 report Energy and AI, and the agency reiterated the warning in a new projection released in April 2026.
In the 2025 briefing, IEA executive director Fatih Birol warned that “Global electricity demand from data centres is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today.” He added that “There is no AI without energy – specifically electricity” and that “Thousands of data centres are set to be built in the next five years.” The figures behind those words are hard to ignore.
IEA data show global data‑center electricity use rose from 415 TWh in 2024 to 447 TWh in 2025 and 565 TWh in 2026. Gartner’s 2026 forecast confirms the upward trend. The agency projects the figure will nearly double to 950 TWh by 2030, a growth rate of roughly 15 % per year. At that level, data centres would account for about 3 % of worldwide electricity demand.
Artificial intelligence is the main engine behind the surge. The IEA’s updated outlook indicates that electricity consumption by AI‑focused data centres will triple between 2025 and 2030. In 2025, AI‑centric facilities saw a 50 % increase in power use, compared with a 17 % rise across all data‑centre types. The agency notes that the United States and China together could supply nearly 80 % of the global growth in data‑centre demand through 2030.
On the supply side, renewables are expected to meet almost half of the additional demand by 2030. Natural gas and coal together are projected to supply more than 40 %, while nuclear power is expected to become increasingly important toward the decade’s end and beyond. The IEA cautions that whether the grid can accommodate this demand depends on policy choices that remain unsettled.
Efficiency gains per AI task have been significant. The IEA reports that energy use per AI operation has fallen by at least an order of magnitude annually. However, total consumption continues to climb, a pattern that researchers attribute to Jevons’ paradox: making a technology cheaper and more efficient can lead to greater overall use. A paper by Luccioni and colleagues argues that rebound effects may undermine the assumption that technical efficiency alone will reduce environmental impact.
The projected energy footprint raises questions for regulators and utilities. The IEA’s analysis suggests that meeting the 950 TWh target will require substantial investment in generation capacity, transmission infrastructure, and grid management. The agency notes that building new transmission lines, substations, or generation facilities can take a decade or more, a timeline that may lag behind the pace of data‑centre expansion.
In the short term, the IEA’s findings highlight the need for coordinated policy responses. While the agency does not prescribe specific measures, it emphasizes that governments, utilities, and technology companies must decide how to balance the growth of AI services with the sustainability of the electricity system.
The article concludes that the data‑centre sector is on a trajectory that will match Japan’s current electricity consumption by 2030. The IEA’s projections, grounded in 2024–2026 consumption data and growth trends, underscore the scale of the challenge. The sector’s continued expansion, driven largely by AI workloads and concentrated in the United States and China, will test the resilience of global power grids and the effectiveness of renewable and fossil‑fuel supply strategies. The outcome will depend on policy decisions that have yet to be finalized.
In the 2025 briefing, IEA executive director Fatih Birol warned that “Global electricity demand from data centres is set to more than double over the next five years, consuming as much electricity by 2030 as the whole of Japan does today.” He added that “There is no AI without energy – specifically electricity” and that “Thousands of data centres are set to be built in the next five years.” The figures behind those words are hard to ignore.
IEA data show global data‑center electricity use rose from 415 TWh in 2024 to 447 TWh in 2025 and 565 TWh in 2026. Gartner’s 2026 forecast confirms the upward trend. The agency projects the figure will nearly double to 950 TWh by 2030, a growth rate of roughly 15 % per year. At that level, data centres would account for about 3 % of worldwide electricity demand.
Artificial intelligence is the main engine behind the surge. The IEA’s updated outlook indicates that electricity consumption by AI‑focused data centres will triple between 2025 and 2030. In 2025, AI‑centric facilities saw a 50 % increase in power use, compared with a 17 % rise across all data‑centre types. The agency notes that the United States and China together could supply nearly 80 % of the global growth in data‑centre demand through 2030.
On the supply side, renewables are expected to meet almost half of the additional demand by 2030. Natural gas and coal together are projected to supply more than 40 %, while nuclear power is expected to become increasingly important toward the decade’s end and beyond. The IEA cautions that whether the grid can accommodate this demand depends on policy choices that remain unsettled.
Efficiency gains per AI task have been significant. The IEA reports that energy use per AI operation has fallen by at least an order of magnitude annually. However, total consumption continues to climb, a pattern that researchers attribute to Jevons’ paradox: making a technology cheaper and more efficient can lead to greater overall use. A paper by Luccioni and colleagues argues that rebound effects may undermine the assumption that technical efficiency alone will reduce environmental impact.
The projected energy footprint raises questions for regulators and utilities. The IEA’s analysis suggests that meeting the 950 TWh target will require substantial investment in generation capacity, transmission infrastructure, and grid management. The agency notes that building new transmission lines, substations, or generation facilities can take a decade or more, a timeline that may lag behind the pace of data‑centre expansion.
In the short term, the IEA’s findings highlight the need for coordinated policy responses. While the agency does not prescribe specific measures, it emphasizes that governments, utilities, and technology companies must decide how to balance the growth of AI services with the sustainability of the electricity system.
The article concludes that the data‑centre sector is on a trajectory that will match Japan’s current electricity consumption by 2030. The IEA’s projections, grounded in 2024–2026 consumption data and growth trends, underscore the scale of the challenge. The sector’s continued expansion, driven largely by AI workloads and concentrated in the United States and China, will test the resilience of global power grids and the effectiveness of renewable and fossil‑fuel supply strategies. The outcome will depend on policy decisions that have yet to be finalized.
