The International Energy Agency's annual data on data centre electricity consumption provides the most authoritative global benchmark for tracking how digital infrastructure is reshaping electricity demand. The IEA's 2026 reporting indicates that data centres, data transmission networks, and cryptocurrency operations together consume a growing share of global electricity, with data centres alone projected to approach 10 percent of total global electricity consumption as AI training and inference workloads scale.
US data centres represent the largest national share of data centre electricity demand globally, driven by the concentration of hyperscale cloud infrastructure operated by major technology companies. Total US data centre electricity consumption has roughly doubled over the past five years, reflecting both the growth of cloud services and the more recent acceleration of AI infrastructure deployment.
The energy intensity of AI workloads significantly exceeds that of conventional cloud computing. Large language model training runs require continuous operation of dense GPU clusters for weeks or months, while AI inference workloads generate persistent electricity demand proportional to query volume. As AI services scale to broader user populations, inference workloads are expected to grow faster than training in absolute electricity consumption terms.
Water consumption by data centres, which use cooling systems that evaporate large quantities of water, has emerged as a secondary resource concern alongside electricity. Communities hosting large data centre campuses have begun raising questions about municipal water supply impacts alongside grid capacity questions.
Source: International Energy Agency -- https://www.iea.org/energy-system/buildings/data-centres-and-data-transmission-networks