- Data centers currently consume approximately 1.5% of global electricity, with AI-specific facilities accounting for about 0.5%.
- Electricity demand is highly concentrated; in the US, data centers account for 5% of total demand, reaching over 25% in specific states like Virginia.
- Individual AI query energy usage is relatively low: a median text-based query consumes about 0.24 watt-hours (Wh), though complex agentic tasks can require up to 50 Wh.
- Future energy projections are uncertain and rely heavily on efficiency gains in hardware versus the growth in user demand.
Data Center Energy Consumption
- Global data center electricity demand reached roughly 485 TWh in 2025, according to the IEA.
- AI-focused data centers represent about one-third of total data center power usage.
- Discrepancies between data sources (such as the IEA versus S&P/Energy Institute) stem partly from the inclusion of cryptocurrency mining in some datasets and varying methodological approaches.
Environmental and Local Implications
- The primary concern is not total global consumption but local grid strain and the carbon intensity of the energy sources powering these centers.
- Emissions impacts depend more on whether the power source is coal-based, nuclear, or renewable rather than the absolute amount of electricity consumed.
- AI adoption may act as a bottleneck for its own expansion if power infrastructure cannot scale to meet regional demand spikes.
This summary was generated by AI from the original article and may omit nuance or later updates. How everytldr works