AI Energy Consumption in 2026

Artificial intelligence is becoming one of the largest new sources of electricity demand worldwide. According to the International Energy Agency, AI-related electricity consumption reached approximately 560 TWh in 2025 — more than the total electricity consumption of countries like Sweden or Poland. By 2030, this figure could reach 1,000 TWh, equivalent to Japan's entire electricity demand.

The energy intensity of AI comes from two main sources: training (building models) and inference (running them). Training a frontier model like GPT-4 consumed an estimated 3,600 MWh of electricity. But inference — the daily queries from millions of users — accounts for the majority of ongoing energy use. A single ChatGPT query uses roughly 10x the electricity of a Google search.

Data center construction is booming in response. Over 12 GW of new compute capacity is under construction in the US alone, with the largest planned facilities requiring up to 700 MW — enough to power a city of 500,000 people. This buildout raises questions about grid capacity, water usage (Microsoft used 6.6 billion liters in 2023 for cooling), and carbon emissions.

The environmental picture is mixed. Approximately 48% of AI data center capacity uses renewable energy, and major tech companies have committed to carbon neutrality. However, absolute emissions continue to rise — Google reported a 48% increase in CO₂ emissions from 2019 to 2023, largely driven by AI workloads. The challenge ahead is scaling AI infrastructure while meeting climate commitments.