- Data centers consume around 1.5% of global electricity, with AI-focused centers accounting for about 0.5% of that total.
- AI energy consumption is dominated by inference (running models) rather than training, and individual queries vary from 0.24 Wh for simple text to 50 Wh for complex agentic tasks.
- Demand is geographically concentrated: in the US, data centers use 5% of electricity, with some states like Virginia exceeding 25%, and Ireland surpasses 20%.
- Future projections are uncertain, with AI expected to drive most growth, potentially reaching 3% of global electricity by 2030, but efficiency gains and clean energy sources will be critical.
- For individuals, AI query energy is minimal compared to daily consumption, but local grid strain and carbon emissions depend heavily on the electricity mix powering data centers.