How much energy do data centers and artificial intelligence use? (ourworldindata.org)

Recent analyses shed light on the significant energy demands of data centers, particularly those focused on artificial intelligence (AI). In 2025, data centers consumed an estimated 485 terawatt-hours (TWh) of electricity, accounting for about 1.5% of global electricity generation. Notably, while AI-focused data centers consumed around 155 TWh—approximately 0.5% of the world's electricity—the growing reliance on AI raises concerns about carbon emissions, local energy supply strains, and the potential bottleneck this could impose on technological advancement. Projections suggest that by 2030, data centers may reach 3% of global electricity consumption, with AI facilities likely matching the energy use of non-AI counterparts. Understanding the energy consumption of individual AI queries reveals that while they have a minimal impact, the cumulative effect can be significant due to the rapid AI growth. For instance, a typical text query to models like Google's Gemini or OpenAI's ChatGPT is estimated to consume between 0.24 and 0.34 watt-hours—comparable to a microwave running for a fraction of a second. However, with high-density AI usage, especially during complex queries, energy demands can rise sharply. This raises pressing questions about the sustainability of AI growth and the ability of local power grids to handle increased demand while minimizing environmental impacts, especially in regions already dependent on data centers for a substantial portion of their electricity.