Published 14 days ago • loading... • Updated 12 days ago
Using AI chatbots? Here are 4 ways to reduce the energy drain
Experts say shorter outputs and smaller task-specific models can cut energy use by as much as 90%, according to a UNESCO study.
As AI data centers' electricity demand jumped 17% in 2025, experts are identifying strategies to reduce the environmental footprint of generative AI tools used for tasks ranging from medical diagnosis to code generation.
Smaller language models consume between 15 and 50 times less energy than all-purpose models, a 2025 UNESCO study found, with shifting to these specialized tools resulting in a 90% decrease in energy use overall.
University College London computer scientist Ivana Drobnjak found that instructing models to halve their output reduced energy consumption by 50%, while shortening user prompts provided only a 5% energy saving.
Gunter Klambauer, an AI expert at Johannes Kepler University in Austria, suggests users avoid AI for simple tasks like weather checks and select "Web results only" to save energy during searches.
With around 3.2 billion queries sent daily to ChatGPT, individual choices matter; Drobnjak noted that users who select the right tools for specific tasks can make "a real difference" amid broader tech industry efforts.