Electricity use of AI coding agents (www.simonpcouch.com)

In 2025, researchers have begun to quantify the electricity and water usage of AI chatbots, revealing that for average users, daily interactions with these models consume negligible amounts of electricity and water. Hannah Ritchie's analysis suggests that engaging with AI chatbots represents a minor fraction of an individual's overall environmental footprint, with even frequent usage resulting in a minimal impact compared to everyday activities like driving or showering. This finding is significant as it provides a clearer understanding of the environmental trade-offs associated with AI technologies, encouraging users to focus on larger lifestyle adjustments for carbon reduction. However, for "extreme power users" like software developers who rely heavily on coding agents such as Claude Code, the energy consumption becomes substantially greater. These sessions are markedly more resource-intensive, given that they involve numerous tool calls and longer queries that significantly exceed the energy cost of a standard "median query." By estimating the energy consumption per token and analyzing real session data, it becomes apparent that a typical coding session could demand orders of magnitude more energy than casual usage scenarios. This highlights the need for further research into optimizing the energy efficiency of AI systems, particularly for high-demand applications in software development and other fields.