Electricity use of AI coding agents

Summary

Throughout 2025, we got better estimates of electricity and water use of AI chatbots. There are all sorts of posts I could cite on this topic, but a favorite is this blog post from Our World in Data’s Hannah Ritchie. On the electricity front: In short, “unless you’re an extreme power user, asking AI questions every day is still a rounding error on your total electricity footprint.” A similar story applies to water usage. This one from Benjamin Todd: The average American uses 1600 liters of water per day, so even if you make 100 prompts per day, at 2ml per prompt, that’s only 0.01% of your total water consumption. Using a shower for one second would use far more. Generally, these analyses guide my own thinking about the environmental impacts of my individual usage of LLMs; if I’m interested in reducing my personal carbon footprint, I’m much better off driving a couple miles less a week or avoiding one flight each year. This is indeed the right conclusion for users of chat interfaces like chatgpt.com or claude.ai. That said, 1 or 10 or 100 median prompts a day is many orders of magnitude off from my own personal use of LLMs; I likely am, in Hannah Ritchie’s words, an “extreme power user.” I work in software and spend much of my workday driving 2 or 3 coding agents, like Claude Code, at a time. Thus, a much more relevant question for me is how much energy does a typical Claude Code session consume? (I’m not going to discuss water use in this post.) tl;dr, much more: There are so many considerations and assumptions and pieces of shorthand one must use along the way to answer this sort of question. I’ll do my best to call those out throughout this post, but please do understand this is still just Sunday afternoon napkin math from Some Guy. Coding agents and the ‘median query’ tl;dr: In this section, I point out that a Claude Code session should use orders of magnitude more energy than a ‘median query.’ The Gemini paper, Sam Altman blog post, and Epoch AI analysis of Chat...