The below sources are where I drew my estimate of Claude Code’s energy consumption, from both cloud/datacenter processing AND my local machine, based on rough measurements for Sonnet 3.7 and extrapolated up for Opus 4.5 by about x5. Numbers are with heavy use on the max plan (which I don’t actually have). I tried to estimate conservatively here.

• https://arxiv.org/html/2505.09598v1
• https://github.com/anthropics/claude-code/issues/4580
• https://www.sustainabilitybynumbers.com/p/carbon-footprint-chatgpt
• https://support.claude.com/en/articles/11145838-using-claude-code-with-your-pro-or-max-plan
• https://www.fastcompany.com/91336991/openai-anthropic-deepseek-ai-models-environmental-impact
• https://www.datacenterknowledge.com/supercomputers/project-rainer-aws-anthropic-complete-massive-ai-supercomputing-cluster
• https://epoch.ai/gradient-updates/how-much-energy-does-chatgpt-use
• https://undetectable.ai/blog/how-much-energy-does-chatgpt-use/
• https://www.sunearthtools.com/tools/CO2-emissions-calculator.php
• https://llm-tracker.info/_TOORG/Power-Usage-and-Energy-Efficiency
• https://arxiv.org/html/2407.16893v1
• https://hyperdev.matsuoka.com/p/critical-memory-leak-in-claude-code