AI Water & Energy Usage: ChatGPT vs Claude vs Gemini (2026)
Summary
Analysis of energy and water consumption for various AI models, comparing standard and reasoning models alongside the sustainability strategies of major cloud hyperscalers.
Key quotes
The efficiency gap between AI models spans over 200x, choosing the right model is itself a sustainability decision.
Reasoning models like o3 or DeepSeek-R1 can consume 10-70x more per prompt.
AI-driven data centres could withdraw up to 6.6 billion m3 of freshwater by 2027, more than four times Denmark’s annual usage.
The article provides specific energy (Wh) and carbon (g CO2) benchmarks for models including GPT-4, Claude 3.7, and DeepSeek-R1. It also compares the PUE and WUE metrics of Microsoft, AWS, Google, and Meta.