AI and sustainability — how much energy does AI consume?
Training and running AI models costs enormous amounts of energy. But AI can also help with climate solutions. How does the balance work out?
The energy problem of AI
Training a large language model is estimated to consume as much energy as the annual consumption of hundreds of households. The IEA predicts data centers will consume as much electricity in 2026 as all of Japan.
Inference vs. training
Training is the most energy-intensive phase, but only happens once. Inference scales with usage.
Water scarcity
Data centers cool servers with water. Microsoft's data centers consumed 6.4 million m³ of water in 2022.
AI as a climate solution
- Google DeepMind reduced data center cooling energy by 40% using reinforcement learning
- AI optimizes energy grids and predicts power demand
- Climate modeling is faster and more accurate with AI
- Discovery of new battery and solar panel materials via AI simulations
More efficient models
Models like Mistral 7B and Phi-3 offer strong performance at a fraction of the compute. This efficiency race can limit the net impact.
Author: Claude claude-sonnet-4-6