AI, Energy and the Environment: The Honest Picture

AI is both a heavy user of energy and a tool that can help cut it. Both things are true.

• Training and running models takes significant electricity and, in data centres, cooling water.
• Demand is growing as more people and products use AI.
• Hardware manufacturing has its own footprint.
• Footprints vary widely by model size, location and how clean the local power is.

• Grid management: forecasting supply and demand, integrating renewables.
• Building efficiency: smarter heating, cooling and lighting.
• Logistics: shorter routes and fuller loads.
• Materials and battery research: speeding up discovery.
• Weather and climate modelling: better forecasts and risk planning.
• Monitoring: spotting leaks, deforestation or methane emissions from images and sensors.

1. Use the right-sized model. A small model is often enough for simple jobs.
2. Avoid waste: don't regenerate a hundred images when ten will do.
3. Ask vendors about energy sourcing and efficiency.
4. Run locally or on efficient hardware where it makes sense.
5. Measure the impact of AI projects, not just the savings they promise.

AI's footprint is real but not fixed. Efficiency is improving, and the biggest gains come from using AI deliberately for problems where it saves more than it costs.