Artificial intelligence is frequently sold as a climate solution and criticised as a climate problem — and both are true. Used well, AI is a powerful tool for cutting waste and emissions; built carelessly, it's an energy-hungry contributor to them. The honest picture requires holding both.
Where AI genuinely helps
- Energy grids — forecasting demand and balancing supply, integrating intermittent renewables that are hard to manage with traditional methods.
- Logistics and transport — optimising routes and loads to cut fuel and emissions.
- Precision agriculture — applying water, fertiliser, and pesticide only where needed, reducing inputs and runoff.
- Buildings — smart HVAC and lighting that cut energy use without sacrificing comfort.
- Materials and climate science — accelerating discovery of better materials and modelling climate systems.
AI's value to sustainability is mostly in optimisation — squeezing waste out of systems too complex for humans to tune by hand.
The cost side, honestly
Training and running large AI models consumes significant electricity and water (for data-centre cooling). As AI use scales, so does its footprint. Ignoring this — or assuming AI is automatically "green" — is part of the problem. Responsible use means weighing whether the efficiency gained outweighs the resources spent, and pushing for cleaner data-centre energy.
The net picture
For most practical applications, AI applied to optimise a wasteful system saves far more than the model costs to run — a route-optimisation model's energy is trivial against the fuel it saves across a fleet. The risk is using AI indiscriminately, where the footprint isn't justified by the benefit. The discipline is intentionality: deploy it where it meaningfully reduces waste, and account for its own cost honestly.
For organizations
The opportunity is real and aligns with frameworks like ISO 50001 (energy) and ISO 14001 (environment): use AI to find and cut the inefficiencies those systems are designed to surface, while measuring and managing the technology's own footprint as part of the equation.
Use AI for a greener operation
My AI and Sustainability course covers where AI genuinely reduces environmental impact, its real resource costs, and how to deploy it intentionally rather than indiscriminately.
View the course →Questions
Is AI good or bad for the environment?
Both — it's a powerful optimisation tool that can cut large amounts of waste, but training and running models consumes energy and water. Net impact depends on using it where the benefit clearly outweighs the cost.
Does AI use a lot of energy?
Large models do, plus water for cooling. For a well-chosen application the savings usually dwarf that cost, but it should be measured, not assumed away.