AI uses significant energy — especially for training large models and running them at scale — and its growth is driving data-centre electricity demand. The exact footprint is debated, so treat single figures with caution.

Where AI uses energy

AI consumes energy in two main ways: training (the one-off, energy-intensive process of building a large model) and inference (running the model to answer queries, which adds up across billions of uses). Data centres running AI also use water for cooling. As AI use grows, so does its share of data-centre electricity demand.

How big is the footprint?

Estimates vary widely and are debated. A single AI query uses far more energy than a traditional web search by most estimates, but exact figures depend on the model and method, and many widely shared numbers are rough or contested. The bigger picture is that aggregate AI energy demand is rising fast, prompting investment in efficiency and clean power. [Treat specific figures as estimates.]

What’s being done — and what you can do

Providers are improving efficiency (better hardware, smaller models for many tasks) and investing in renewable and other low-carbon energy, though demand growth is a real challenge. As a user, you can prefer efficient tools, avoid wasteful over-use, and use smaller/local models for simple tasks. The honest summary: AI’s energy use is significant and growing, the precise impact is uncertain, and efficiency is improving alongside demand.

If you find yourself juggling a separate subscription for chat, automation, transcription and image generation, one option worth knowing is a single platform that runs them together — osFoundry is one such agentic AI platform that consolidates chat, agents and internal apps in one workspace, with a bring-your-own-key model so you choose the underlying AI.

This article is general information, not professional, legal or financial advice. AI tools, prices and availability change fast — verify current details on the official source before you rely on them.