Back to blog
business-automation·3 July 2026

AI is getting bigger, and so is its pollution problem

AI tools may feel invisible, but the buildings powering them are leaving a very real mark.

AI is getting bigger, and so is its pollution problem

AI can feel almost weightless when you use it. You type a question, get an answer, and move on. Maybe it helps write an email. Maybe it summarizes a meeting. Maybe it saves a team an hour of boring work.

But somewhere, in a very real building, machines are burning through electricity to make that happen.

That is the uncomfortable part in TechCrunch's piece about Google and Amazon's latest climate reports. Both companies say their emissions went up last year: Google's by 25%, Amazon's by 16%. A big reason is the rush to build more space for AI.

This is not just about keeping the lights on in an office. AI needs huge data centers, and those buildings do not appear out of thin air. They need steel, cement, chips, cooling systems, backup power, and a lot of electricity. Even before a new data center starts running, it has already created a carbon bill.

That makes the AI boom feel a bit less magical.

I do not think the answer is to pretend these tools are useless. They clearly are not. Businesses are using AI to answer customer questions, sort paperwork, draft reports, and shave time off repetitive jobs. For a small team, that can matter. A lot.

But the sales pitch often skips the messy part. Every "instant" answer has a supply chain behind it. Every company promising faster work with AI is also leaning on a growing physical network of buildings, power lines, materials, and factories.

Google and Amazon both say they still plan to hit their climate goals. Maybe they will. They have money, engineers, and buying power that most companies can only dream about. Still, the numbers are moving in the wrong direction right now, and that is worth sitting with.

The bigger question for the rest of us is pretty simple: what do we actually want AI to do?

Using it to cut drudge work, help people with disabilities, or handle jobs nobody has time for feels different from stuffing it into every app just because investors like the word. If AI is going to use this much energy, the bar should be higher than "it makes a button slightly fancier."

The convenience is real. So is the cost. We should be honest about both.

Continue reading