AI runs on
water and power.
Here are the real numbers, and what changes when the AI is on your PC.
Data centres are growing fast.
Source: IEA, Energy and AI (2025).
How much water do your questions use?
Data centres are cooled with water that evaporates. Slide to see your year.
Low end: Google, 0.26 mL per Gemini prompt (2025). High end: UC Riverside, about 500 mL per 20 to 50 questions (2023), which also counts the water used to make the electricity. Estimates vary a lot.
Your PC isn’t magic.
Your PC uses electricity too.
Running AI at home uses power, and big data-centre chips get more answers out of each watt than a home PC.
But no cooling tower runs for you.
Your PC is cooled by a fan, not by evaporating water. And nobody builds a new data centre because you asked a question.
AI isn’t the main problem.
Even in 2030, data centres are a few percent of the world’s electricity. Planes, cement and steel are bigger. We’d rather tell you the truth.
Quick answers.
How much water does one AI question use?
Somewhere between a few drops and a couple of tablespoons, depending on who measures. Google says 0.26 mL for a Gemini prompt; UC Riverside estimated about 10 to 25 mL per question for ChatGPT, counting the water behind the electricity.
How much electricity do data centres use?
About 415 TWh in 2024, around 1.5% of the world’s electricity, heading for about 945 TWh by 2030, according to the IEA.
Is GreenCube greener?
It uses no data-centre cooling water and needs no new data centre. Your PC still uses electricity, so it’s not zero impact.
Is AI destroying the planet?
No. It’s growing fast, but it’s still a small part of the world’s energy. How it’s built and powered matters most.
Want more? How much water does AI use?