Why Do AI Data Centres Use So Much Water? | Explained
The short version: Business Today lays out the global math on AI's water appetite — we're asking who's checking the local math in your town.
Our Take
Credit to Business Today for putting a number on something most people never think about: every AI query has a water bill attached, and the industry-wide tab could hit 6.6 trillion liters a year by 2027. That's the kind of stat that sounds abstract until you realize it's not spread evenly — it lands hardest on whatever county happens to host the server farm.
Global estimates are useful, but they can also let local officials off the hook. The real question isn't "how much water does AI use worldwide," it's "how much water is this specific facility pulling from this specific aquifer, and did anyone tell the neighbors before the permit got signed?" Plenty of these deals get approved with vague efficiency promises and no binding water caps, in places already stretched thin by drought or agriculture.
If you want to see what's being built near you and who approved it, check our facility map. And if your community is staring down a new data center proposal, our take-action page has resources for asking the water and power questions before the concrete gets poured, not after.
This is GridWatch the USA’s original commentary. The video above is the work of Business Today, published on YouTube — full credit to the creator.