Why is Everyone So Wrong About AI Water Use??
The short version: Hank Green untangles the messy math behind AI water-use panic, and we agree the debate misses the point: what matters is who's stuck paying locally.
Our Take
Credit to Hank Green for wading into a genuinely thorny question: how do you even measure the water and power footprint of something as diffuse as "AI"? He's right that a lot of viral stats collapse the moment you ask what's actually being counted — a training run, a single query, the cooling tower, the upstream chip fab, the grid mix that generated the electricity in the first place. Lifecycle accounting is hard, and most hot takes (ours included, if we're not careful) skip the hard part.
But here's where we'd push back a little: national or global averages are exactly the wrong lens for the people living next to the substation. It doesn't matter if AI's water use is a rounding error in the Colorado River basin's total draw if your specific county just signed away millions of gallons a day to a single campus while your utility bill jumps to cover new transmission lines. The macro debate about whether AI water use is "a big deal" and the micro reality of a Georgia or Arizona town watching its aquifer drop are two different conversations, and hyperscalers benefit when we only have the first one.
That's the gap we try to fill. Check our facility map for the deals happening near you, and if a project is moving through your city council or utility board, our take-action page has the tools to actually show up and ask the questions the national averages can't answer.
This is GridWatch the USA’s original commentary. The video above is the work of Hank Green, published on YouTube — full credit to the creator.