Seems like a sensible move to save water would be for fewer people to eat as much meat, while each of the 1000 equivalent heavy AI users refrained from being as heavy themselves.

If one is good then both would be better.

How good could it get anyway?

Under ideal conditions I guess that would mean no meat at all, which may seem extreme but in the end it's not going to change what a lot of people are doing with their life outside of mealtimes.

But if for every person who becomes a vegetarian, there were 1000 vibe coders who were not doing anything like that at all any more, that might be a much more significant improvement in the overall scheme of things.

Depends on what they did with their time instead, but 1000:1 looks like some pretty good odds.

[dead]

I kept seeing articles about how much water AI uses (data centers consumed 1.7 trillion gallons in 2024) — and I wondered how it actually compares to other things we do every day. Turns out producing one pound of beef uses about 1,800 gallons of water. A single ChatGPT query uses roughly 16oz, which is an excellent size for a steak.

So I did the math: one person switching from a typical American diet to vegetarian saves enough water annually to offset over 1,000 heavy AI users. The citations are real, the pricing is not--still working on that.

Curious what HN thinks — is the framing of AI water usage overblown relative to other consumption, or is it a legitimate infrastructure concern at scale?