The water use actually isn't all that high - it's just easy to make "a million gallons of water every year" sound like a lot, but compared to a 500 acre farm which could easily use 3 million gallons every day, it's not very big.
The electricity use is really substantial though, but that's harder for people to visualise so gets less media attention.
Right, this argument completely lacks context. Agricultural use of water is astounding. And even that is generally much less than the enormous amount of water that is available in agricultural states (CA notwithstanding).
Minnesota where I live gets approximately 3x10^13 gallons of rain / year. Yes, almost none of that is captured for use, but it's not like we're talking about a fundamentally physically limited resource here. It's just that there's a bad time/phase mismatch.
Hell, a 500 acre data center has >200 million gallons drop onto it out of the sky in MN, each year (20in avg * 500 acre = 10,000 acre-inches)
Not necessarily because if you have a closed loop system then that vastly decreases the amount of water usage and increases the amount of electricity (the water has to be cooled)
Water use is not necessarily linked with energy use. Open up Google's annual environmental report and look at the water consumption for each facility. Unrelated to the size/power of those sites.
It is strongly linked when you adjust for local outdoor temperature and humidity. Pretty much all sites use evaporative chillers whose properties are well known.
> pretty close proxy for how much computation is happening.
[citation needed]. See the vastly different power budget and cost of AWS graviton ARM vs x86 compute. Looking even at power use directly is only going to give a very low precision proxy for aggregate compute, with water usage even more indirect.
Looking at power use directly and making some educated guesses about average FLOPs/watt is probably the most effective way to estimate aggregate compute.
Even at Amazon I wouldn't be surprised if it's the primary way they do it, and I would be interested in some research. I'm trying to think of other ways, and accurately aggregating CPU/GPU load seems virtually impossible to do in a very rigorous way at that scale.
And yes, as an outsider you might have trouble knowing the relative distribution of ARM/x86, but that's just another number you want to obtain to improve your estimate.
Counterpoint: you have no factual basis for believing anything about the energy used by various CPUs in EC2, none of which are publicly available parts.