I Was Wrong About Data Center Water Consumption
construction-physics.com
construction-physics.com
In the US, golf courses consumed 1.68 million acre-feet of water in 2022[0] (down 29% from 2006, they are pleased to report), which works out to 1,499 million gallons per day.
So golf courses consume three times as much water as AI. Were's the proportional level of outrage over that issue, I wonder?
[0] https://gcmonline.com/course/environment/news/water-manageme...
either way, if a murderer says “why are you charging me? someone else was murdereded last night in the next town over!” People would rightfully laugh at the murderer and continue to charge him, and rightfully so.
Do I really need to say that ecological impacts and murder are very different things?
The funny part is your "murderer in the next town over" example is almost exactly how cap-and-trade works (plus a price mechanism), which is the most effective system yet devised to reduce smog from power plants. So in this area of public policy we already let off the "murderer" (eyeroll) with that excuse, and rightfully so...
The golf course number was surprising so I shared it. Forgive me.
It's telling that for AI we're only given impressive-sounding water numbers, but never in the context of total consumption. It has all the hallmarks of a media-driven moral panic, not serious activism.
The fact that HN's response is to compare (these particular) water users to murderers[0] and cry whataboutism and unseriousness against me... it's not exactly helping convince me that this isn't 90% motivated moral panic and 10% legitimate grievance.
If you bought that golf course and built something productive that produced value for the community, like apartments, detached houses, a factory, a store, anything at all, you'd suddenly have to pay more tax on it.
A land value tax would push golf courses further from city centers.
I am willing to believe that one golf club may be worth subsidising, but not three, not the ones which have locks on the gate to prevent community access, and not when the joining and membership fees are so high.
The AI comparison mentioned I believe is really just data center consumption which goes much further than just AI but a large amount of important compute power and storage, with most of it unseen.
Golf courses are at least pleasant to look at and whilst they might use a lot of water, most people can understand it given their own need to water their own lawns. Whereas a data center uses both comparative levels of water, but also enormous amounts of electricity. Whereas a golf course doesn't use even a small fraction of the same amount of electricity.
Even large parklands and other green areas need large amounts of water, but most people are not going to disagree with that.
https://www.twdb.texas.gov/innovativewater/desal/doc/2024_Th...
The current projection is ~200k acre/ft by 2070 from seawater sources. It'd be cool if the AI/data center folk would combine our three ares of strength: (1) wind; (2) seawater desalination; and, (3) mass liquid piping. We'd use excess offshore wind to desalinate seawater and pump it inland.
That depends entirely on the depth of the river and the depth of the reservoir. If the average depth of the reservoir is deeper than the average depth of a river there is less surface area.
Given that Earth averages towards smooth, the rate of width increase overwhelms the rate of depth increase when dams fill.
The only exceptions might be narrow canyon dams with nowhere to spread laterally, or dams on already-wide rivers. But those are rare.
No it doesn't.
The only thing that matters (in this oversimplified calculation which only takes into account surface area) is average depth of the freshwater while it is on land. If the reservoir is on average deeper than the rivers the freshwater otherwise would be flowing in, there will be less evaporation per liter of freshwater available for use.
Now a dam also increases the total amount of freshwater that's kept on the land in a steady state situation compared to if the water flowed free into the sea. It would be absurd to count this as "extra evaporation" when this extra freshwater otherwise would've simply be lost when it would flow into the sea instead of being kept in the reservoir.
Furthermore the fact that dams increase evaporation though this mechanism is an easily verifiable scientific fact. https://riverresourcehub.org/resources/how-dams-affect-water...
> Does it make sense to include this water evaporation in the share of water consumed by data centers? I think it’s debatable.
Even without pawing over the exact numbers, having a cursory understanding of how a datacentre works should highlight to you that excessive water consumption is going to be a problem. I couldn't imagine writing essays to try and argue against that in the first place. And to be an order of magnitude wrong in your position... sheeesh.
Ground water is not fungible economic value. When its gone it is gone because the aquifers collapse. The CA central valley dropped 9M from 1925 to 1977 [1]
The space the water was in, is gone. Datacenters pumping groundwater will do the same exact thing.
[1] https://www.reddit.com/r/pics/comments/1jby06/this_is_the_im...
This is true when examining the environmental impact of anything.