Oct 22, 2026, 9:15 PM@nirak *sigh* That is not something I'd considered. The website that they cited is a *marketing* company. The sound you are hearing is me hitting my head against the wall in frustration.
"Showcase your company, solutions, services, products, reports, surveys, events, or other content..."
Oct 22, 2026, 10:24 PM@JessicaAnnEng and all those estimates are from self reported data from the companies, which of course is going to be 100% accurate
I really think they love the water issue because it's the easiest to shoot down, though. What about gas generators running near people's homes? What about all the rare earth materials being mined to make GPU's that will only last 5-8 years? What about mass clearing land to build these things? Etc etc etc
I hate this for your kid :(
@nirak @JessicaAnnEng Very much this. The most common go-to rebuttal I hear re: the environmental concern is either one of two things:
1. "Data centers use a closed loop system" (yet provide no reciepts re: how often this "uses less water" system is being used in existing (AI or non-AI) data centers. They also never address how the water in that loop is recooled nor address the fact that closed loop systems might add coolant chemicals to the water in that loop
2. "You are against data centers.... why dont you go after golf courses, too."
Even knew someone who used both of these talking points in the same conversation.
@Catwoman69y2k @nirak @JessicaAnnEng The “closed loop” thing is technically true, but also irrelevant, because that’s not the water use people are talking about. Commodity servers mostly dump their heat into air. The air dumps its heat via exchanger into the closed-loop water (or various other refrigerants, but water is common). The closed-loop water dumps its heat via exchanger into the outside air. That takes a *LOT* of exchanger volume on the water-to-outside-air side. To increase efficiency, they use something like a swamp cooler for the outside air. *That* water is used once and evaporates.
Incidentally, this is also why “datacenters in space” is such a dumb idea.
@bob_zim @nirak @JessicaAnnEng What you just explained is (of course) a part I have never known re: closed loop systems. And of course its easy to see 2 reasons why that is not brought up
@Catwoman69y2k @nirak @JessicaAnnEng It reminds me of the “Masks don’t work!” goons from a few years ago. They take one technically-correct aspect of the thing being discussed, and treat it as the only thing being discussed.
Like … sure, datacenters don’t *need to* use the pre-chillers for the outside air all the time, but in practice they do use them about 80% of the time, because they would need several times as much exchanger capacity without the pre-chilling. Especially in arid environments like New Mexico and Utah.
@bob_zim @nirak @JessicaAnnEng Very much agreed about the parallels with some of the COVID anti-maskers that were so proud of thier ignorance
When it comes to that 80%, I wonder how that compares since we got all this talk about "hyperscaling" . One big issue with alot of the techbros' "technological advances" is that they often want to scale things up way faster than what is reasonable. Sometimes they just assume they can scale up at a rate that isnt even physically and/or technologically possible. Common issue in people who have that amount of hubris driving thier thought processes.
Oct 23, 2026, 6:24 AM@bob_zim @Catwoman69y2k @nirak @JessicaAnnEng closed loop makes the argument worse, because all the water consumption we're talking about is for retopping and evaporative cooling of the closed loop, which CANNOT be closed.
@f4grx @Catwoman69y2k @nirak @JessicaAnnEng Exactly. The closed part does need to be topped up occasionally, and it does have some nasty biocides in it, but it mostly doesn’t matter. If they didn’t use water, they would use R-134a or some other refrigerant for the whole coolant loop. The issue is those evaporative systems they pretend they don’t use.
For anybody unfamiliar, these systems are also called pre-chillers, and they operate like swamp coolers, as I mentioned. They humidify the outside air before it passes through the heat exchangers to take heat from the coolant loop. Humid air carries more heat than dry air, so this lets them dump the heat with much lower airflow, meaning they can handle a given heat load with far fewer exchangers.
Power plants have those big steam towers for the same reason. Non-photovoltaic electricity is almost exclusively generated through the movement of heat. Water is heated into super-pressurized steam, the steam turns a turbine, then the steam cools back into water. The limit on the power a given facility can generate is almost always its ability to dump the heat from the steam. Same with the work a datacenter can do.
@Catwoman69y2k @nirak @JessicaAnnEng Did they know that the only reason golf courses get less attention is because they have fewer immediatly vulnerable angles, and that a lot of people actually do legitimately hate golf courses?
@Epic_Null @nirak @JessicaAnnEng I cannot say that i know. This is because a) I dont know this person well enough and I do my best (in life) to ask before I assume and 2) Im usually only listening in these conversations because, at this point, everyone (pro-AI or against AI) have their mind made up. Re: those who are pro-AI...many of them are too cooked to see any reason
...but the thing you are wondering about is definitely something I have mulled over in my head. Additionally, while I have hated previous iterations of "rich people push into a community and start sucking resources exponentially higher than the everyday-Joe (in those surrounding areas)..... I dont know if comparing the environmental impact of AI data centers vs the environmental impact of golf courses is like comparing apples to oranges. My reasoning for that is that I could guess that AI data enters' environmental impact is 100x worse than the golf courses (to be clear, I say "I could guess" because I do not know for sure. )