Hyperscalers have come under scrutiny for their impact on water quality and availability.
See full article...
See full article...
How data center operators are tackling their water use problems
- Thread starter JournalBot
- Start date
There are numerous underground, abandoned mines that are flooded in the Midwestern and Appalachian coal fields. Why not use them for geothermal heating?
Because of their honeycomb tunnel layouts, each one offers multiple square miles of surface area for heat transfer between the mine walls and the water in these mines.
It’s too dangerous for humans to reenter abandoned mines but vertical wells slowly pumping large volumes out and then back in should be doable.
You could put solar panels on the strip mined areas that are often nearby. There’s also natural gas in these areas.
Because of their honeycomb tunnel layouts, each one offers multiple square miles of surface area for heat transfer between the mine walls and the water in these mines.
It’s too dangerous for humans to reenter abandoned mines but vertical wells slowly pumping large volumes out and then back in should be doable.
You could put solar panels on the strip mined areas that are often nearby. There’s also natural gas in these areas.
Upvote
29
(34
/
-5)
I've heard about such projects before and its a lot of smoke and mirrors.
Big Hyperscaler hires a technology company to help conserve water with an irrigation company or water agency. Maybe it works, maybe it saves some number of gallons per year. But the technology company runs back to Big Hyperscaler and misrepresents the amount of water saved (because they get paid by the gallon), and that misrepresentation is used in their sustainability reports.
Upvote
93
(96
/
-3)
ubercurmudgeon
Ars Praefectus
...and if all that impinges on profits too much, they could always start campaigns about limiting one's "water footprint", subtly shifting the responsibility onto individuals and consumers, meanwhile funding politicians who'll tear up any water use regulations that apply to them.
Upvote
80
(83
/
-3)
Any business is going to try to get away with paying out the least amount of money possible for expenses. Due to the political and regulatory environment in America, actually avoiding damage to the environment will, for the foreseeable future, be more expensive than merely making empty claims to that effect. So making empty claims is what they'll do.
Upvote
46
(48
/
-2)
One of my responsibilities is to try and figure out potential water demand impacts for our utility from data center implementations and it is proving to be a challenge. I want to just recommend a wait and see approach but there is the threat that they can build these centers out faster then we can upgrade our infrastructure.
Issue one is "how real is it", as in are we actually going to see any centers get built on our network. There is all sorts of talk, a lot of bluster, but I am not seeing much real progress. Doesn't mean it isn't real and there is a lot of pressure for us to not be the development constraint. This also happened with hydrogen, BIG water users but none of the generation facilities got built (they were much more transparent on their needs too and had at least gotten as far as to secure land).
IF the datacenters are real, well I am getting huge variability in water use expectations and no clear commitments. On the low end, I have been told to expect water use that mostly supports sanitary needs (so a warehouse essentially), on the high end we need to build new transmission to meet their ask. Everybody keeps saying "closed loop water cooling" to me like a mantra but are leaving it up to me to figure out how that might translate into demand.
Issue one is "how real is it", as in are we actually going to see any centers get built on our network. There is all sorts of talk, a lot of bluster, but I am not seeing much real progress. Doesn't mean it isn't real and there is a lot of pressure for us to not be the development constraint. This also happened with hydrogen, BIG water users but none of the generation facilities got built (they were much more transparent on their needs too and had at least gotten as far as to secure land).
IF the datacenters are real, well I am getting huge variability in water use expectations and no clear commitments. On the low end, I have been told to expect water use that mostly supports sanitary needs (so a warehouse essentially), on the high end we need to build new transmission to meet their ask. Everybody keeps saying "closed loop water cooling" to me like a mantra but are leaving it up to me to figure out how that might translate into demand.
Upvote
56
(56
/
0)
You make a good point Alphabet, you are now forbidden from using dirty energy in your new datacenters. You want to build a new datacenter, use green energy. If that means that less people get to make AI slop for cheap, I see that as a win-win
Upvote
52
(56
/
-4)
Upvote
0
(11
/
-11)
I feel like you're being put in a bad spot there-- It's surely not your responsibility to manage & predict the system architecture of your potential users... that's like asking the Gas Station down the road to predict the fuel economy on my next car before I know what I'm buying.
Upvote
21
(22
/
-1)
there are probably some water experts around here and I’m genuinely curious. Presumably evaporative cooling increases humidity, and some portion of that probably returns as rain. Are there studies on amount?
I imagine it’s regional and probably low.
I imagine it’s regional and probably low.
Upvote
16
(17
/
-1)
Except...Those abandoned mines aren't just empty caverns. Pumping liquid into them likely means you're going to dissolve a lot of, more often than not toxic, minerals, that will then leach out elsewhere
Also...you need water to pump. Which isn't just everywhere
Also also, you risk destabilizing geological formations causing problems to above ground structures.
TANSTAAFL.
As to OP article where:
Pledges are just that. These "pledges" are as useless and meaningless as "carbon credits" have been. Toothless promises with farcical follow through that all but never accomplishes anything IRL. Just a whole bunch of performative paper shuffling and corporate greenwashing to fool those who don't want to watch too closely.
Upvote
29
(31
/
-2)
Ugh. Read this and thought "ars wouldn't be this technically sloppy" then scrolled up to check the byline.
Upvote
13
(13
/
0)
They're not, but their PR department is on top of it.
"There are no water issues with our building, I can assure you of that. Yes, indeedy. Permit please? Oh, almost forgot. Here's a lot of money for your representatives."
Upvote
23
(23
/
0)
This whole issue is just a massive failure by policy makers and regulators allowing these installations to use as much water as they want with complete disregard. The thermal engineering challenge here is a solved problem and if all these bozo's really believe they can run an orbital datacenter, then running a datacenter on Earth with a closed cycle water loop is child's play.
Upvote
22
(23
/
-1)
RickRoyLeonPrisZhoraRacha
Ars Scholae Palatinae
None of these companies would do a study on the evaporated water and its climate effect down wind. Remember that past shuttle launches created clouds that changed weather. Now with all the moisture that "wasn't a lake or resevoir" evaporating...that moisture will dump somewhere, and that means flooding.
Just like North America burning coal back in the day, would create acid rain in Europe.
Just like North America burning coal back in the day, would create acid rain in Europe.
Upvote
-6
(4
/
-10)
If you pump water in and out you’ll have a closed loop
By definition, a flooded mine is full of water
Just about any subsidence that was going to occur happened during active mining. It’s not unheard of, but not much happens after abandonment.
Pumping large volumes very slowly should not destabilize the existing supports (wooden cribbing, steel roof bolts).
I was involved with a large, long distance water pipeline. It moved water slowly to minimize erosion of the pipe and was designed for decades of use.
You’re not changing the water level
Eventually, much of your mine is going to collapse in bits and pieces over many, many years whether or not you’re pumping water
.
Free lunch? Well, it cost someone else many, many millions to develop your tunnel network.
Last edited:
Upvote
-3
(11
/
-14)
I'm certain (a lot of) people have looked into this already, but why not use things other than water to cool the centers? Is it just because water is just so much cheaper than alternatives? Is there really nothing else you could use that wouldn't be better?
Upvote
10
(10
/
0)
1) No you won't have a closed loop. Mines and geologic features are not impermeable. See what happens when fracking companies inject waste water into wells...it has a habit of bubbling back to the surface.
2) If you move slowly is the key. You do realize you're talking about an industry whose entire motto was literally "move fast and break things"...right?
Upvote
20
(25
/
-5)
I read a news story about the controversial one in Utah that is 40,000 acres and can potentially consume 9 gigawatts of power.
To put that in prospective, I am currently working at a data center that supports large scale traditional computer processing (websites with petabytes of data, millions of transactions a day), and that data center fits on a 1 acre lot.
To put that in prospective, I am currently working at a data center that supports large scale traditional computer processing (websites with petabytes of data, millions of transactions a day), and that data center fits on a 1 acre lot.
Upvote
23
(26
/
-3)
So along with taking people's jobs, increasing wealth inequality, driving up energy bills, being a blight on the land, it'll also take one of the only two things humans need to survive...water. Arguably it'll also impact the other one by using vast amounts of fossil fuels for energy and thus making our planet even hotter which turn makes food harder to grow.
What a win for us all
What a win for us all
Upvote
27
(28
/
-1)
"Lawrence Berkeley National Laboratory predicted in a 2024 report that hyperscale data centers could consume up to 33 billion gallons of water by 2030 if they relied heavily on evaporative cooling."
33 billion gallons of water for data center cooling 2030 is certainly worth the attention it is getting. It is also important to look at other large water consumers for scale. For example, Almond agriculture in California alone eclipses data centers demands with it's 1.6 - 1.9 trillion gallons per year. https://www.c-win.org/cwin-water-blog/2022/7/11/california-almond-water-usage) @ 326,000 gallons per acre-foot.
We all need to be better stewards of our natural resources.
33 billion gallons of water for data center cooling 2030 is certainly worth the attention it is getting. It is also important to look at other large water consumers for scale. For example, Almond agriculture in California alone eclipses data centers demands with it's 1.6 - 1.9 trillion gallons per year. https://www.c-win.org/cwin-water-blog/2022/7/11/california-almond-water-usage) @ 326,000 gallons per acre-foot.
We all need to be better stewards of our natural resources.
Upvote
25
(28
/
-3)
It’s been many years since I was in a coal mine or looked at a mine map, but a long-operating mine might spread underground over an area of 5 to 10 square miles.
If continuous miners were used, it would consist of long, parallel 15’ wide tunnels with 15’ cross tunnels and 15’ square pillars of coal left in between tunnels to help hold up the roof. The tunnel height would match the seam height: typically 3’ to 8’
On a map, this looked like a honeycomb
Caveat: it’s been a long time since- I may be off 50% on the tunnel widths and coal block sizes.
When the section was fully developed, miners would “retreat” the mine, blasting pillars one at a time and retrieving some of the remaining coal. The mine roof behind the remaining roof would collapse but still leave large voids in the “gob” since you’d still have your wooden cribbing and/or roof bolts in old tunnels.
In some cases, the mine may have closed for economic reasons before the pillars were blown, leaving intact tunnels.
Starting in the 1980s some mines in these 2 regions moved to long-wall mining. Those mines have potential, too, but perhaps less than what I’ve described. See Wikipedia for more information
If the mine was “below drainage” (approximately the elevation of the nearest creek), it would fill with water after the pumps were turned off.
There’s enough cooling potential to support a lot of computing power.
If continuous miners were used, it would consist of long, parallel 15’ wide tunnels with 15’ cross tunnels and 15’ square pillars of coal left in between tunnels to help hold up the roof. The tunnel height would match the seam height: typically 3’ to 8’
On a map, this looked like a honeycomb
Caveat: it’s been a long time since- I may be off 50% on the tunnel widths and coal block sizes.
When the section was fully developed, miners would “retreat” the mine, blasting pillars one at a time and retrieving some of the remaining coal. The mine roof behind the remaining roof would collapse but still leave large voids in the “gob” since you’d still have your wooden cribbing and/or roof bolts in old tunnels.
In some cases, the mine may have closed for economic reasons before the pillars were blown, leaving intact tunnels.
Starting in the 1980s some mines in these 2 regions moved to long-wall mining. Those mines have potential, too, but perhaps less than what I’ve described. See Wikipedia for more information
If the mine was “below drainage” (approximately the elevation of the nearest creek), it would fill with water after the pumps were turned off.
There’s enough cooling potential to support a lot of computing power.
Upvote
2
(8
/
-6)
The thing about THIS water use is that it's fresh water. Something the world is already seeing issues with. The supplies themselves are under duress from sea water intrusion turning fresh water brackish, and it's only going to get worse. In areas where water is depleted, evaporative cooling only wastes that water. It's not going to becoming back down IN THAT REGION where it's needed. So areas seeing less rainfall now from climate change are going to get far worse water issues even without data centers fucking with the soup.
And corporate pledges?
Seriously?
That's like the 5 year old promising to leave the open cookie jar alone.
PLEASE let AI die before that shit hits the fan.
Upvote
34
(38
/
-4)
The only water use that matters is when they use it in evaporative cooling. And they use water by the millions of gallons per day for that (1000 million per 365 days is 2.7 million gallons per day). What are you going to evaporate that you could just release into the atmosphere, and that also is available by the millions of gallons per day?
Upvote
12
(13
/
-1)
Are pledges to use less water or make investments to replenish more water than the AI company consumes legally binding statements? Will the fine for not following through and using up all the water and or massively increasing heat and other pollution in the local environment be met with a one dollar fine?
Upvote
12
(12
/
0)
Capex for Alphabet in 2019 was $25B. They are expecting to spend $180B this year.
Upvote
-4
(1
/
-5)
There’s no comparison.
Fracking and waste injection involve extremely high pressures, so of course they’re violently pushing toxic water into the rock. They’re also wedging open cracks with stuff like sand.
When the mine previously filled with water, it reached static equilibrium with the surrounding groundwater. You’re not increasing the pressure in the mine by more than several psi at the return wells. So toxic migration is not going to change from what you’d already experience.
“Move slowly” - that was the water flow I was talking about, not tech bros.
Last edited:
Upvote
10
(14
/
-4)
There's something that just clicked in my brain about this cooling water usage:
Out of all the common materials and substances we have, evaporating/boiling/vaporizing water takes a ridiculous amount of energy. So much energy that if you live in a dry climate where water can evaporate quickly at room temperature, you can easily use it for home cooling even in some relatively water-scarce regions of the U.S.
Council Buffs, Iowa (mentioned in the article) isn't exactly wetlands, but (correct me if I'm wrong here), it's certainly not a high-desert where only plants specialized for water scarcity grow. And Ashburn VA, pictured in the article's headline, is actually in a wet climate region. The only industry that would normally make places like those blink at your coolant water usage would be power plants, the kind that power cities and all of the industries within them. This energy doesn't come from nowhere, and if it were solar generated, we would see fields of panels on the roofs and surrounding area.
For that Google Facility, it's "over 1 billion gallons of water per year" is over 310MW average (if I did my math right). Megawatts, not Megawatt hours per year or something dumb like that, their water usage removes 310 Megajoules of waste heat every second.
Normally, I would ask "Where the fuck is that energy coming from?" but I already know the answer to that question: On-site fossil fuel generators that may be in flagrant violation of environmental and pollution laws.
View: https://www.youtube.com/watch?v=5p426fSlYH4
Out of all the common materials and substances we have, evaporating/boiling/vaporizing water takes a ridiculous amount of energy. So much energy that if you live in a dry climate where water can evaporate quickly at room temperature, you can easily use it for home cooling even in some relatively water-scarce regions of the U.S.
Council Buffs, Iowa (mentioned in the article) isn't exactly wetlands, but (correct me if I'm wrong here), it's certainly not a high-desert where only plants specialized for water scarcity grow. And Ashburn VA, pictured in the article's headline, is actually in a wet climate region. The only industry that would normally make places like those blink at your coolant water usage would be power plants, the kind that power cities and all of the industries within them. This energy doesn't come from nowhere, and if it were solar generated, we would see fields of panels on the roofs and surrounding area.
For that Google Facility, it's "over 1 billion gallons of water per year" is over 310MW average (if I did my math right). Megawatts, not Megawatt hours per year or something dumb like that, their water usage removes 310 Megajoules of waste heat every second.
Normally, I would ask "Where the fuck is that energy coming from?" but I already know the answer to that question: On-site fossil fuel generators that may be in flagrant violation of environmental and pollution laws.
View: https://www.youtube.com/watch?v=5p426fSlYH4
Upvote
-1
(4
/
-5)
People in this thread are asking why they can't use alternative sources of water such as ground water or seawater instead of fresh (usually potable) water. The reason is because these other sources of water are much more costly to them either in terms of maintenance or the purification they would need to be used with their current evaporative systems. They don't want to pay that cost, they want to take advantage of the fact that potable water systems distribute the cost of water purification/acquisition across the population. To put it simply, it's not that they can't find water, it's that they can't find water at the price they want to pay it.
Just like how they it's never that they can't find a person in the US to hire, it's that they can't find a person in the US to hire at the poverty wages they want to pay.
Just like how they it's never that they can't find a person in the US to hire, it's that they can't find a person in the US to hire at the poverty wages they want to pay.
Upvote
25
(27
/
-2)
Because water is just about the best coolant in the universe, especially when the operating temperature range of thing being cooled perfectly matches the liquid phase range of unpressurised water. You pretty much couldn't dream for a better coolant for this application than water.
Upvote
15
(15
/
0)
I was wondering how they would "Replenish" the water. We can't just Harry Potter this. The Law of Conversation of Matter also has a word to say.
Because if they truly could just "Magic" water into existence they are in the wrong business.
But it's all just numbers on a datasheet. Probably crunched using the same AI.
Upvote
5
(6
/
-1)
It’s a nice idea, but probably not workable. I’m only familiar with Appalachia, but acid mine drainage is a major issue here. The mine water can be both toxic and corrosive. Even if you were only doing heat exchange into that water, any equipment or pipes with reasonable heat transfer could get damaged being submerged in a toxic acid bath.
Flooded mines are just generally treacherous too. Any activity in them could release toxic gases or destabilize parts of them. If you did breach or destabilize old, flooded mines, you could trigger new/worse mine drainage which is already a major problem for PA/WV streams and rivers.
Upvote
16
(18
/
-2)
Assuming an inlet temperature of 50 degrees F (15 C) and an return temperature of 104 F (40 C), a system moving 1 cubic meter of water per second will remove 160 MW
A big abandoned mine may hold 50 million cubic meters of water in its tunnels and voids.
Upvote
-1
(3
/
-4)
Water is actually phenomenal at this job due to its high specific heat and the energy required to evaporate it. The only other way to really get this waste energy out of these plants is to try to dump it into the surrounding air directly (due to the scale of energy involved, I don't think you can just dump it into the ground and expect it to cool off before you start causing problems). It's hard to put energy into the air because, unlike water, air's an insulator, has a low specific heat, and is already a gas so you can't take advantage of the energy required to push it through a state change.
How all of this works is squarely in the topic of thermodynamics, so resources dedicated to learning about evaporative cooling and air-conditioning are going to teach you much better than off-the-cuff comments, but the relevant material properties at play are:
- Specific heat (how much energy it takes to increase a material's temperature by a single degree)
- Enthalpy of vaporization (how much more energy it takes to vaporize a liquid that's already at it's boiling temperature)
- Thermal Conductivity
Upvote
17
(18
/
-1)
Use stainless steel piping: already widely used in power generation.
Abandoned mines of any sort are always treacherous and as I noted, you can expect periodic local rockfalls with or without pumping. Water currents right at the wells might be on the order of 1-2 mph. Away from the wells, you’re looking at just a few inches per minute at most.
You’re slowly moving water over such a large area that localized roof collapses won’t cause much trouble until enough of them accumulate in a few decades.
Unless it’s a freshly closed mine, most of your likely roof falls have already happened
The gas you’d worry about would be methane; most of that leaked out of nearby rocks when first mined. You’ll still get occasional spikes in dissolved methane; you can scavenge that at the top.
The only other toxic gas I can think of is carbon monoxide but only if you previously had a mine fire. You’d know about and avoid those sections ahead of time.
Last edited:
Upvote
-2
(6
/
-8)
My biggest problem with these data centers is the local officials. They are public servants - everything they do is suppose to be public knowledge yet they are signing NDA's?!?!?!? How do they get away with signing NDA's???? They get away with it because they know their bosses (you and me) aren't going to do anything except bitch and moan about it. College football is more important. But we are at the beginning of the pain stage of history repeating itself (currently known as enshitification) and while the pain stage has historically been the longest of the three stages at least some people are starting to look around and figure things out. We are at least getting closer to the time when college football will take a backseat to saving the constitution.
Upvote
8
(9
/
-1)