I can totally see why a company wants to keep this info secret.
Competitors would really like to know.
The electricity use is really substantial though, but that's harder for people to visualise so gets less media attention.
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)
> pretty close proxy for how much computation is happening.
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.
You could spend more electricity if needed to up the airflow to get the same cooling power without humidifying.
https://web.archive.org/web/20200604033055im_/https://www.nr...
That PUE of 1.028 is really good ( https://en.wikipedia.org/wiki/Power_usage_effectiveness ). And even with all of their closed loop parts of the system, they're still needing to reject heat from the cooling towers.
https://www.nrel.gov/computational-science/reducing-water-us...
> The initial design of the data center used evaporative coolers to eliminate the added expense of energy-demanding chillers. However, while the cooling towers were more efficient and less expensive, they would consume approximately 2 million gallons of water annually to support cooling of the IT load—approaching an hourly average of 1 megawatt.
Industrial scale closed loop cooling is relatively recent technology.
> In August 2016, a prototype thermosyphon cooler—an advanced dry cooler that uses refrigerant in a passive cycle to dissipate heat—was installed at NREL. The thermosyphon was placed upstream of the HPC Data Center cooling towers at the ESIF to create a hybrid cooling system. The system coordinates the operation for optimum water and operating cost efficiency—using wet cooling when it's hot and dry cooling when it's not.
It is a goal though... https://www.microsoft.com/en-us/microsoft-cloud/blog/2024/12...
> Although our current fleet will still use a mix of air-cooled and water-cooled systems, new projects in Phoenix, Arizona, and Mt. Pleasant, Wisconsin, will pilot zero-water evaporated designs in 2026. Starting August 2024, all new Microsoft datacenter designs began using this next-generation cooling technology, as we work to make zero-water evaporation the primary cooling method across our owned portfolio. These new sites will begin coming online in late 2027.
More northerly data centers are likely able to achieve a lower WEU. Again from Microsoft - https://datacenters.microsoft.com/sustainability/efficiency/
It’s an explanation of why so many companies suddenly appeared to go “woke”, or why they did a complete 180 when the political climate changed. Even powerful companies like Apple must grovel for favor.
Ok, when we're considering how much water a person uses, are we going to include the water used to grow the almonds you ate? Because agriculture is going to dwarf anything that data centers use.
Beef too. It uses the same amount of water but people eat 30x as much annually.
Beef and milk are harvested ready to eat. Vegan substitutes are all highly processed. Processed food consumption is associated with greater cancer and diabetes risk.
It isn't meaningless due to industrial farming. Chickens and pigs are even more likely to be industrially farmed than cattle are.
If we lowered our meat consumption by about 90% then we wouldn't need to industrially farm meat and the 10% would be much more ecologically justifiable.
Then the problem is that the soy we've replaced our meat with is industrially farmed...
If you want to talk about pigs or chickens, that's an entirely different story. Those do get raised full life cycle in factory-like industrial facilities. But those aren't cows.
Compared to an almond? Who the fuck eats almond steaks? It's a nonsensical comparison. If you want less fat and more protein per calorie, chicken beats beef. Chicken also has a lower water and carbon footprint.
> Vegan substitutes are all highly processed
Beans aren't "highly processed". Learn to cook and you'll understand that there are options besides processed food for vegetarians and vegans.
This book is science, front to back, cementing the idea that animal products are not ideal for human health. https://www.amazon.com/Food-Revolution-Your-Diet-World/dp/16...
And also, what? You'll die if you don't eat meat today? Because that's what "immediately" means. That's news to all the world's vegetarians.
Most people attribute this difference to the GDP/capita ($89,600 vs. $12,132 PPP) or the number of doctors per capita (36.082 vs. 7.265 per 10k), not the diet.
But agreed that diet is also not the only link to longevity. Although a Mediterranean diet is higher on plant based foods, considered healthier, and those countries do have better life expectancy than USA.
https://peterattiamd.com/rhondapatrick3/
The protein doesn't necessarily have to come from beef, although that is one of the highest quality sources in terms of digestibility (for most people) and essential amino acid balance.
Presumably, after a cow is done being used for milk, it can then be sold for meat.
So you wouldn't want an old cow steak, but a stew or burger made from an old cow can be awesome.
That said, Italy and France are known for smoking a lot, which supresses the appetite. Your original observation was swiss though (land of milk, chocolate and cheese)?
Not exactly the same, but can I interest you in a caramel-waffle-oat-milk latte?
The name "almond milk" has been used since at least the 1500s.
While almond milk is an incomplete substitute for bovine mammary secretions, it is so much closer than candy corn that it has been used as a substitute for the last 800 or so years, and shared the "milk"-ness in the name before we had an English language:
The word “milk” has been used since around 1200 AD to refer to plant juices.
This makes this use of the word older than English people spelling the thing chickens lay as "eggs" rather than "eyren": https://www.youtube.com/watch?v=rZ5znvym68k
The Romans called lettuce "lactūca", derived from lac (“milk”), because of the milky fluid in its stalks: https://en.wiktionary.org/wiki/lettuce
Similar examples abound.
For example, I grew up in the UK, where a standard Christmas seasonal food is the "mince pie", which is filled with "mincemeat". While this can be (and traditionally was) done with meat derivatives, in practice those sold in my lifetime have been almost entirely vegetarian. The etymology being when "meat" was the broad concept of food in general: https://en.wikipedia.org/wiki/Mincemeat
Further examples of this: today we speak of the "flesh" and "skin" of an apple.
Personally, I don't like almond milk. But denying that something which got "milk" in its name due to it's use as a milk-like-thing, before our language evolved from cross-breeding medieval German with medieval French, to argue against someone who said "Insofar as it is used for milk", is a very small nit to be picking.
Candy corn in my coffee wouldn't taste anywhere near as good
Almonds also consume 15% of California's water. But California produces 80% of the world's almonds. We're talking about an order of magnitude difference in water consumption, almonds are far more efficient and beef is both far more wasteful and far more common.
Animal agriculture is wildly inefficient and honestly it's not surprising because you have to keep living moving animals around for it.
California agricultural water is so fucking cheap, you can buy foreign land, start a farm, grow a bunch of grass, and ship it over to your country.
And that's cheaper than just growing grass locally.
That's insane
Most problems California has are the same: Systems that were initially designed and built a hundred or more years ago to support a state of like a hundred people, and an utter refusal to update those agreements because it would slightly inconvenience some really wealthy farms.
Growing Almonds and Alfalfa in California would be fine if they paid market rates for the water, and would therefore be more conscientious about using it and not wasting it, and that would dramatically improve the water situation of California and upstream places.
But it's way cheaper to pay for people to run absurd narratives on Fox News to make it a culture war issue so that you never have to care.
https://www.youtube.com/watch?v=aux22FHTFXQ
The situation is so fucked up. It's a war of the rich against the richer. Wealthy farm corporations all have lobbying groups, and instead of lobbying for a more free market distribution of water, where they could have all they want as long as they pay for it, they run political campaigns to ensure California never reforms it's water rights system and continues to die of thirst while giving away 90% of it's water for rates decided 100 years ago. All the political agitation about water in California is over an absolute minuscule fraction of the water distribution, because actually fixing the problem would mean these farms paying market rates for water, which they do not currently do.
A microcosm of the US problem basically.
It feels reasonable that we should have the same detail of information for data centers.
You seem to think there is only one way to count. That's completely wrong. The important thing is whether you are clear about what you are counting and why.
Your comment is odd. But let me be the first to give you permission to count things how you want. Just make sure you explain the criteria and reasoning. Have a good day.
Likewise, the power company is going to generate electricity regardless of whether a data center is there or not. The power company has various means of generating electricity available which use more or less water. The amount of water used in generating electricity is attributed to the power company, not the data center.
Contrary to your belief that it is about clickbait, it's actually just about how to accurately evaluate environmental impact of data centers backed by science and basic logic.
1. Numbers from a pro-almond group: https://www.almonds.org/why-almonds/growing-good/water-wise
I actually wonder if there is not single moderately sized industry that does not have some interest group...
They're basically necessary to do business in an advanced society that has rules governing every aspect of life. Those rules (even if good) often have unintended consequences and advocacy groups can help ensure that their industries are considered when rule-making.
lobbyists will advocate for taking the water right from under the noses of the workers, and the workers will turn around and praise their employer for maintaining their jobs... it's often some kind of perverse shell game
at the end of the day the owners fly off to wherever with a pile of money and the workers are left without jobs or water — these false dichotomies of "if it weren't for lobbyists all jobs would be regulated away" is often used to disenfranchise people from actually changing these systems
It only becomes advocacy when you take the money out of it.
>In a series of secret meetings in 1994, the Resnicks seized control of California’s public water supply. Now they’ve built a business empire by selling it back to working people.
https://perfectunion.us/how-this-billionaire-couple-stole-ca...
That’s a good idea. Like nutrition facts but for everyday economic climate decisions.
It’s also absolutely true that “agricultural usage dominating data center usage” is a dirty little secret that a lot of people are very, very incentivized to keep secret. Amazon can’t outright say that, because uh whutabuht mah poor farmers.
The water normally continues to exist, so presumably it's some other resource we are using. This may seem pedantic but it's not - raw groundwater, or unprocessed grey water is not potable as in "water a person uses" for drinking, which is a subset of the water a person uses overall (directly in showers etc, or by proxy in bought products, building materials etc).
In each case, water is more of a "carrier" for some other resource or property. If in CA the almonds go through a lot of water (is this due to perspiration? i.e. their cooling mechanism?), the water will still create clouds that I presume increases rainfall elsewhere? In fact, the water now holds more energy (from solar) that might be useful somehow.
Similar comments around "land usage", entirely depends on opportunity loss otherwise.
Will it though
I am critical of this metric though, since the water isn't really lost in many cases. Especially if datacenters use water never meant for consumption.
If you look at it as a power generation problem it become much more plastic. That is of course as long as the water doesn't get expended in regions that lack it.
Although if you want to compare datacenter usage to agriculture, you could say that one is more essential than the other. Even if modern agriculture is a high tech industry that uses datacenters.
The real waste is the energy required to produce, clean, and transport the water that is being “wasted.”
sustainability, availability and maximum marginal price matters, just as with electric power generation
If your tap water comes from a river and flows back to a river, leaving it running mostly just wastes energy.
If it is the power company, then the relevant metric is the amount of power consumed, as it would be up to the power company to find a more water-efficient way to produce power.
That does create a grim implication that the value of a person must be weighed against the value of the data center, and that perhaps one is a less useful eater than the other.
One gallon of cow milk uses more water than one gallon of almond milk.
Water use for all of AI is inconsequential compared to agriculture.
In addition, water is almost never wasted, only moved around.
Energy is the important input.
Unfortunately, media sound bites can't distinguish meaningless water usage from meaningful usage.
Technically yes, vapor goes to the atmosphere etc. But in certain areas, data centers are effectively removing water that was previously used for farming.
https://www.context.news/ai/thirsty-data-centres-spring-up-i...
In addition, increasing human productivity through technological innovation is the only thing that ever let us escape the malthusian trap.
How so?
[1] https://www.npr.org/2023/08/09/1192996975/amid-a-water-crisi...
But there is definitely an impact to pulling too much water out of one place too fast, which must be ethically addressed when building datacenters.
Beyond potential impacts to other local residents in terms of reduced access to local water or price increases to meet demand, there is also the danger of disturbing the local environment and reducing the quality of local water.
We've seen the stories about increased sediment in local residents' water supply after a new datacenter moves in next door, but I'd like to share an example from my own city.
Our city is known for its soft water. It's one of the only nice things about the city. Well, we have a local Exxon plant that sits right on top of the highest point in our water table. For oil refinement, the purer the water, the better.
For decades, the vacuum created by this plant's continuous suction has created fault lines that have been leeching increasing amounts of sediment and salt water into our water table, ruining the drinking water and in some cases making it entirely undrinkable.
"In Louisiana, industry uses more groundwater than in any other state except California, according to the US Geological Survey. For decades, industrial users have been able to pump water out of Baton Rouge’s aquifer effectively without limitations – no withdrawal caps on individual wells and no metering requirement"
When you try to push against them and raise awareness, you get discredited or sued. They are dedicated to protecting their unfettered access to our clean drinking water through whatever means necessary. I do not for a moment think Amazon any different. They are an ethically bankrupt company.
https://www.theguardian.com/environment/2022/jun/08/louisian...
If you're talking about the New York Times Article "Their Water Taps Ran Dry When Meta Built Next Door" described in https://andymasley.substack.com/p/the-ai-water-issue-is-fake , that NYT article was so misleading I'd call it basically a deliberate lie. The article was about a household that used well water and started having more problems with sediment in their well water when Meta started constructing a data center within a mile of them (that was not operating yet because it wasn't done being built yet). It's unclear if the construction of the data center was actually related to their sediment issues, and even if it was, the fact that it was a data center being constructed as opposed to some other type of large building was irrelevant.
That's why I thought to offer an example from my own backyard that I can verify myself, and has a much clearer story and is also in a non-datacenter industry as to avoid hype and focus on the importance of reasonable water usage restrictions.
You can literally day the same thing about energy. Electricity is never wasted, its just different afterwards.
Water use absolutely does matter, because „being moved around“ in the quantities we do, is far from trivial. Its also different than agriculture. Agriculture still has a somewhat closed local water cycle while the water used for evaporative cooling is basically gone locally.
It matters a whole lot for where you are. If you‘d do evaporative cooling with salt water from the ocean, nobody would bat an eye. The problem is that it is done with fresh water, which is becoming increasingly scarce in an increasing number of regions around the world.
> The company has negotiated a pair of agreements with The Dalles city officials that would significantly reduce property taxes Google must pay on the new development and secure for the company the water it needs for its expanded operations.
> The deal to deliver groundwater to Google has drawn skepticism from members of the public who’ve grown wary of Oregon’s water stability in a changing climate, and that suspicion was on full display at a recent City Council meeting.
> ...
> “Without this agreement, [Google] or any other industry could use those wells as they wanted just as the aluminum plant previously did,” Anderson, the public works director, told the Council.
> Anderson said the amount of water that can be withdrawn from The Dalles groundwater aquifer annually without causing a decline is 5,500 acre-feet per year. Only about 1,800 acre-feet per year are being drawn out of the aquifer currently.
---
Most surface water has a "you can't drain this" compact. For example, the Great Lakes compact - https://en.wikipedia.org/wiki/Great_Lakes_Compact
Part of the sticking point for the Foxconn mess in Wisconsin was the water use within the watershed of the Great Lakes. https://observatory.sjmc.wisc.edu/2018/05/10/great-lakes-wat...
Microsoft's plan use: https://www.wpr.org/news/microsoft-data-centers-8-million-ga...
> The data obtained by WPR from the city shows the first phase of Microsoft’s data center campus would use a peak of 234,000 gallons per day or 2.8 million gallons per year. Under subsequent phases, the campus would use a peak of 702,000 gallons per day or 8.4 million gallons annually.
https://www.wpr.org/wp-content/uploads/2025/09/MKE-Regional-... shows the demand and discharge request.
Isn't there some better way we can, perhaps, turn some of the heat back into something useful? Maybe heat a building? Or turn it back into electricity. It doesn't have to be an efficient conversion because it's now 100% wasted.
I think it could be higher with germanium semiconductors, as you could run them at a higher temp and get superheated steam
I'm a big fan of district heating, but it's something that needs to be built before the datacenter is. It also doesn't really work well if the datacenter isn't in an already cool region.
> Or turn it back into electricity.
The temps aren't high enough to do that easily. You need boiling water to generate electricity, and chips don't like running at or above 100C.
It's possible you could use a heat pump to turn hot water into boiling water, but that will stop working when temps get out of band. You might be able do it with a sterling engine, but you'd, ironically, need a supply of cool water to keep those running.
We should question the motives of whoever orchestrated a "story" out of this non-story and is pushing it in the media. It obviously isn't being done in good faith.
From https://www.usgs.gov/water-science-school/science/total-wate...:
> Water use in the United States in 2015 was estimated to be about 322 billion gallons per day (Bgal/d), which was 9 percent less than in 2010.
It doesn't seem to be very much water at all.
If so, why?
If not, does it matter how much water is used?
Even if seawater was easy to use and datacenters were near the shore, it would produce very saline brine which would be difficult to safely get rid of.
HN commentary on water use by so-called "tech" companies usually includes a number of mindlessly-parroted, bad faith "arguments"
One of these is to try to compare the new (additive) water use by non-essential data centers with existing (non-additive) water use by agriculture
Putting aside that (a) data centers are non-essential and not comparable to food, water or shelter and (b) agricultural use is not new, these "arguments" are also ignoring that (c) the so-called "tech" companies are trying to hide the data
Employees of these so-called "tech" companies might be experiencing guilt over this dishonest tactic, but not enough to make them quit
When their employer hides the data this makes accurate comparisons, e.g., to existing water use by other recipients, difficult if not impossible
Does agriculture also try to hide its water use
If it did, then HN comments could not attempt bad faith comparisons
Because there would be no data to cite
There is a plan for constructing a new high-capacity datacenter [edit: near my city]. And a lot of discussions in the media are done through an emotional tone around water and electricity usage.
The media generally frames it as if installing a new datacenter would put the neighbors in risk of not having water or electricity. I'm not arguing that a datacenter doesn't bring any problems, everything has pros and cons.
Both sides seems to be using bad faith/misleading arguments, and I thinks that's really bad because we end up with solutions and agreements that don't improve the lives of the people affected by these new developments.
I wish I'd known what was coming, and gone to the meetings to oppose it.
avalys•2h ago
Where does the water go? If they simply take in cold water and release hot water, that water is still available for other uses.
Or do data centers use evaporative cooling just like power plants?
xnx•2h ago
Yes. Not always, but evaporative cooling is much more energy efficient than heat exchange to outside air.
That said, stories about data center water use are a distraction from much bigger water consumers like golf courses and agriculture (e.g. to ship alfalfa to the middle east).
teeray•2h ago
That's something of a fallacy of relative privation. When water is scarce, all frivolous uses should be under scrutiny. The others you mention have been well-known for a long time. The current stories simply highlight a new consumer people haven't thought of before.
wahnfrieden•2h ago
neoromantique•1h ago
Yes.
Romario77•1h ago
Agriculture uses about 70% of all freshwater while datacenters are less than 0.5%
Some leaky channel will cost more than all the datacenters.
egorfine•2h ago
oceanplexian•2h ago
harddrivereque•2h ago
With agriculture, water usage is necessary as eating is not something optional and everyone needs to eat to survive. From the PR point of view, of course. We couldn't live without agriculture, as we had agriculture 20 years ago too.
Golf courses are unrelated as they don't use nearly as much water as agriculture or data centers.
PR is everything, the narrative is what makes the difference. There is a lot of hypocrisy in this field, which is why I try to avoid it, but there is also some truth in it - we really didn't need that many data centers 20 years ago.
triceratops•2h ago
"More expensive" means spending more on air conditioning. Ergo more electricity used, higher electricity demand, more natural gas burned and carbon emissions, higher consumer power prices. So a different kind of PR disaster.
harddrivereque•1h ago
undersuit•1h ago
aseipp•1h ago
teuobk•34m ago
[1] https://www.usga.org/content/dam/usga/pdf/Water%20Resource%2...
teeray•2h ago
johncarlosbaez•1h ago
mbac32768•2h ago
darkwater•2h ago
Well, first you need to cool it down in a way that's good for the environment, I presume. You should not pour down hot water in a cold river with all its fishes and plants.
LgWoodenBadger•1h ago
e.g. https://www.nytimes.com/2025/07/14/technology/meta-data-cent... https://www.bbc.com/news/articles/cy8gy7lv448o