> noted that it would be harder to reach its internal target if its calculations included “secondary” use—water used in generating the electricity to power its data centres, according to the document.
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.
To put this in context, it takes about 1 gallon of water per 1g of almond [1]. And in California's dry climate, this water comes from groundwater that doesn't renew as fast as it's depleted. So the next time someone chastises you for your non-low flow showerhead that uses more than 2 almonds of water per minute, remember to put the numbers in context.
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.
Lobbying as practiced by these advocacy groups is basically American flavored corruption. If the system were in place anywhere else we’d call it corruption, but it’s our system so we call it Lobbying.
Yes but that ship has sailed, so now you need anti-corruption corruption to hopefully corrupt the corruption so it's less corrupt.
Like, the only reason "almond advocacy" even exists is because dairy and beef are some of the most lobbied and blessed industries in America. They can practically do whatever they want, whenever they want, so fuck you, pal.
this assumes that the needs of both the workers and the owners (who pay for the lobbying) are aligned, and they're often not
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
I've built enough things to know that you don't build understanding by insulting people. You seem to disagree with my position, but are unable to find the words to defend yours.
This sounds a bit sensationalized, and I'm not sure if it's the source from which I originally learned of the issue, but:
>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.
Almond trees are stuck as almond trees forever. They can't switch to something else. Anywhere an investment in something is entrenched like this, you'll find lobbying.
More specifically, farming and urban use 50% of the water California gets (river outflows to the sea are the other main place water ends up) Urban use is 10% and farming is 40% of that total water input. Almonds alone are 7% of ALL water, almost as much as the 10% urban.
The level of subsidy the relatively unimportant crops get with their basically free water is astounding, especially considering the high urban prices of water
Tangential to the point, I think we should be careful about the almond talking point. Insofar as it is used for milk, almond milk uses almost half as much water as dairy milk, uses 1/18 the land and emits 1/5 the amount of carbon. As food, it is eaten in such a vanishingly small quantity compared to other water-hungry foods (meat) as to be insignificant. https://ourworldindata.org/grapher/environmental-footprint-m...
I thought the problem is that that growing almonds is usually done in arid regions, so the issue is not that it uses lots of water, but that it uses lots of water in areas that frequently experience droughts. This is an honest question (I truly don't know, although I suspect I can guess): are dairy farms also common in those areas?
Just for a single point of comparison, California's alfalfa consumes 15% of the state's water. All that alfalfa is going to feed cattle. California only produces 9% of the USA's alfalfa so it's easy to say that this is a tremendous amount of water.
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.
World beef production is approx. 60x almond production by mass, and that doesn't include dairy. That isn't the whole story, because cattle use more water than just what is used to grow alfalfa, but you are comparing apples to oranges here.
I was drawing a point of comparison in response to the question about almonds disproportionately using scarce water resources in arid areas, which is a different question from overall water usage. My point was that almond water usage is literally only an issue in California, and that their water usage is not that extreme given the size of the market California serves.
The comparison you are bringing up actually just gets to the heart of the issue.
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.
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.
It’s been some years but I recall that at one point probably around 2016-2017, California produced 80% of the world’s almonds. This was notable because at the same time, California was experiencing historic droughts.
Yes- about half of the total water siphoned from the Colorado river goes to cows. This number astounded me when I first read it, and I hope it has a similar effect on you. I don't like almond milk, for what it's worth, and I don't think we should ignore plant-based foods with a high climate impact, but animal agriculture is the most environmentally devastating institution we have individual power to transition from. https://www.npr.org/2024/09/09/nx-s1-5002090/colorado-river-...
GP's point works just as well if you substitute almonds with diary milk or hamburger meat though. The specific water use of almonds is indeed completely tangential
I've read here for a long time but just made my account because I have been feeling very compelled by the data surrounding the huge economic effects of the animal agriculture industry and how otherwise pro-science and pro-data people find themselves with deeply entrenched unscientific viewpoints. Should I link my Google scholar to prevent people from seeing conspiracies everywhere??
I think the almond talking point takes hold because, like a lot of complaining about LLM usage, large parts of the blame gets directed elsewhere rather than the choices that we're all way more likely to be making and could influence. Like, it's 2025. Even the people most likely to be drinking almond milk have largely moved to oat, whose water usage is great.
But it's okay. This has been solved very recently as in last week. We are going to now be getting our beef from Argentina. Not only has the prices of beef issue been fixed, it'll also fix the country's water shortage issue as a bonus. /s
Almond milk is not dairy milk, but it is absolutely "milk", in the sense of a white liquid derived from plants - a definition that has existed in English for hundreds of years.
The name "almond milk" has been used since at least the 1500s.
I don’t drink ‘almond beverage’ but given the amount of uses it has substituting milk (and the amount of people that accept them) it seems like a very relevant comparison. Maybe I’m not sofisticated enough but I’m yet to see a candy corn mlik latte be ordered.
> “Almond milk” is not milk. You know what else is less carbon-intensive than milk? Candy corn. But that is also _not milk_, and so equally irrelevant!
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.
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.
Yes: my understanding is that it’s rather common practice to at least make a best-effort estimation of all these secondary impacts.
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.
This also tends to include the rainwater that falls on the ground to grow the grass that cattle eat that would fall on the grass anyway, or that the cow farts aren't really more greenhouse emissions than the grass rotting out over the summer anyway. It's distortive.
IMO, water is a renewable resource and what should count is the use of water in scarce environments from scarce sources directly in excess of what gets renewed. If you're right by the Mississippi river and often see flooding in your region, I don't think using the water for cooling a reactor (steam as the byproduct) is necessarily something that should be considered a negative... it'll come back down as rain somewhere.
I'm not sure why Amazon is "using" so much water, assuming their cooling systems are a closed loop... otherwise, if using evaporation for cooling, like a reactor, it depends on the location, source and usage. "it's complicated"
> assuming their cooling systems are a closed loop
My understanding is that the closed-loop systems are rather new; like, "we started using these things in 2024 literally because everyone's been moaning about our AI water usage" new. I'd assume that many new constructions are starting to leverage them, but its not 100%, and existing DCs would be slow to upgrade.
And cheaper cow production, focused on grain-feeding, uses even more water than grass feeding; after spending a month in Switzerland, it's wild to see how addicted we Americans are to cheap beef.
Ha ha, but seriously unlike other nutrients the human body can't really store protein reserves for any length of time. So you do need to keep taking in quite a bit every day or the body will break down muscle in order to sustain other tissues.
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.
You need protein for sure. But unless you want to win Mr Universe you will probably get enough through regular nutrition. And there are a lot of other things besides protein you need in your food. The current “Protein in everything” trend is just a fad created by industry
How much is "enough" exactly? Can you quantify that for us? There's a huge difference between enough to avoid acute malnutrition in a young person versus enough to optimize healthspan in an older person. Did you even listen to the podcast I linked above?
It is totally possible to get enough protein (by whatever metric you choose) through regular food. And yet the reality is that many people don't.
The upper limit seems to be 1g per lb (454g) of body weight. This is for people who lift heavy regularly to build and grow muscle. The lower bound is about 1/3 of that.
A 165 lb person needs about 55g of protein daily for maintenance. This is about one cup of black beans and 3 eggs. Huevos rancheros for breakfast will cover most or all of an average person's daily protein needs and that's before even counting the protein in the cheese or tomatoes.
While vegetables don't have much protein it still adds up. A single medium tomato has 1g, a potato has 3g, 1g in an onion, 0.5g in a carrot, and so on. Eat your daily portion of veggies and you can get maybe 10g of protein. That's nearly 20% of what you need!
Carb-heavy foods such as breads and rice also have some protein. More than veggies, less than lentils, eggs, dairy, and meat. A standard serving of pasta has 7g of protein. A cup of cooked brown rice has 5g.
All of this is to say even vegans who eat a healthy, varied diet are not deficient in protein. If they are ovo-lacto vegetarians it's a layup. They may need to supplement with whey and creatine or eat some chicken a few times a week if they want to squat 3 plates.
Insufficient protein is mostly not a problem in the first-world. The problem is eating crap. Eating grass-fed steaks every meal doesn't compensate for a crap diet.
Lots of sources of complete protein on there besides beef. Don't limit yourself. A combination of quinoa, buckwheat, eggs, soy, pea, milk, whey, beans, and nuts will provide all the amino acids you need. Even beef, despite being a complete protein, is deficient in tryptophan.
I like meat. I'm not a militant vegetarian or vegan or anything like that. I dislike the laundering of people's love of eating meat as "essential for health", when it's anything but (special dietary needs aside). Eating meat is a mostly a choice.
I'll repeat again: varied diets are essential. A single protein source won't cut it. And most people get enough protein going just by raw numbers.
I don't know why you felt the need to post that little rant. No one here claimed that people should only eat beef as their sole source of protein. You're arguing against a strawman.
I'm very pro-protein - we just tend to see better metabolic health outcomes when we eat more of it. Diets with higher proportions of protein result in better body composition, less fat, less chance of diabetes, etc. I think the vast majority of people in America at least would be better off with higher protein diets.
But "Did you even listen to the podcast I linked above" is an insane thing to say in a random online conversation. No, you cannot expect people to listen to a whole podcast to make whatever random unsubstantiated point you have. Explain the important things you took away from the podcast and link it as a reference for if people disbelieve you for whatever reason.
The difference in life expectancy between the USA (where beef is heavily consumed) and India (where you may get lynched on suspicion of eating it) is 78.39 - 72 = 6.39 years at birth.
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.
At the risk of stating the obvious, meat is not the only difference in 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.
This stands against the evidence. Beef is causally linked with the largest killers of Americans, including heart diabetes, diabetes, and obesity in general. "a mostly plant-based diet could prevent approximately 11 million deaths per year globally, and could sustainably produce enough food for the planet’s growing population without further damage to the environment."
A "Mediterranean diet" is more healthy than the average American diet because it is more plant-based.
https://www.sciencedirect.com/science/article/abs/pii/S01406...https://www.health.harvard.edu/blog/with-a-little-planning-v...
I guess it depends. I have a friend who has dairy cows, grass-fed, very well cared for. He names each one and hates to slaughter them, but when they are done producing, he does take them to the slaughter house. I bought a quarter of a cow from him a few years back, and it was excellent. It did have a hint of gaminess redolent of venison or lamb, but it was delicious.
You are misinformed. I regularly eat our retired dairy cows. They taste absolutely excellent, but they were raised on pasture and ate top quality feed.
Dairy cows from commercial confinement dairies are a different matter. They eat almost exclusively grain and develop horrible health problems.
Yet the same can be said of any meat produced in a CAFO, when compared to that produced on most small farms. Garbage in, garbage out.
My guess is that they have specialized breeds and how they're raised and fed and such, so that dairy cattle wouldn't make for good beef and vice versa. They probably do use them somehow, but maybe for dog food or some similar use.
Beef is more nutrient efficient though and has better macros for human consumption. It has 4 times more protein, less fat, and no carbs. Seems to me at least Americans could do well to eat more beef and less products created from processed almonds mixed with refined vegetable oils and sugar.
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 has 4 times more protein, less fat, and no carbs
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.
"Factory-farmed beef" doesn't even exist. All cows get raised in fields for their first 12-18 months. The ""factory-farming"" is just a feed lot they get taken to for 3-6 months to eat grain before slaughter.
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.
Not all pigs and chickens are "factory farmed" but most are. And unlike cows, the term is relatively meaningful and descriptive when applied to the way those are raised. Bro heard about how pigs and chickens get raised in tiny cages never seeing the light of day then assumed that's how it works for cows too, which he now realizes isn't the case because I shamed him into looking it up.
You're going to have to cite strong sources or else this is either heavy cope or straight-up denial.
It is inconceivable that American's consume as much beef as they do, yet production has been able to scale without resorting to factory-farming. Every other commodity food is factory farmed. It's asinine to think beef is immune to that.
Bro, get out of the city or just look at a map, America is absolutely enormous. There is no shortage of land to let cows graze on. Feedlots for cattle aren't even necessary at all, they're only used to increase profit.
What's really asinine is that you have such strong opinions about a subject you know nothing about and demand that other people do research for you.
Cope harder, bro. Show me sources. Cattle is a massive industry and it should not be hard to find sources proving me wrong. Calves live on pasture for the first few months, but are then transferred to backgrounding farms, that are technically still pastures, but are fed a highly supplemented diet and have less room to roam. STFU about me not knowing what I'm talking about.
Feedlots are absolutely necessary at the levels that American's consume beef. It might look like there isn't a shortage of pasture for cows, but the truth is there is not enough land to transition to 100% grass-fed with American's level of consumption. [1]
> Calves live on pasture for the first few months, but are then transferred to backgrounding farms, that are technically still pastures, but are fed a highly supplemented diet and have less room to roam
Glad to see you finally looked it up and saw that I was right. Your attempt to save face is amusing though. From "factory farmed" to "the pastures are a bit small" is funny.
And no, feed lots aren't necessary. They make beef cheaper by making cows bigger, therefore making the industry more profitable, but if you did away with them Americans would still be eating enough beef to make urban vegan weenies seethe.
Cattle water consumption should be meaningless. If a cow is drinking water from a surface water source and breathing/sweating/peeing it out in a pasture, that's the same process that would be happening if humans didn't exist.
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...
The problem is the assumption that cows (and other similarly intensive animals) would exist in the same quantity. You'd expect something like 5x less cow-like biomass without people intervening.
I highly recommend you read up on the actual research of what you're talking about. It points to the exact opposite of basically every sentence you wrote
Not all processed foods are created equal. Almost all of the elevated health risk from processed foods comes from processed meats and sugary drinks.
https://www.nytimes.com/2024/05/06/well/eat/ultraprocessed-f...
Whole grain breads are ultra-processed, and I don't think many are arguing against those.
Beef has absolutely devastating effects on human health including elevated cancer risk, diabetes risk, dramatically higher incidences of heart disease (the greatest killer of Americans). Plant-based substitutes are scientifically shown to lead to better outcomes. Better yet, soy based whole foods are excellent for human health, contrary to the bro-science talking points. Turns out, beans are good for you!
https://hsph.harvard.edu/news/the-bottom-line-on-ultra-proce...https://www.uclahealth.org/news/article/health-benefits-soy
At least that is generally the case for calculating the ecological footprint of a person.
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.
I generally find criticism of water waste misplaced. Water’s lifecycle is cyclic. The water isn’t lost unless it is being salted or otherwise poisoned.
The real waste is the energy required to produce, clean, and transport the water that is being “wasted.”
Depletion of local water resources is a good thing to measure. But by and large, this is not what we're measuring. Instead, we're coming up with absurd statistics that imply any water put to beneficial use just disappears forever.
If your tap water comes from a river and flows back to a river, leaving it running mostly just wastes energy.
That's a fair point. If you want to calculate the real total water usage of any person, you must first invent the universe. You have to cut it off somewhere.
I find this loses the point, because agriculture is essential. If we preserve water it's because we want to keep enough of it for essentials like agriculture and such. Not that agricultural practices shouldn't better conserve water, but it's the usage of water for non-essential things that I think most people find wasteful.
Does Amazon control the generation of this electricity? If so, then I definitely think it should be included, because it is something that Amazon has control over, and could possibly reduce water use.
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.
We can. We do. It depends on what you are trying to count. If you think there is a single number that will answer everything, no. Are you going to include all the water ancestors used? After all, without that water, I wouldn't be here.
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.
There is an honest way to count and that is to count the water directly used by the person or process in question. There are many other ways to count to make big scary number for clickbait article. I suppose we could also count all the water used by data center employees to take showers as "data center water usage" but that is dishonest. Those are people who exist and they are going to take showers whether they work at the data center or not.
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.
Kind of surprised I have to point this out. Power companies do not generate the same amount of power regardless of whether that power would be consumed on their grid or not. They increase generation based on demand. Whether that power comes from high water consumption generation is based on location which determines which power sources are available in the local grid. A major part of why indirect water consumption from power generation is included in the standard WUEsource - water usage effectiveness metric - is because it makes it clear what the impact will be when assessing data center location and size. In other words, it's important to choose locations near power generation that doesn't consume water heavily. Yes, the amount of water used in generating electricity and the efficiency there is controlled power companies choices for power generation at their locations. Data centers control the location they are built in and without an estimate of indirect water usage they cannot strategize locations with lower environmental impact.
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.
A power company builds a generation plant of a certain size. That plant is going to run whether there is a data center or not. Maybe there is some incremental additional water usage if it is running at 85% capacity instead of 75% capacity but this is probably marginal compared to the plant running or not running at all.
Depending on the method of power generation, it might also need a certain consistent base load to run most efficiently, so adding in a data center that will be a consistent load running 24/7 could actually increase the efficiency.
What does that mean though: "water used"? The only actual use of "water" is an E = mC^2 conversion to raw energy, or a chemical/location change such that it's locked-up/removed from availability permanently.
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.
I find it really odd this recent push for discussions around the development of new datacenters.
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.
One went in 1/4 mile from my home a couple years ago. I ignored the notices of development because I thought it was far enough that it wouldn't affect me, but it blocks the view of the mountains that I used to enjoy, and sometimes I can hear noise from its cooling system (I assume).
I wish I'd known what was coming, and gone to the meetings to oppose it.
Large buildings 1000 feet from you are going to have some impact, but your complaint has little to with being a data center specifically. They could have put in a large warehouse and your view gotten blocked just the same, similarly the noise from the cooling system can be managed well or poorly on any building.
usually large warehouses will appear where there’s good highway connections and lots of cheap unskilled labour. A DC might catch a lot of people “in the sticks” by surprise.
I appreciate that my view isn't the only consideration for that kind of decision, but when a new building goes up much larger than anything else in the area, and affects the skyline for thousands of people, I think that should be one of the considerations.
What I'm trying to say is that everything we build has positive and negative effects in our society. And if we want to create a better society we need to have a good understanding of these effects.
I think your article about 'Lack of water, unclean water due to data centers' is a good example of bad faith arguments. It start the article talking about someone that lost access to their private well after a datacenter was constructed. This article don't do it, but I've se people go from arguments like this (a specific water-related disruption) to 'thousands of residencies will loose access to water'.
What strikes me as odd is the fact that datacenters aren't all that special when compared to other infrastructure projects(roads, warehouses, hospitals, power plants, garbage disposal, water dams, ...) but the way we are discussing it seems unique. For every other infrastructure project the discussion seems to be 'how do we make sure that X, Y and Z won't be a problem for the society?'. But when it comes to datacenters it becomes 'datacenters are bad and we should not build them', which seems bad way to approach this issue.
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)
> 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.
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.
> Or do data centers use evaporative cooling just like power plants?
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).
The problem with data center water usage is that it is unnecessary from the PR point of view. Data centers can run on air cooling just as good, but more expensive. For all I know, we could also do just as good without data centers, like we did 20 years 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.
Golf course water usage vastly dwarfs data center water usage. Google used something like 1 billion gallons a year for their DCs. Single golf courses in arid regions can use upwards of a hundred million gallons a year, and in those areas there can be dozens of courses. It's not even close in terms of water usage.
Golf courses in the USA used about 2.1 billion gallons of water per day circa 2004 [1]. In other words, the annual usage of Amazon's datacenters per the article, 7.7 billion gallons, is less than the amount of water used on just American golf courses in four days.
> Data centers can run on air cooling just as good, but more expensive
"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.
The difference in energy usage won't be noticeably higher for PR purposes. Of course, the difference comes at a price, cutting which is the main incentive for water usage.
> That said stories about data center water use are a distraction from much bigger water consumers
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.
Also - Will data center water usage remain "negligible" if AI succeeds at wide adoption and scales to 100x current deployment? If 100x current usage levels become a concern, I don't know why people pretend that current usage is not a concern for a tech that many of those same people are projecting to scale.
While technically true, if your datacenter is in Phoenix and you just consumed a few acre feet of water to raise the relative humidity by 0.000001%, for all intents and purposes that was a massive waste of water.
If the data-center depletes the water table used for surrounding wells, whether the "released water" is still "available" is irrelevant to those dependent on wells for water.
> Where does the water go? If they simply take in cold water and release hot water, that water is still available for other uses.
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.
> In addition, water is almost never wasted, only moved around.
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.
In an increasing number of regions, yes, but not where AWS data centers are located. People are not dying of thirst because of us-east-1. The one has nothing to do with the other.
You have to know this isn't a valid comparison. When people say the water isn't lost to cooling, they literally mean the water ends up back in the water table --- on a human time scale. When you burn fuel to generate electricity, you don't get the fuel back.
> In addition, water is almost never wasted, only moved around.
Technically yes, vapor goes to the atmosphere etc. But in certain areas, data centers are effectively removing water that was previously used for farming.
It should be inconsequential. Sometimes it isn't. If you're pumping water from an aquifer in the desert for evaporative cooling, that's highly consequential.
Unfortunately, media sound bites can't distinguish meaningless water usage from meaningful usage.
You might have a point if it was wheat for human consumption vs datcenter, but those aren't the water hogging plants, which are stuff like almonds, alfalfa (for export)[1]. Comparing those instead, it's unclear whether those are "more important and more beneficial to humanity" than AI, which also genuinely provides utility to people (as evidenced by its popularity).
Honestly, it's hard to tell. Humanity benefits a lot from the massively complex set of technologies that require the existence of big data centers. Including agricultural production itself.
Both in terms the amount of food we grow and the types of food we choose to produce we are way past the necessity of feeding ourselves and firmly in the territory of producing luxury goods that harm both ourselves and our environment. It's not as different as it seems at first glance
The water usage for these new numbers of power consumption, in the GW range, is thousands of tonnes per day (if my maths is right haha). It's a HUGE amount of water.
I agree that the anti-datacenter hype is not much different than anti-nuclear or anti-vaccine insanity, and uses the same tactics of deception and obfuscation.
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.
> 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.
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.
Yes, the materials I have seen have not convinced me, either.
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.
Useage is one issue and should be monitored, but I think you have to understand that in some cases the tech company purchases the water supply and towns become dependant or placed at will of the tech company's interest. On top of that, there is a cost increase to utilities even if the water is moving around a closed loop.
Why do companies actively lie in their advertising about being eco-friendly, instead of just keeping a low profile?
Is it because we tend to focus only on current events and quickly forget their past track record? Indeed, if people soon forget the lies, the risk is minimal.
One reason is that large institutional investors or lenders enforce certain agendas by only giving money to companies that meet certain criteria. Thus companies will posture themselves as meeting those qualities to attract money and investment.
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.
> 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.
> 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.
Concerns about inconsequential water usage at datacenters is a far more welcome problem for the the industry than the other real issues they could be dealing with. People distracted by water concerns will not notice the very real energy usage and AI ethics/practices issues.
Say what you want but the industry has figured out how to manage public perception and sentiment. Water usage problem is easy to fix, while energy usage is a far tougher nut to crack.
I am assuming the 7.7 billion gallons(29b liters) a year is all surface water. It better be. It would be hideously irresponsible to use any ground water for cooling their data centers.
> 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.
> 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.
What happens to water that a data center “uses”? Is it warmed up and returned to the environment? Contaminated and sent to wastewater? Evaporated into the air?
Its freshwater and has to be freshwater because it goes through pipes and/or is evaporated. Corrosion, scaling and fouling are all issues.
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.
""It would be better if they could own up to it," said a current Amazon software developer, who asked to remain anonymous for fear of retaliation. "Even if they said it was a low priority, at least that would be honest.""
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
no opinion on the rest of the point, but, why do you keep writing "so-called" in front of tech companies? They are called that because they are tech companies; the word's meaning is widely agreed upon.
This is a reaction to the unwarranted fear-mongering in the media over water usage by data centers. The amount of water is utterly inconsequential. Growing corn on the same land the data centers currently occupy would consume more water and provide far less value.
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.
Agree in spirit, but 'Growing corn on the same land the data centers currently occupy would consume more water and provide far less value' makes the argument weak because those 'agriculture is barely 2% of GDP' argument are like 'brain is barely 2% of your mass'.
This isn't really about using water as much as dealing with all of the heat that comes out of computation. The water is just the simplest way to dispose of the heat.
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'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.
> Amazon’s data centres were projected to use 7.7 billion gallons of water a year by 2030, according to the leaked strategy memo, which was circulated within the company in 2022.
> noted that it would be harder to reach its internal target if its calculations included “secondary” use—water used in generating the electricity to power its data centres, according to the document.
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.
To put this in context, it takes about 1 gallon of water per 1g of almond [1]. And in California's dry climate, this water comes from groundwater that doesn't renew as fast as it's depleted. So the next time someone chastises you for your non-low flow showerhead that uses more than 2 almonds of water per minute, remember to put the numbers in context.
1. Numbers from a pro-almond group: https://www.almonds.org/why-almonds/growing-good/water-wise
I don't know why, but the fact that a "pro-almond group" exists chuckled me up.
There is corruption for everything money touches...
I actually wonder if there is not single moderately sized industry that does not have some interest group...
Industry advocacy groups ≠ corruption.
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.
Lobbying as practiced by these advocacy groups is basically American flavored corruption. If the system were in place anywhere else we’d call it corruption, but it’s our system so we call it Lobbying.
Yes but that ship has sailed, so now you need anti-corruption corruption to hopefully corrupt the corruption so it's less corrupt.
Like, the only reason "almond advocacy" even exists is because dairy and beef are some of the most lobbied and blessed industries in America. They can practically do whatever they want, whenever they want, so fuck you, pal.
Nah dude look at France and their farm owners
this assumes that the needs of both the workers and the owners (who pay for the lobbying) are aligned, and they're often not
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
This assumes no such thing.
You're right, it asserts it. Lobbying is not necessary, and in some other wealthy countries it would be considered bribery.
It only becomes advocacy when you take the money out of it.
This is a Netflix-political-drama enthusiast's understanding of lobbying. Turn off House of Cards or whatever and try to build something.
I've built enough things to know that you don't build understanding by insulting people. You seem to disagree with my position, but are unable to find the words to defend yours.
This sounds a bit sensationalized, and I'm not sure if it's the source from which I originally learned of the issue, but:
>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...
When the water wars start, they will be the first against the wall.
If there is and you name it, the interest group will be formed.
Almond trees are stuck as almond trees forever. They can't switch to something else. Anywhere an investment in something is entrenched like this, you'll find lobbying.
Is that the Almond Front of California, or the Californian Almond Front?
I'd non-ironically join that. Almonds are the bees knees. Almond milk is an atrocity though.
Almonds alone are like half of all urban water use in california.
More specifically, farming and urban use 50% of the water California gets (river outflows to the sea are the other main place water ends up) Urban use is 10% and farming is 40% of that total water input. Almonds alone are 7% of ALL water, almost as much as the 10% urban.
The level of subsidy the relatively unimportant crops get with their basically free water is astounding, especially considering the high urban prices of water
Tangential to the point, I think we should be careful about the almond talking point. Insofar as it is used for milk, almond milk uses almost half as much water as dairy milk, uses 1/18 the land and emits 1/5 the amount of carbon. As food, it is eaten in such a vanishingly small quantity compared to other water-hungry foods (meat) as to be insignificant. https://ourworldindata.org/grapher/environmental-footprint-m...
I thought the problem is that that growing almonds is usually done in arid regions, so the issue is not that it uses lots of water, but that it uses lots of water in areas that frequently experience droughts. This is an honest question (I truly don't know, although I suspect I can guess): are dairy farms also common in those areas?
Just for a single point of comparison, California's alfalfa consumes 15% of the state's water. All that alfalfa is going to feed cattle. California only produces 9% of the USA's alfalfa so it's easy to say that this is a tremendous amount of water.
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.
World beef production is approx. 60x almond production by mass, and that doesn't include dairy. That isn't the whole story, because cattle use more water than just what is used to grow alfalfa, but you are comparing apples to oranges here.
You can check the per calorie stats.
Animal agriculture is wildly inefficient and honestly it's not surprising because you have to keep living moving animals around for it.
I was drawing a point of comparison in response to the question about almonds disproportionately using scarce water resources in arid areas, which is a different question from overall water usage. My point was that almond water usage is literally only an issue in California, and that their water usage is not that extreme given the size of the market California serves.
tbf some of the alfalfa is shipped to China it's not all used on cows here
The comparison you are bringing up actually just gets to the heart of the issue.
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.
Corporations never want to pay for the problems that they cause. They would rather go bankrupt and start a new company.
It’s been some years but I recall that at one point probably around 2016-2017, California produced 80% of the world’s almonds. This was notable because at the same time, California was experiencing historic droughts.
Yes- about half of the total water siphoned from the Colorado river goes to cows. This number astounded me when I first read it, and I hope it has a similar effect on you. I don't like almond milk, for what it's worth, and I don't think we should ignore plant-based foods with a high climate impact, but animal agriculture is the most environmentally devastating institution we have individual power to transition from. https://www.npr.org/2024/09/09/nx-s1-5002090/colorado-river-...
GP's point works just as well if you substitute almonds with diary milk or hamburger meat though. The specific water use of almonds is indeed completely tangential
Account created 10 minutes ago, hitting talking points of the almond growers industry association.
They are replying to talking points of the dairy industry.
Even if true, you’ve not countered a single point. Are there un-truths among those points? If so, let’s hear them.
I've read here for a long time but just made my account because I have been feeling very compelled by the data surrounding the huge economic effects of the animal agriculture industry and how otherwise pro-science and pro-data people find themselves with deeply entrenched unscientific viewpoints. Should I link my Google scholar to prevent people from seeing conspiracies everywhere??
I think the almond talking point takes hold because, like a lot of complaining about LLM usage, large parts of the blame gets directed elsewhere rather than the choices that we're all way more likely to be making and could influence. Like, it's 2025. Even the people most likely to be drinking almond milk have largely moved to oat, whose water usage is great.
Almond's get brought up because it distracts from the fact that beef and dairy milk production uses a lot more water than almonds do.
No one wants to be reminded that their 4 oz burger they had for lunch used 460 gallons of water to produce.
Except that almond trees thrive in hot dry climates. Cows thrive in the rain.
But it's okay. This has been solved very recently as in last week. We are going to now be getting our beef from Argentina. Not only has the prices of beef issue been fixed, it'll also fix the country's water shortage issue as a bonus. /s
“Almond milk” is not milk. You know what else is less carbon-intensive than milk? Candy corn. But that is also _not milk_, and so equally irrelevant!
Almond milk is an economic substitute for dairy milk, making the comparison appropriate. No need to be dense about it.
Almond milk is not dairy milk, but it is absolutely "milk", in the sense of a white liquid derived from plants - a definition that has existed in English for hundreds of years.
The name "almond milk" has been used since at least the 1500s.
I don’t drink ‘almond beverage’ but given the amount of uses it has substituting milk (and the amount of people that accept them) it seems like a very relevant comparison. Maybe I’m not sofisticated enough but I’m yet to see a candy corn mlik latte be ordered.
> I’m yet to see a candy corn mlik latte be ordered.
Not exactly the same, but can I interest you in a caramel-waffle-oat-milk latte?
https://mightydrinks.com/cdn/shop/files/Barista-CW.png
You certainly can sir!
But that's such a good idea though
> “Almond milk” is not milk. You know what else is less carbon-intensive than milk? Candy corn. But that is also _not milk_, and so equally irrelevant!
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:
- https://www.ift.org/news-and-publications/food-technology-ma...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.
I mean I'm drinking a coffee with almond milk right this second.. which coincidentally replaced the dairy milk
Candy corn in my coffee wouldn't taste anywhere near as good
Yes: my understanding is that it’s rather common practice to at least make a best-effort estimation of all these secondary impacts.
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.
This also tends to include the rainwater that falls on the ground to grow the grass that cattle eat that would fall on the grass anyway, or that the cow farts aren't really more greenhouse emissions than the grass rotting out over the summer anyway. It's distortive.
IMO, water is a renewable resource and what should count is the use of water in scarce environments from scarce sources directly in excess of what gets renewed. If you're right by the Mississippi river and often see flooding in your region, I don't think using the water for cooling a reactor (steam as the byproduct) is necessarily something that should be considered a negative... it'll come back down as rain somewhere.
I'm not sure why Amazon is "using" so much water, assuming their cooling systems are a closed loop... otherwise, if using evaporation for cooling, like a reactor, it depends on the location, source and usage. "it's complicated"
> assuming their cooling systems are a closed loop
My understanding is that the closed-loop systems are rather new; like, "we started using these things in 2024 literally because everyone's been moaning about our AI water usage" new. I'd assume that many new constructions are starting to leverage them, but its not 100%, and existing DCs would be slow to upgrade.
Categorical rejection of alternatives is premature without context.
> are we going to include the water used to grow the almonds you ate?
Beef too. It uses the same amount of water but people eat 30x as much annually.
And cheaper cow production, focused on grain-feeding, uses even more water than grass feeding; after spending a month in Switzerland, it's wild to see how addicted we Americans are to cheap beef.
We need our protein or we will die immediately.
Ha ha, but seriously unlike other nutrients the human body can't really store protein reserves for any length of time. So you do need to keep taking in quite a bit every day or the body will break down muscle in order to sustain other tissues.
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.
You need protein for sure. But unless you want to win Mr Universe you will probably get enough through regular nutrition. And there are a lot of other things besides protein you need in your food. The current “Protein in everything” trend is just a fad created by industry
How much is "enough" exactly? Can you quantify that for us? There's a huge difference between enough to avoid acute malnutrition in a young person versus enough to optimize healthspan in an older person. Did you even listen to the podcast I linked above?
It is totally possible to get enough protein (by whatever metric you choose) through regular food. And yet the reality is that many people don't.
> How much is "enough" exactly?
The upper limit seems to be 1g per lb (454g) of body weight. This is for people who lift heavy regularly to build and grow muscle. The lower bound is about 1/3 of that.
A 165 lb person needs about 55g of protein daily for maintenance. This is about one cup of black beans and 3 eggs. Huevos rancheros for breakfast will cover most or all of an average person's daily protein needs and that's before even counting the protein in the cheese or tomatoes.
While vegetables don't have much protein it still adds up. A single medium tomato has 1g, a potato has 3g, 1g in an onion, 0.5g in a carrot, and so on. Eat your daily portion of veggies and you can get maybe 10g of protein. That's nearly 20% of what you need!
Carb-heavy foods such as breads and rice also have some protein. More than veggies, less than lentils, eggs, dairy, and meat. A standard serving of pasta has 7g of protein. A cup of cooked brown rice has 5g.
All of this is to say even vegans who eat a healthy, varied diet are not deficient in protein. If they are ovo-lacto vegetarians it's a layup. They may need to supplement with whey and creatine or eat some chicken a few times a week if they want to squat 3 plates.
Insufficient protein is mostly not a problem in the first-world. The problem is eating crap. Eating grass-fed steaks every meal doesn't compensate for a crap diet.
You're not accounting for protein quality and essential amino acids. You can't just add up grams. That's not how this works.
https://en.wikipedia.org/wiki/Protein_quality#Protein_source...
Lots of sources of complete protein on there besides beef. Don't limit yourself. A combination of quinoa, buckwheat, eggs, soy, pea, milk, whey, beans, and nuts will provide all the amino acids you need. Even beef, despite being a complete protein, is deficient in tryptophan.
I like meat. I'm not a militant vegetarian or vegan or anything like that. I dislike the laundering of people's love of eating meat as "essential for health", when it's anything but (special dietary needs aside). Eating meat is a mostly a choice.
I'll repeat again: varied diets are essential. A single protein source won't cut it. And most people get enough protein going just by raw numbers.
I don't know why you felt the need to post that little rant. No one here claimed that people should only eat beef as their sole source of protein. You're arguing against a strawman.
https://news.ycombinator.com/item?id=45722288
You've been pretty reasonable. Not everyone is.
I'm very pro-protein - we just tend to see better metabolic health outcomes when we eat more of it. Diets with higher proportions of protein result in better body composition, less fat, less chance of diabetes, etc. I think the vast majority of people in America at least would be better off with higher protein diets.
But "Did you even listen to the podcast I linked above" is an insane thing to say in a random online conversation. No, you cannot expect people to listen to a whole podcast to make whatever random unsubstantiated point you have. Explain the important things you took away from the podcast and link it as a reference for if people disbelieve you for whatever reason.
Do we need beef protein? What's wrong with chicken or eggs or beans?
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.
Yeah people who eat beef are actually healthier than vegetarians.
The difference in life expectancy between the USA (where beef is heavily consumed) and India (where you may get lynched on suspicion of eating it) is 78.39 - 72 = 6.39 years at birth.
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.
At the risk of stating the obvious, meat is not the only difference in 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.
This stands against the evidence. Beef is causally linked with the largest killers of Americans, including heart diabetes, diabetes, and obesity in general. "a mostly plant-based diet could prevent approximately 11 million deaths per year globally, and could sustainably produce enough food for the planet’s growing population without further damage to the environment." A "Mediterranean diet" is more healthy than the average American diet because it is more plant-based. https://www.sciencedirect.com/science/article/abs/pii/S01406... https://www.health.harvard.edu/blog/with-a-little-planning-v...
Beef is not causally linked to diabetes.
There are proposed diabetic mechanisms which are exacerbated by cholesterol. Plant based food are completely devoid of cholesterol and have dramatically lower incidences of diabetes. So there you have a mechanism, and the demonstrable effect of abstaining from that source of food. What else do you want? https://diabetesjournals.org/diabetes/article/56/9/2328/1265... https://www.mdpi.com/2218-273X/13/2/224
A causal link.
Define "healthier".
Source?
I wonder if cow milk products would be more expensive if people ate less beef.
Presumably, after a cow is done being used for milk, it can then be sold for meat.
Dairy cows don’t taste good. They’re used for the bottom of the barrel low grade ground meat. Think dog food.
Old cows are generally very flavorful. They're more beefy and gamy than the young cattle we're used to eating. They're also stringy and tough.
So you wouldn't want an old cow steak, but a stew or burger made from an old cow can be awesome.
I guess it depends. I have a friend who has dairy cows, grass-fed, very well cared for. He names each one and hates to slaughter them, but when they are done producing, he does take them to the slaughter house. I bought a quarter of a cow from him a few years back, and it was excellent. It did have a hint of gaminess redolent of venison or lamb, but it was delicious.
You are misinformed. I regularly eat our retired dairy cows. They taste absolutely excellent, but they were raised on pasture and ate top quality feed.
Dairy cows from commercial confinement dairies are a different matter. They eat almost exclusively grain and develop horrible health problems.
Yet the same can be said of any meat produced in a CAFO, when compared to that produced on most small farms. Garbage in, garbage out.
My guess is that they have specialized breeds and how they're raised and fed and such, so that dairy cattle wouldn't make for good beef and vice versa. They probably do use them somehow, but maybe for dog food or some similar use.
"It's wild to see how Americans are able to procure an extremely high quality protein source inexpensively"
It's pretty easy if you don't care about animal welfare and environmental issues. As soon as you care, prices go up steeply.
American cows lead very good lives, up to including death. Better than pretty much any mammal on earth except for humans and 1st world pets.
without paying now for the externalities of its production
how dare you, he spent a whole month in switzerland. what an exalted being
Only 30x? Y'all eat a lot of almonds.
Beef is more nutrient efficient though and has better macros for human consumption. It has 4 times more protein, less fat, and no carbs. Seems to me at least Americans could do well to eat more beef and less products created from processed almonds mixed with refined vegetable oils and sugar.
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 has 4 times more protein, less fat, and no carbs
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.
Factory-farmed beef is the worst source of pollution in the food industry. We definitely need less of that.
"Factory-farmed beef" doesn't even exist. All cows get raised in fields for their first 12-18 months. The ""factory-farming"" is just a feed lot they get taken to for 3-6 months to eat grain before slaughter.
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.
You've never heard of free-range chickens?
I agree pigs are raised in abominable conditions.
Not all pigs and chickens are "factory farmed" but most are. And unlike cows, the term is relatively meaningful and descriptive when applied to the way those are raised. Bro heard about how pigs and chickens get raised in tiny cages never seeing the light of day then assumed that's how it works for cows too, which he now realizes isn't the case because I shamed him into looking it up.
You're going to have to cite strong sources or else this is either heavy cope or straight-up denial.
It is inconceivable that American's consume as much beef as they do, yet production has been able to scale without resorting to factory-farming. Every other commodity food is factory farmed. It's asinine to think beef is immune to that.
Bro, get out of the city or just look at a map, America is absolutely enormous. There is no shortage of land to let cows graze on. Feedlots for cattle aren't even necessary at all, they're only used to increase profit.
What's really asinine is that you have such strong opinions about a subject you know nothing about and demand that other people do research for you.
Cope harder, bro. Show me sources. Cattle is a massive industry and it should not be hard to find sources proving me wrong. Calves live on pasture for the first few months, but are then transferred to backgrounding farms, that are technically still pastures, but are fed a highly supplemented diet and have less room to roam. STFU about me not knowing what I'm talking about.
Feedlots are absolutely necessary at the levels that American's consume beef. It might look like there isn't a shortage of pasture for cows, but the truth is there is not enough land to transition to 100% grass-fed with American's level of consumption. [1]
[1] https://grazingfacts.com/land-use
> Calves live on pasture for the first few months, but are then transferred to backgrounding farms, that are technically still pastures, but are fed a highly supplemented diet and have less room to roam
Glad to see you finally looked it up and saw that I was right. Your attempt to save face is amusing though. From "factory farmed" to "the pastures are a bit small" is funny.
And no, feed lots aren't necessary. They make beef cheaper by making cows bigger, therefore making the industry more profitable, but if you did away with them Americans would still be eating enough beef to make urban vegan weenies seethe.
Pork and chicken have better feed conversion ratios and water consumption
Cattle water consumption should be meaningless. If a cow is drinking water from a surface water source and breathing/sweating/peeing it out in a pasture, that's the same process that would be happening if humans didn't exist.
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...
The problem is the assumption that cows (and other similarly intensive animals) would exist in the same quantity. You'd expect something like 5x less cow-like biomass without people intervening.
If humans didn't farm them there'd be far fewer cows alive and drinking water.
And far more bison. Not the same, but similar.
ok!
I highly recommend you read up on the actual research of what you're talking about. It points to the exact opposite of basically every sentence you wrote
Not all processed foods are created equal. Almost all of the elevated health risk from processed foods comes from processed meats and sugary drinks. https://www.nytimes.com/2024/05/06/well/eat/ultraprocessed-f... Whole grain breads are ultra-processed, and I don't think many are arguing against those. Beef has absolutely devastating effects on human health including elevated cancer risk, diabetes risk, dramatically higher incidences of heart disease (the greatest killer of Americans). Plant-based substitutes are scientifically shown to lead to better outcomes. Better yet, soy based whole foods are excellent for human health, contrary to the bro-science talking points. Turns out, beans are good for you! https://hsph.harvard.edu/news/the-bottom-line-on-ultra-proce... https://www.uclahealth.org/news/article/health-benefits-soy
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...
At least that is generally the case for calculating the ecological footprint of a person.
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.
I generally find criticism of water waste misplaced. Water’s lifecycle is cyclic. The water isn’t lost unless it is being salted or otherwise poisoned.
The real waste is the energy required to produce, clean, and transport the water that is being “wasted.”
aquifers deplete, the water table lowers, wells dry up
sustainability, availability and maximum marginal price matters, just as with electric power generation
Depletion of local water resources is a good thing to measure. But by and large, this is not what we're measuring. Instead, we're coming up with absurd statistics that imply any water put to beneficial use just disappears forever.
If your tap water comes from a river and flows back to a river, leaving it running mostly just wastes energy.
That's a fair point. If you want to calculate the real total water usage of any person, you must first invent the universe. You have to cut it off somewhere.
I find this loses the point, because agriculture is essential. If we preserve water it's because we want to keep enough of it for essentials like agriculture and such. Not that agricultural practices shouldn't better conserve water, but it's the usage of water for non-essential things that I think most people find wasteful.
It's not essential that almonds be grown in California.
It sounds like we know the water usage for growing almonds.
It feels reasonable that we should have the same detail of information for data centers.
Like the emissions we can have various scopes and different targets per scope.
Does Amazon control the generation of this electricity? If so, then I definitely think it should be included, because it is something that Amazon has control over, and could possibly reduce water use.
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.
> are we going to include the water used to grow the almonds you ate?
That’s a good idea. Like nutrition facts but for everyday economic climate decisions.
Knee jerk reaction for me every time I see "almonds"
One gallon of cow milk uses more water than one gallon of almond milk.
Almond milk is 2% almond. Cow milk is 100% cow milk.
An gallon of raw almonds requires a massive amount of water.
Using numbers to create illusions of comparison is the marketing equivalent of parsley in your tooth.
OK but rain water != tap water.
It's not rainwater. (it's not tapwater either)
We can. We do. It depends on what you are trying to count. If you think there is a single number that will answer everything, no. Are you going to include all the water ancestors used? After all, without that water, I wouldn't be here.
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.
There is an honest way to count and that is to count the water directly used by the person or process in question. There are many other ways to count to make big scary number for clickbait article. I suppose we could also count all the water used by data center employees to take showers as "data center water usage" but that is dishonest. Those are people who exist and they are going to take showers whether they work at the data center or not.
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.
Kind of surprised I have to point this out. Power companies do not generate the same amount of power regardless of whether that power would be consumed on their grid or not. They increase generation based on demand. Whether that power comes from high water consumption generation is based on location which determines which power sources are available in the local grid. A major part of why indirect water consumption from power generation is included in the standard WUEsource - water usage effectiveness metric - is because it makes it clear what the impact will be when assessing data center location and size. In other words, it's important to choose locations near power generation that doesn't consume water heavily. Yes, the amount of water used in generating electricity and the efficiency there is controlled power companies choices for power generation at their locations. Data centers control the location they are built in and without an estimate of indirect water usage they cannot strategize locations with lower environmental impact.
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.
A power company builds a generation plant of a certain size. That plant is going to run whether there is a data center or not. Maybe there is some incremental additional water usage if it is running at 85% capacity instead of 75% capacity but this is probably marginal compared to the plant running or not running at all.
Depending on the method of power generation, it might also need a certain consistent base load to run most efficiently, so adding in a data center that will be a consistent load running 24/7 could actually increase the efficiency.
> Because agriculture is going to dwarf anything that data centers use
Will it though
What does that mean though: "water used"? The only actual use of "water" is an E = mC^2 conversion to raw energy, or a chemical/location change such that it's locked-up/removed from availability permanently.
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.
I'd rather have the almonds than another data center.
Oh right are almonds actively extracting money from you because their business model is exploitative?
I find it really odd this recent push for discussions around the development of new datacenters.
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.
Most new datacenters are powered with gas turbines (because utility grid connections are slow to deploy) and that'll surely affect everyone nearby.
One went in 1/4 mile from my home a couple years ago. I ignored the notices of development because I thought it was far enough that it wouldn't affect me, but it blocks the view of the mountains that I used to enjoy, and sometimes I can hear noise from its cooling system (I assume).
I wish I'd known what was coming, and gone to the meetings to oppose it.
Large buildings 1000 feet from you are going to have some impact, but your complaint has little to with being a data center specifically. They could have put in a large warehouse and your view gotten blocked just the same, similarly the noise from the cooling system can be managed well or poorly on any building.
usually large warehouses will appear where there’s good highway connections and lots of cheap unskilled labour. A DC might catch a lot of people “in the sticks” by surprise.
Hopefully we will emerge from this with a legislative framework that says "fuck your view"
I appreciate that my view isn't the only consideration for that kind of decision, but when a new building goes up much larger than anything else in the area, and affects the skyline for thousands of people, I think that should be one of the considerations.
As long as it includes compensation for lowering house prices, of course!
Let’s use a little logic. What are the “both sides” here? Megacorps with billions vs “local media”? Citizens?
Yes, that’s a balanced equation.
Now, do we have any evidence to back complaints?
Lack of water, unclean water due to data centers:
https://www.bbc.com/news/articles/cy8gy7lv448o
Data centers causing energy prices to increase:
https://www.bloomberg.com/graphics/2025-ai-data-centers-elec...
Data center natural gas generators flooding communities with pollution:
https://www.politico.com/news/2025/05/06/elon-musk-xai-memph...
Yes, “both sides”.
What I'm trying to say is that everything we build has positive and negative effects in our society. And if we want to create a better society we need to have a good understanding of these effects.
I think your article about 'Lack of water, unclean water due to data centers' is a good example of bad faith arguments. It start the article talking about someone that lost access to their private well after a datacenter was constructed. This article don't do it, but I've se people go from arguments like this (a specific water-related disruption) to 'thousands of residencies will loose access to water'.
What strikes me as odd is the fact that datacenters aren't all that special when compared to other infrastructure projects(roads, warehouses, hospitals, power plants, garbage disposal, water dams, ...) but the way we are discussing it seems unique. For every other infrastructure project the discussion seems to be 'how do we make sure that X, Y and Z won't be a problem for the society?'. But when it comes to datacenters it becomes 'datacenters are bad and we should not build them', which seems bad way to approach this issue.
Water use is a pretty close proxy for power use which is a pretty close proxy for how much computation is happening.
I can totally see why a company wants to keep this info secret.
Competitors would really like to know.
Yes if we squint really hard then this could be the reason. It could also be because it's a PR disaster.
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)
Only because people are innumerate, though.
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)
You think that's more of a concern than public backlash with NIMBYs and local governments?
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.
You just proved their point though
Please send your CV over to PR
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.
Is this water even “used” in the same sense that water is used for bathing or agriculture?
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?
> Or do data centers use evaporative cooling just like power plants?
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).
The problem with data center water usage is that it is unnecessary from the PR point of view. Data centers can run on air cooling just as good, but more expensive. For all I know, we could also do just as good without data centers, like we did 20 years 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.
Golf course water usage vastly dwarfs data center water usage. Google used something like 1 billion gallons a year for their DCs. Single golf courses in arid regions can use upwards of a hundred million gallons a year, and in those areas there can be dozens of courses. It's not even close in terms of water usage.
Golf courses in the USA used about 2.1 billion gallons of water per day circa 2004 [1]. In other words, the annual usage of Amazon's datacenters per the article, 7.7 billion gallons, is less than the amount of water used on just American golf courses in four days.
[1] https://www.usga.org/content/dam/usga/pdf/Water%20Resource%2...
> Data centers can run on air cooling just as good, but more expensive
"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.
The difference in energy usage won't be noticeably higher for PR purposes. Of course, the difference comes at a price, cutting which is the main incentive for water usage.
OK, so we need to include the effect of introducing a green house gas, water vapor, into the environment.
> That said stories about data center water use are a distraction from much bigger water consumers
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.
I don't think it's a fallacy, it's much easier to optimize water usage for something that much larger.
Agriculture uses about 70% of all freshwater while datacenters are less than 0.5%
Some leaky channel will cost more than all the datacenters.
Also - Will data center water usage remain "negligible" if AI succeeds at wide adoption and scales to 100x current deployment? If 100x current usage levels become a concern, I don't know why people pretend that current usage is not a concern for a tech that many of those same people are projecting to scale.
>Also - Will data center water usage remain "negligible" if AI succeeds at wide adoption and scales to 100x current deployment?
Yes.
Even in case of evaporative cooling the water is not used. It's returned to the environment.
While technically true, if your datacenter is in Phoenix and you just consumed a few acre feet of water to raise the relative humidity by 0.000001%, for all intents and purposes that was a massive waste of water.
If the data-center depletes the water table used for surrounding wells, whether the "released water" is still "available" is irrelevant to those dependent on wells for water.
e.g. https://www.nytimes.com/2025/07/14/technology/meta-data-cent... https://www.bbc.com/news/articles/cy8gy7lv448o
> Where does the water go? If they simply take in cold water and release hot water, that water is still available for other uses.
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.
Datacenters can use a few different kinds of liquid cooling, including the one you describe. It varies quite a bit by geography.
Even if it's evaporative cooling, couldn't the water vapor just be... condensed back into water?
Yes - by cooling it. See the problem?
You don't have to use energy to cool it though do you?
Couldn't it just flow into a big passive outdoor radiator?
This is incredibly meaningless reporting.
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.
> In addition, water is almost never wasted, only moved around.
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.
In an increasing number of regions, yes, but not where AWS data centers are located. People are not dying of thirst because of us-east-1. The one has nothing to do with the other.
You have to know this isn't a valid comparison. When people say the water isn't lost to cooling, they literally mean the water ends up back in the water table --- on a human time scale. When you burn fuel to generate electricity, you don't get the fuel back.
> In addition, water is almost never wasted, only moved around.
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...
It should be inconsequential. Sometimes it isn't. If you're pumping water from an aquifer in the desert for evaporative cooling, that's highly consequential.
Unfortunately, media sound bites can't distinguish meaningless water usage from meaningful usage.
Isn't agriculture objectively more important and more beneficial to humanity than Big data centers?
You might have a point if it was wheat for human consumption vs datcenter, but those aren't the water hogging plants, which are stuff like almonds, alfalfa (for export)[1]. Comparing those instead, it's unclear whether those are "more important and more beneficial to humanity" than AI, which also genuinely provides utility to people (as evidenced by its popularity).
[1] https://www.npr.org/2023/08/09/1192996975/amid-a-water-crisi...
yes and no. Most agriculture is not necessary for pure survival, especially water-needing crops in the desert. It's more luxury food products.
In addition, increasing human productivity through technological innovation is the only thing that ever let us escape the malthusian trap.
> increasing human productivity through technological innovation is the only thing that ever let us escape the malthusian trap.
How so?
https://www.epi.org/productivity-pay-gap/
Many such charts as this: https://www.researchgate.net/figure/World-economic-history-a...
In absolute terms, yep. In marginal terms, not so much. See also: paradox of value
Honestly, it's hard to tell. Humanity benefits a lot from the massively complex set of technologies that require the existence of big data centers. Including agricultural production itself.
Agriculture produces food... it feels silly to compare the two.
Both in terms the amount of food we grow and the types of food we choose to produce we are way past the necessity of feeding ourselves and firmly in the territory of producing luxury goods that harm both ourselves and our environment. It's not as different as it seems at first glance
The water usage for these new numbers of power consumption, in the GW range, is thousands of tonnes per day (if my maths is right haha). It's a HUGE amount of water.
I agree that the anti-datacenter hype is not much different than anti-nuclear or anti-vaccine insanity, and uses the same tactics of deception and obfuscation.
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...
> 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.
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.
Yes, the materials I have seen have not convinced me, either.
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.
Useage is one issue and should be monitored, but I think you have to understand that in some cases the tech company purchases the water supply and towns become dependant or placed at will of the tech company's interest. On top of that, there is a cost increase to utilities even if the water is moving around a closed loop.
Why do companies actively lie in their advertising about being eco-friendly, instead of just keeping a low profile? Is it because we tend to focus only on current events and quickly forget their past track record? Indeed, if people soon forget the lies, the risk is minimal.
One reason is that large institutional investors or lenders enforce certain agendas by only giving money to companies that meet certain criteria. Thus companies will posture themselves as meeting those qualities to attract money and investment.
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.
Oh, I didn't know it worked that way. Thank you for the information.
Is data center water use not closed loop? If not, why?
NREL (National Renewable Energy Laboratory - https://en.wikipedia.org/wiki/National_Renewable_Energy_Labo... ) had a neat dashboard of their cooling system for the HPC data center. https://www.nrel.gov/computational-science/hpc-data-center
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/
They take in ambient air and release 100% humidity warmer air, basically a giant server-heated humidifier.
You could spend more electricity if needed to up the airflow to get the same cooling power without humidifying.
Concerns about inconsequential water usage at datacenters is a far more welcome problem for the the industry than the other real issues they could be dealing with. People distracted by water concerns will not notice the very real energy usage and AI ethics/practices issues.
Say what you want but the industry has figured out how to manage public perception and sentiment. Water usage problem is easy to fix, while energy usage is a far tougher nut to crack.
I am assuming the 7.7 billion gallons(29b liters) a year is all surface water. It better be. It would be hideously irresponsible to use any ground water for cooling their data centers.
https://www.opb.org/article/2021/09/29/google-water-data-cen...
> 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.
or 24,000 acre-feet in the terms that we normally use to speak about water. In other words, basically none.
What happens to water that a data center “uses”? Is it warmed up and returned to the environment? Contaminated and sent to wastewater? Evaporated into the air?
> a commitment to “return more water than it uses by 2030”.
How would that even be possible?
Some are adding/funding water treatment plants tre reduce the load from the city water main
Dumb question, but is this done with fresh water?
If so, why?
If not, does it matter how much water is used?
Its freshwater and has to be freshwater because it goes through pipes and/or is evaporated. Corrosion, scaling and fouling are all issues.
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.
Thanks!
Sensing this is the new global warming threat replacement
""It would be better if they could own up to it," said a current Amazon software developer, who asked to remain anonymous for fear of retaliation. "Even if they said it was a low priority, at least that would be honest.""
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
no opinion on the rest of the point, but, why do you keep writing "so-called" in front of tech companies? They are called that because they are tech companies; the word's meaning is widely agreed upon.
d2FzICsz
This is a reaction to the unwarranted fear-mongering in the media over water usage by data centers. The amount of water is utterly inconsequential. Growing corn on the same land the data centers currently occupy would consume more water and provide far less value.
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.
Agree in spirit, but 'Growing corn on the same land the data centers currently occupy would consume more water and provide far less value' makes the argument weak because those 'agriculture is barely 2% of GDP' argument are like 'brain is barely 2% of your mass'.
The company that literally named itself after a river and the threatened habitat it runs through.
This isn't really about using water as much as dealing with all of the heat that comes out of computation. The water is just the simplest way to dispose of the heat.
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.
> Maybe heat a building?
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.
> "You need boiling water to generate electricity"
https://en.wikipedia.org/wiki/Mercury_vapour_turbine
the efficiency of converting it to electricity would be low, due to low absolute temperature difference and thus carnot effeciency.
I think it could be higher with germanium semiconductors, as you could run them at a higher temp and get superheated steam
The level of obsession some people have over data center water usage is completely unhinged.
> Amazon’s data centres were projected to use 7.7 billion gallons of water a year by 2030, according to the leaked strategy memo, which was circulated within the company in 2022.
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.
[dupe] Some more discussion: https://news.ycombinator.com/item?id=45710025