edit: corrected to 736, its per year.
The US uses about 2 billion gallons of water per day on golf courses.
It's possible they've gotten more efficient in the past 14 years, but it's also possible there are more golf courses today. I haven't looked into it.
[0]: https://www.usga.org/content/dam/usga/pdf/Water%20Resource%20Center/how-much-water-does-golf-use.pdfFor context, the city of London uses about 2.6 billion liters, or about 680 million gallons, per day.
So that's about four days of London water usage per year, give or take -- or just over 1% of London's water usage.
That doesn't seem like that much, really. The Seattle metro area isn't huge; that consumption is only 0.7% of New York's, and Amazon runs the largest network of data centers in the world.
California pistachios consume about 500–600 billion gallons of water per year.
You could also argue that Amazon data centers are crucial commercial infrastructure that used for a lot of logistics necessary to move food around.
I had to say I was wondering why basic non polluting water use would need special attention like this. Apparently it’s a PR meme to divert attention from real problems.
I think a ton of people were REALLY misinformed about how much water AI data centers use. I know I was at one point.
Now there may be people pushing that narrative still on purpose because it clearly works. But I don’t think the average person who uses that talking point is doing anything other than expressing a concern based on the (terrible) information they got somewhere.
There's a reason certain types of misinformation become popular and others fizzle. The environmental concerns around AI are starting from the goal 'disliking AI' and going in search of a reason for many people. The environment is a convenient reason because it links to an existing left-wing cause & doesn't require conceding the frame of AI rapidly becoming extremely capable (scary! don't like to think about that!) so it's all comfort and outrage without stakes.
Sure, they shouldn’t be powered by gas or coal. And the local effects are significant.
But the macro environmental effects are minor and the local effects can be resolved with the most trivial regulations.
But they mostly are....
According to new data quietly published this week, energy use by AI datacentres in the UK could cause the emission of up to 123m tonnes of carbon dioxide (CO₂) – about as much as generated by 2.7 million people – over the next 10 years.
That latest figure replaces a previous estimate – since deleted – that claimed emissions would reach a maximum of 0.142m tonnes of CO₂ in a single year.
To my understanding, the only thing that changes when water is used for data centers is its temperature.
That's a lot different than residential use, where it's used in toilets and needs to undergo significant wastewater treatment to be cleaned enough to be re-used.
So how do we compare apples to oranges for these very different use cases?
Update: It appears my assumption about the only thing changing in the data center case is temperature. For much of that water, a phase change occurs (evaporative cooling), so it is no longer accessible to be recycled.
* In air-cooled datacentres an approach is "direct evaporative cooling". They might take 30°C outside air and spray a fine mist into it, cooling it to perhaps 25°C before it enters the servers. After passing through the servers the air might leave at around 38°C. The water is now dispersed as humidity in a large volume of exhaust air. Recovering that water would require condensing it back out of the air, which means removing huge amounts of latent heat, it would be cheaper to just use 'traditional' compressor based cooling in the first place.
Cooling towers (which are used in many 'ai' facilities) have essentially the same problem. Servers reject heat into a water loop, and the cooling tower then cools that water by evaporating a portion of it into the atmosphere. The water that leaves as vapour is the "consumed" portion. While some liquid water remains in the system and a small amount is discharged as concentrated blowdown that can be treated and reused relatively easily, the majority of the consumed water has been converted into atmospheric moisture as above.
Power consumption and effect on electricity infrastructure is so, so, so much more consequential and dangerous. It alone is way more than enough on which to base a very solid anti-DC campaign. The water argument weakens the whole anti-DC position by being so refutable.
EDIT: with probable exceptions in specific local instances where water supply is already very constrained, like Utah.
Personally, I find it ironic to see people going on and on about data centers on platforms like Threads, Reddit, and X. It's like, do you know where your data is going when you press that button?
I actually have a great deal of respect for the average person. Most of the time, the intuitive model of the world is very good at getting workable answers. But it completely falls apart when something is outside the universe of what people deal with on a day-to-day basis. Try asking the people in your family what the profit margin of a grocery store is. People might go to grocery stores all the time and know exactly how to comparison shop to optimize their spend. But most actually have no idea about the numbers involved at each step of the supply chain. Trying to explain inflation to people over the last few years has been literal hell, because virtually nobody understands the differences between price levels and the first and second derivatives thereof.
In some cases, they are.
The Colorado River basin waters seven states and is in extreme drought. There are proposed data centers in the area that would require water from the Colorado or from already distressed aquifers.
They are being built with evaporative cooling which is not closed loop.
With that said, the vast majority of the yelling about water issues is overblown, power issues are far more likely to be a problem.
Also, I was wondering, what does 2.5B gallons of water equate to? Here's the answer for curious minds:
> Using EPA’s cited 82 gallons per person per day figure, 2.5B gallons/year equals the annual household water use of about 83,500 people.
I did some further math... If 1bn users world wide leverage AWS services in their daily routine (netflix, whatever, ...), the formula becomes this one:
2.5B gallons/year ÷ 1B users = 2.5 gallons/user/year
> Compared with the EPA-style U.S. household benchmark we used earlier of about 82 gallons/person/day, that would be: > > 0.00685 ÷ 82 ≈ 0.0084%
So the AWS data centers make up roughly an additional 0.01% of daily water usage. Why is this worth a bloomberg article?
I live somewhere that's had a lot of interest from companies wanting to put up new data centers (northwest Iowa, southwest Minnesota – open farmland basically). The water usage thing is easily the top concern that people cite in their arguments against data centers.
(Just a side note on how things are going here: three surrounding counties have already implemented a 5-year moratorium on any new data centers and solar power plants because people see them as inextricably linked here.)
There is nothing to indicate 2.5B is too much, too little or just right
But HN replies are extraordinarily reactionary and defensive when there is any discussion of "AI" water use
The replies are reminiscent of those in threads under submissions reporting facts about "crypto" not too long ago, before SBF went to prison and the crypto hype subsided
Saying "if it's edible then it doesn't matter how much water it uses, it's justified" isn't a good position to take.
Plus California was/is in a drought for years, that's why people bring up the almonds example.
A lot of ways people use water can actually end up back in the source area after treatment. That should be considered differently than water evaporated in a desert that rarely receives rains.
Demanding closed loop cooling is just as important as demanding self built renewable power for new data centers.
I see no reason we should grandfather in 'heritage carbon emitters' when we are emitting way more than we ought to.
MrWiffles•1h ago
Edit: seems even this is just the summary first two paragraphs.