Artificial intelligence promises revolutionary solutions to global challenges, but the water costs to produce and power AI hardware and infrastructure may exceed the benefits.
Initial research shows that AI has a significant water footprint. It uses water both for cooling the servers that power its computations and for producing the energy it consumes. As AI becomes more integrated into our societies, its water footprint will inevitably grow.
The growth of ChatGPT and similar AI models has been hailed as “the new Google.” But while a single Google search requires half a millilitre of water in energy, ChatGPT consumes 500 millilitres of water for every five to 50 prompts.
So when computer systems use water, it’s typically in a closed cooling loop. The water is heated by the computer components and then cooled in a radiator before being returned to the computer components to absorb more heat and the cycle repeats.
So why do these articles always read like it’s consuming water in a way that eliminates it from existence?
As far as I’m aware they’re not taking water and turning it into something else like concrete, so what exactly is happening that it’s reducing our fresh water supply on Earth?
Data center water cooling isn’t a closed loop. They generally don’t use it like PC water cooling. There are exceptions, but servers are typically air cooled.
What they did is look for a less energy intensive way to cool the air than traditional air conditioning. So they turned to evaporative cooling, and also misting the incoming air. This reduced their energy use, but at the expense of water use.
Most of the overall amount of “operational” water that Google used in 2021 is related to these data centers; it withdrew 6.3 billion gallons during that fiscal year, according to its 2022 Environmental Report. Of that amount, 1.7 billion was discharged.
They’re evaporating away a lot more water than they return.
They’re evaporating away a lot more water than they return.
Ok, but the point is once water evaporates it doesn’t stay evaporated forever. It condenses and turns into rain or snow.
Where exactly do people think this water is going when it evaporates? Space?
We don’t have a shortage of water we can drink, we have the potential for drinking water shortages in the future assuming that only demand continues to grow. We don’t even really have industrial and agricultural shortages anywhere either, not in the developed world at least - we have areas where we’re utilizing a high percentage of available water and where water use is regulated for large consumers and areas where ground water tables are being diminished which is the only real problem.
The other key thing to remember is that water supply is incredibly geographic, the Arizona River basin for example is a common example of a water system in danger which is why someone placing a facility that uses a lot of water is unlikely to build there if they’re doing a non location specific thing like running a date center. You’d be far more likely to put it somewhere like Virginia near the Potomac where there’s no shortage at all or in germany and Ireland where again they have huge amounts of water. (I use these examples because it’s where aws data centers are)
If I don’t flush my toilet then it’s not going to help the watertable in Capetown or Arizona, but if an Arizonian telephones me and says ‘I want to do some math but my calculator uses too much water’ then I can do that math for him using the huge amount of water endlessly falling from the sky here and that will save water in Arizona.
According to the UN development program the most common form of water scarcity is economic, largely due to the complexity of infrastructure projects - the chicken and egg braindrain problem where no one that has the education to build and run such systems wants to live in such difficult areas. The internet is already moving towards improving this. Remote learning is allowing people to study to a worldclass level without needing to physically have that infrastructure locally, ai improves this hugely as it can explain things in people’s first language and as it’s abilities continue to grow this will only continue. Likewise automation of construction and monitoring allows high efficiency and environmentally sustainable facilities to be established so water availability and quality can be improved thus reducing the global water scarcity.
Also in developed nations reduced cost of infrastructure makes it ever easier to create more water efficient systems especially with active monitoring of agricultural crops and targeted water supply, etc. And as you rightly point out places with limited supply can always just build a desalination plant, they’re getting cheaper all the time and the ones currently in operation are working really well, Isreal and Jordan for example working together to refil the sea of Galilee (a fresh water lake that was declining due to over use).
These stories have no substance, they exist because ai is trending news story and everyone loves to find reasons to bad mouth anything new. Yes, It’s a valid concern that must be taken into consideration but an article is junk if it doesn’t start by explaining using water in areas with huge amounts of it isn’t even slightly a problem and that this is of course standard operating procedure when locating data centers.
Freshwater is not an infinite source. It’s something that I was taught in school in the 90s when people actually gave a shit about the planet and each other.
No. People are tracking useable water supplies. If it gets out of that, we don’t care what happens to it.
We’re draining aquifers to give people and industry drinkable, useable water (no matter how we feel about that). The water “still existing” somewhere else is an entirely pedantic point, and a huge waste of everyone’s time.
I think it goes out of the cycle somewhat, kinda like all that is held up
If 100L rain down a day, and you use 10L for cooling then you will have still have 100L flowing but now only 90L for actual use. And then datacenters take up a significant amount of the total then you have a lot less to use elsewhere such as watering fields for example.
Where I live we have this huge river around our city that provides most of the province with freshwater (along with all of the rivers that feed into it, but the population concentrates around that one big river)
That one big river is also a place for ships to go through, and an ecosystem (despite all of the disruption).
More water in use by all kinds of facilities still manages to lower the level of the river significantly, to the point where there have been worries raised about the ecosystem and where shipping capacity was reduced.
If you’re taking water and putting it in a closed loop, you’re effectively removing it from the natural water cycle until you remove it from said closed loop, no?
I agree the articles make it sound more like they are just burning water out of existence lol.
Right. Makes not sense at all, but people gobble it up the same way they blame bitcoin for the environment because the narrative is that it uses a lot of power. It’s not about the power, it is about how carbon intensive the generation method is. Such a great distraction.
The article seems to conflate AI, Farm specific AI, and technological advancements. They call out a lot of ways that tech advancements and targeted applications of AI can help save water / find new solutions. It’s lightweight on the fidnigns, but whatever. They then go to talk about data centers using a lot of water. It’s related, but not really the same thing.
I partly agree. AI has really little chance to produce anything useful if we use it the way we do now. I’m not so sure with the blockchain technology. We needed more decentralized networks in our economy and society, and blockchain is just one technology that can help here imho. It’s true that the vast majority of crypto projects represents a blend of scams and get-rich-quick schemes, but there are some fine projects that do a good job imo.
You may not be fully aware of where AI is being used. The LLMs get a lot of press for both being impressive and at the same time not living up to expectations. However, there are other AI efforts that are having real results.
I support radiology imaging at a large U.S. health system, and there are several different AI systems being tested and deployed that assist with diagnosis. It may not get a lot of mainstream articles published, but it allows doctors to treat patients more efficiently, which has the potential to both reduce costs and increase access to care.
I’m sure there are similar efforts in every other industry.
Yes, I know. I don’t say it’s all bad. It improves human decision making in a lot of things. What I meant is that it has been doing also a lot of harm in the last few years, e.g., in the U.S. where insurer UnitedHealth allegedly used an AI model with 90% error rate to deny care, or in The Netherlands and in France, just to name examples. And I’m afraid this is just the tip of the iceberg
But I’d agree that it’s not the technologies, it’s the way we humans use them.
surely nestle and coke are using orders of magnitude more. Also don’t many Americans drink bottled water when tap water is completely fine in most areas? I read somewhere it was a weird cultural quirk, you know like refusing to use the metric system. I’m not saying shouldn’t check AIs footprint, but water seems a werid one to go after.
