Is it possibly excessive? Yes. But this is insane hyperbole.
Of course that's how you get attention, I guess. Rational discourse is ignored. To get people to pay attention you have to make everything extreme and stupid.
I ran some numbers on energy use a while back. In general: if you skip one 20 mile drive, you make up for heavy use of e.g. Claude Code for a week.
So telework one day and you're making up for your AI use.
If AI is using too much water, it means the data centers are not being charged enough for that water. Charge them more and they'll magically discover ways to reduce water consumption.
The Texas natural gas mega-data-center story: that's a political problem. Tax carbon emissions and they'll suddenly realize they can power it better with solar, wind, or nuclear power. I'd also speculate that Texas made gas power a contingency to build it, since fossil fuels own Texas.
That figure represents ~40% of extant grid capacity for the entire country in just a single state.
edit: I'd also add that historically, power generation has accounted for 4% of all water consumption in Texas. The highest recorded load demand on the ERCOT grid, to date, is ~87.5 GW. Scaling generation capacity by 500% is not possible to achieve within a short-term time frame, but even if it were, the water demand of the additional generation sites would completely drain many critical aquifers and result in skyrocketing water rates for citizens of a state already suffering terrible droughts regularly.
Everything has a downside attached. Most things are terrible for the environment if you look at the raw numbers, and if you're not interested in them or they're things you don't use then it's easy to just dismiss them as pointless nonsense that we should stop doing.
However, if you're only looking at the downside then you're being dishonest. The point is never "Let's stop this because it's bad for the planet", but "Is the net benefit worth the cost?"
That could still be a no for AI, but there are things we ought to stop doing long before we come for AI. Notably golf. It's a terrible waste of land and water, and boring.
"“This report is not a case against artificial intelligence, a technological transformation that is improving the lives of billions of people around the world,” said Professor Kaveh Madani, Director of UNU-INWEH who led the investigation team. “It is a call for using it responsibly and addressing its unintended impacts proactively to make it sustainable and equitable.”"
It is impossible to judge of the net benefit is worth the cost if you don't know the cost. More specifically, "The report finds that AI’s environmental cost is being systematically mismeasured."
Its key policy messages even include: "AI within planetary limits is achievable. The report’s central argument is constructive. Capability and stewardship can grow together, but only with measurement, transparency, and shared responsibility across the ecosystem."
You can also see it's not anti-AI by how it argues, for example: "Governments should integrate AI infrastructure into energy planning, water governance, and land-use permitting, and require standardized environmental footprint reporting."
The report itself praises AI even more, calling it "a defining technology of the Fourth Industrial Revolution, moving societies beyond digitization toward the optimization and automation of complex systems", along with an overview of its transformational benefits.
It also has an easily replaceable alternative that would reduce farmland use by 50% , massively reduce water use, eliminate one of the top contributor to climate change, and at little to no cost.
A glass of cow’s milk costs weeks worth of the average AI user’s resource usage.
So you’re right that we have far better options to tackle before we even start worrying about AI.
AI does have a particularly high local impact, due to concentrating so much of the world’s AI energy use into the land on which the data center is located, so that is a real issue to be dealt with, but that’s largely a regulatory and management issue.
FWIW, I have a hard time evaluating the validity of "A glass of cow’s milk costs weeks worth of the average AI user’s resource usage."
My back-of-the-envelope estimation, with total power and water consumed, ended up with a few un-resolvable factors.
First, what is "the average AI user"? Is that just people, or does it include businesses, militaries, the police and other government surveillance organizations, etc?
As specific points, does the data collection and AI analysis by both the NSA and McDonald's count as part of the average AI user’s resource usage?
The report says "A spam filter and an image generator are both forms of AI" so am I an AI user because I have a Bayesian filter built into my email client? Because if so, my minuscule power use is nothing like the what's needed for typical image generation, causing the overall average use to decrease.
Second, the wholesale prices I see for power, water, and milk, are all subsidized values, either from direct government subsidizes (like funding dams in the US southwest, or farm subsidizes) or from incomplete pricing (like the environmental impact of burning natural gas for AI). How do I even convert a hundredweight of milk for the wholesale price into the size of a glass? Or did you use the retail price of milk?
In other words, how did you end up with your conclusion?
there have been plenty of articles about golf courses being a threat to water sources; just because golf courses haven't been closed doesn't mean we shouldn't be concerned about the threat posed by AI. You're still free to crusade against golf courses if you like.
This ignores that no one drives to a movie theatre or play every day or even several times a day but this is a common usage pattern for Netflix or that Netflix is not really an alternative to these activities but just another form of entertainment piled on top.
these are bad examples, because it would be good if they were all gone, except almond farming in areas naturally hospitable to almonds.
So rather than ask "What is the impact of AI?" which narrows your focus to the impact of AI, if you really care about the natural resources you should be asking "What has a big impact on natural resources?"
If that's AI, and you're ignoring all the things that have a far greater impact on what you claim to care about, then you should ask if you're really just looking for things to complain about around AI, and natural resources are a convenient excuse.
For example, the largest data centre in the US uses approximately the same amount of water as 2 burger restaurants if you include the water used to raise the beef used in the burgers. (https://newsletter.semianalysis.com/p/from-tokens-to-burgers...)
If you care about water resources and you're not a vegan campaigning against the beef industry, but you're focusing on AI, then you have a degree of cognitive dissonance.
If you're complaining about resource usage, but only when that resource is being used by one thing you don't like, you're just a tool for somebody else's agenda.
Specifically, if someone ACTUALLY cared about foolish uses of freshwater, they'd be advocating for reforms in water rights that result in growing extremely water-inefficient crops in semi-arid and desert environments. Mostly almonds and alfalfa, especially when these are for export.
One would also make the distinction of where the water came from important. Some places are flush with fresh water and its usage isn't terribly important, other places are deserts.
Anyway there should just be a "sin tax" for using freshwater for frivolous things like golf courses in the desert and datacenters who don't want to invest in less wasteful cooling solutions.
People act like doing things that are good for the environment is inherently lacking in economic value. My Dad’s environmental consulting business put me through college, while doing good things for the environment and people, and that is really small scale stuff. And his primary customer was a bank wishing to avoid lending on contaminated sites.
There could absolutely be massive money to be made in doing good things for the environment, it is the people making money doing bad things for the environment who do their damndest to keep grasp on the wheel and subvert any attempt to make doing good things for the environment economically advantageous.
These people are virtually selling their children’s future for a few extra bucks today. The earth is going to be around for a long time yet. We should care about our children, grandchildren and so on.
Maybe if we structured our economics around long term plans we wouldn't have such a mess.
But just because I like AI doesn't mean I need to go with the crowd and downvote anything that threatens my livelihood. We can think of solutions instead of just reacting all the time.
Weird take.
> That could still be a no for AI, but there are things we ought to stop doing long before we come for AI.
Hard to have that discussion if these are the responses.
Don't threaten my communist soul with a good time.
>Everything has a downside attached.
Yes. Let's regulate that, please.
>The point is never "Let's stop this because it's bad for the planet", but "Is the net benefit worth the cost?"
Sure. The answer is 'no, it isn't '. These frontier models purposefully chose the most power guzzling, moat digging approach and it was not necessary whatsoever.
> but there are things we ought to stop doing long before we come for AI.
Sure, let's get to it. As well as with AI at the same time.
Is any benefit worth the cost of killing the world? The article proposes a series of recommendations for investors in the "roadmap for sustainability and equity", because the conversation isn't if the benefits of AI are worth the costs, but if we are able to pay these costs at all plan the required infrastructure improvements -- which are always done with taxpayer money regardless of how neoliberal the society claims to be -- in order to accommodate the expansion of AI without massive disruptions.
Golf courses are not booking revenue by exchanging shares in themselves to other golf courses in exchange for golf balls that have not been manufactured yet.
From an environmental perspective we should ban golf and use the land for literally anything else*.
* maybe not data centers.
If they truly wanted to have an honest discussion, they would have outline both the advantages and disadvantages and start a genuinely balanced public debate. The report is more like a doomsday fear-mongering for a lack of better words propaganda.
The sad part is that reputed organizations like UN does this kind of things, it just reduce the trust of public in these orgs even more.
But humans have driven multiple species to extinction and destroyed pretty much all their natural predators. They didn't do that by being stronger or faster or having bigger teeth and claws. They did it by being smarter than the dumb animals.
So it's clearly and obviously a terrible idea to create something smarter than us, in case we get the same treatment.
The report I read does outline the advantages, starting off in the executive summary, after the AI-generate image, with:
"Artificial intelligence (AI) has become a defining technology of the Fourth Industrial Revolution, moving societies beyond digitization toward the optimization and automation of complex systems. AI systems that learn, perceive, reason, and generate content are now embedded across economies and daily lives, supporting services from translation and recommendations to finance, healthcare, energy management, and transportation."
It later says:
"""... AI enables tasks ranging from data processing and translation to diagnosis, forecasting, and autonomous decision-making. Today, AI can analyze massive datasets in seconds, translate languages, detect diseases from medical scans, forecast climate trends, and power everything from personalized recommendations to self-driving cars.
By executing complex operations at speeds and scales far beyond human capacity, AI is more than just a technological tool—it is a transformative force. It is reshaping economies, redefining labor, and influencing how societies interact with technology and with the planet. As AI’s influence accelerates, so too does the urgency to understand its full impact: not only the promise it offers, but also the profound challenges it raises.
AI is a powerful driver of the Fourth Industrial Revolution (4IR)1, or Industry 4.0—a global transformation marked by the convergence of digital, physical, and biological systems. Propelled by rapid advances in everything from AI and robotics to cloud computing and the Internet of Things (IoT), this revolution is reconfiguring how people live, work, and are governed. Yet while much attention has been given to AI’s capabilities and risks, including bias, misinformation, and automation leading to loss of jobs, far less focus has been placed on AI’s environmental impact.
... The launch of ChatGPT in 2022 was nothing short of a technological shockwave. ... Once a niche technology, AI is now woven into daily life, powering much of the digital world, including voice assistants, search engines, recommendation systems, smart energy meters, and smartphones. The influence of AI is transforming how people work, communicate, access information, and consume services. ... AI is employed in diagnostics, medical imaging, and patient risk prediction—improving speed and precision of care, while reducing treatment costs. ... AI helps forecast demand, manage power grids, and integrate renewable sources into power networks. ... AI is also revolutionizing transportation and mobility."""
Could you clarify how that's not a good outline of the advantages, and elaborate what specifically you think is fear-mongering?
What in this report is making your reduce your trust in UN?
My guess is that there’s someone making money for getting DCs permitted. Like the paperwork done.
474 GW of capacity is silly woo woo land. I’m not sure you could get enough chips made to use that in the next 10 years. The cost of fuel or solar/batteries or reactors or gerbils or whatever they plan to use to make that would be bonkers.
This report praises AI as a part of the next technological revolution, but points out there are problems that should and can be addressed to make its benefits more equitably distributed, to minimize its negatives, to promote transparency in resource use, and to address sovereignty issues.
Discussion of water rights reform, water use for almonds, alfalfa, and golf courses, etc. are completely irrelevant to the goal of this report.
/S
True, but do you add chemicals to that water, and does it becomes unusable at the end of the day? At worst, that water gets in soil and goes somewhere else, and some of it evaporates.
Open loop datacenter cooling is not like that though (you add tons of chemicals to that water to prevent scaling and to kill algae), and closed loop is not that harmful, but big AI datacenters don't use that because it's wee bit more expensive.
The most "water hungry" sectors can reuse that water, or at least redirect it for agriculture. AI Datacenters are different. They are built in the cheapest way, hence they are very dirty both in carbon footprint and water-body fouling.
Golf courses at least trap some carbon by growing grass. It's almost bitterly funny in comparison.
For clarity, it seems possible to filter out these chemicals (alongside everything else) with reverse osmosis, but companies which don't invest in closed loop systems (which reduces water demand by ~90% BTW), won't invest in reverse osmosis, which is way more expensive w.r.t. closed loop systems, I believe.
If someone has more precise resources or information which they can point me to, I'd love to update my knowledge.
Also, this doesn't negate CO2 emissions of these plants.
You just think that because it doesn't know about your tiny little niche in as much detail as you, that makes you smarter.
If I gave a general knowledge test to an AI vs you, who would win?