Talk:Environmental impact of AI

Ok, i've updated the template with the corrected source for the first one (I guess I mixed up the two scientific american articles, I think they bothy say that GPT-3 releases 552 metric tons of Co2, but only one talks about the comparison to cars. Thanks for catching that! Also fixed ALT1, thanks for fixing that as well! Apologies for incorrect info, I'll try to read a bit slower with these articles. Bluethricecreamman (talk) 21:45, 10 July 2024 (UTC)

I didn't dig to hard on that spammy site earlier, sorry. I agree that they stole from enwiki and not the other way around. All three hooks are good to go. I think ALT0 is the most compelling. awkwafaba (📥) 01:05, 11 July 2024 (UTC)

• Is this true? jp×g🗯️ 22:59, 12 July 2024 (UTC)

• Yes. The sourcing for the release of CO2 by a lot of AI models is well established, and it was a news item last week that Google's emission had increased by 50% or something cuz of their big AI push. Bluethricecreamman (talk) 23:03, 12 July 2024 (UTC)

Specifically, I am a little concerned that statistics-heavy topics like this, when they're directly related to buzzwordy trending topics in the news cycle, tend to be very ill-suited to being patched out from headlines. That is to say, writing the shark attack and dog attack based solely on news coverage would give the impression that these were both uncommon things that do happen every once in a while, when in reality the yearly rates are eighty versus several million. This article quotes a bunch of figures from news stories back-to-back, and there doesn't really seem to be much attempt at comparison or context. For example, there is a paper cited that projects 85-143 TWh of global power consumption from neural networks by 2027, based on a conjectural optimistic scenario where NVIDIA/TSMC transition their entire manufacturing output to A100s (all of which are installed in data centers and used for nothing but LLM inference). By comparison, this paper estimated that video gaming consumed 34 TWh/year in the United States alone in 2019; this document from the DoE says that aluminum production in the US in 2000 consumed a total of 279.2 TWh/yr. The US consumes about 16% of the world's electricity, so a very rough approximation here would give a ballpark estimate of 212.5 TWh for gaming (the US does about 2% of aluminum production worldwide, giving us around 13960 TWh yearly for global aluminum production). It also seems like there is a lot of ambiguity between training (a fixed cost, as models are only trained once) and inference (the process that happens when running the model, which typically requires billions or trillions of times less computation). I am somewhat worried about having this on the front page. jp×g🗯️ 23:45, 12 July 2024 (UTC)

you are literally using us energy gov stats to WP:SYNTH a whataboutism argument… by the same logic, the criticism section of any industries environmental impact can be called misleading because theres always another more polluting industry you can point to. the comparison you suggest is “missing” is literally that training ChatGPT training took as much energy as 123 gasoline powered cars yearly footprints. and suggesting that including examples from news articles headlining the carbon footprint is a bad thing for wikipedia is inane. most articles on here use secondary souces such as news articles. and most DyK hooks are eyecatching Bluethricecreamman (talk) 00:34, 13 July 2024 (UTC)

For the specific task of literally writing an encyclopedia article about a topic, I would say that answering questions like "why is this statistic relevant" or "what does this number mean" is probably its most important purpose. I gave a couple examples to illustrate what this looks like, obviously there is no reason why aluminum manufacturing in particular is more relevant than anything else. The most appropriate source here would likely be something like a textbook or monograph about electrical consumption by various industries, and how this related to environmental concerns in general. Since this is a subject of rather large significance, it is important that our writing on it be accurate, and not news. I am opposed to running stuff on the main page where vague insinuations are made by factoids of unexplained significance -- e.g. what's a terawatt-hour? what's a gram of CO2? why are some figures given in one and some as the other, and still others given in folksy derivative units like average midsize sedan gasoline consumption rates over average suburban commutes? jp×g🗯️ 01:06, 13 July 2024 (UTC)

Moreover, ALT0 is not supported by the source. Indeed, OpenAI's well-known sleazy refusal to be transparent about any of its operations or research means that the source cannot say that, because the source does not know. The figure claimed as fact in this hook is, in the source, carefully and explicitly presented as an approximate figure derived by estimation (it's from https://arxiv.org/abs/2104.10350 -- note the methods they are using). jp×g🗯️ 01:16, 13 July 2024 (UTC)

I'm not arguing these bad faith arguments after this. The source link is literally right there for anyone to read. There are three links I cite with the 552 metric tons figure in the article. Most of the "environmental impacts of" are similar collections of "factoids" discussing various industry leaders. And by your logic of WP:NOTNEWS, we should remove most news articles discussing long term trends in any industry. WP:NOTNEWS specifically states no original reporting on wikipedia (we don't report the news ourselves, we cite it), and calls for enduring notability of information (we don't do an article if its just one or two articles). if you find a reputable source arguing that this is a false number or a false comparison, feel free to edit the article. I respect your work as an admin, and your experience, but this is silly. Bluethricecreamman (talk) 01:35, 13 July 2024 (UTC)

I left the following feedback for the creator/future reviewers while reviewing this article: Nice article!!

Broc (talk) 11:16, 3 July 2024 (UTC)

These are clearly different topics, but the large energy usage could be a short background for the CO2 part. What I removed maybe could go in that separate section.

For example, AI in Iceland might have practically no "environmental impact" other than the building of new geothermal plants. ---Avatar317^(talk) 23:00, 3 July 2024 (UTC)

Agreed. we can have a section about how AI methods may have differing footprints based on whether they are run in data centers close to renewable energy sources. Bluethricecreamman (talk) 05:25, 4 July 2024 (UTC)

Another possibility is to rename the article to "Energy usage and resulting environmental impact from AI" since I haven't heard of ANY other negative environmental impacts discussed other than energy usage. Are there any? ---Avatar317^(talk) 23:04, 3 July 2024 (UTC)

I've added in a water usage section. I suspect there is e-waste involved with the chips used for AI, but can't find anything yet... I'm assuming GPU chips are still new enough that e-waste isn't a huge issue.

Main issue with making this article is we don't have a real environmental impact of computing article, just a redirect to green computing. This deserves its own article with all the sourcing, but it also should have context of energy/water usage/ewaste of data centers/ Bluethricecreamman (talk) 05:27, 4 July 2024 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 27 August 2024 and 6 December 2024. Further details are available on the course page. Student editor(s): OwenGaspard (article contribs).

— Assignment last updated by OwenGaspard (talk) 16:46, 25 October 2024 (UTC)

I think this paper in Nature is relevant to this article.

https://www.nature.com/articles/s41598-024-54271-x 129.10.78.216 (talk) 20:32, 31 January 2025 (UTC)

Firstly, this was absolutely not published in Nature— it is published in a journal called Scientific Reports which is owned by the same parent company. It has a long history of shoddy peer-review and taking a very long time to retract obvious BS. Look at the controversies section on that journal’s page: falsely saying homeopathic pain medications worked in rat studies, falsely claimed that the HPV vaccine caused brain damage in mice, and the list goes on and on.

Secondly, the title is clickbait. As far as I can tell, they just punched some existing data into some naive formulas, and the conclusion is not what the headline connotes, which is that using AI to create illustrations and write will lower carbon emissions. That’s not what the paper actually says.

And their methods are beyond reductive:

“For the human writing process, we looked at humans’ total annual carbon footprints, and then took a subset of that annual footprint based on how much time they spent writing.”

Since even by their simple calculations the carbon output of humans is constant, unless there’s some society out there producing new humans based on the per-word demand for writing, or some humans can stop their carbon output when not writing, there is no scenario in which carbon output is lower because of AI. The paper doesn’t even consider the useable yield per image or per word of AI output vs human output, which would be crucial for making any point that related to practical use. If the information in that paper is useful in this conversation at all, it’s going to have to make a very explicit point about exactly what’s being compared and to what end.

This is one reason Wikipedia prefers WP:SECONDARY. You can skim a thoughtful article from a reputable source and get a sense for how worthy this paper is for its cited use, but actually reading it takes a lot longer and most people won’t make it past the misleading headline. Andythechef (talk) 04:27, 25 March 2025 (UTC)

There are secondary sources which cite Tomlinson: Kansas University and also ICEF, a platform founded by the Japanese governement. So we could parafrase these secondary sources? PJ Geest (talk) 14:05, 12 November 2025 (UTC)

Scientific reports is generally legit and that they retract is technically a sign of editorial control. Woukd beed to read paper myself User:Bluethricecreamman (Talk·Contribs) 14:36, 12 November 2025 (UTC)

Support including source. The journal appears to be peer-reviewed. And the "simple calculations" involve citing other research, so the article could be considered a secondary source. It makes a statistical inference rather than collecting direct observations, but I don't see why that would exclude it. WeyerStudentOfAgrippa (talk) 15:15, 27 November 2025 (UTC)

full disclosure, just realized the paper i added right after the journal is also from scientific reports.. but is written by a high schooler.
as they are both in the same journal, i think it should be fine, and apparently Saratoga High School (California) is some insane top tier highschool, but should still be noted User:Bluethricecreamman (Talk·Contribs) 17:29, 27 November 2025 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 23 January 2025 and 23 April 2025. Further details are available on the course page. Peer reviewers: Rysmith1273.

— Assignment last updated by Hungryhippouw (talk) 11:50, 2 April 2025 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 7 January 2025 and 15 April 2025. Further details are available on the course page. Student editor(s): Liamxelias (article contribs).

— Assignment last updated by Liamxelias (talk) 19:11, 25 March 2025 (UTC)

This page might benefit from a dedicated environmental justice section. This article raises concerns around third-order environmental impacts of LLMs, including the exclusion of Traditional ecological knowledge from the generated responses and/or database of popular chatbots like ChatGPT. Alternatively, I could see this being filed under the Climate Solutions page. — Preceding unsigned comment added by Liamxelias (talk • contribs) 18:13, 28 March 2025 (UTC)

The Water usage section seems to sometimes arbitrarily mix the terms use/usage and consume/consumption. I assume the water still exists as liquid water or water vapour after use. I'm guessing that as liquid water it is, for example, at least as clean as before use. How much warmer is it, and if returned to watercourses (for example) what's the environmental impact of that? How much of it is evaporated? Does that water vapour add to local rainfall? Are there initiatives to utilise the heat added to the cooling water? The section doesn't inform me as much as I'd like about these topics. 195.112.25.23 (talk) 15:14, 22 April 2025 (UTC)

if there is a source that talks about the distinction in the context of computing or AI, sure.

Until then, it is WP:OR to argue that water isn't being used because it isn't being annhilated, or salted, or polluted. We really cannot include such in the article. Bluethricecreamman (talk) 15:21, 22 April 2025 (UTC)

It seems clear to me in the first statement in that paragraph: "Cooling AI servers can demand large amounts of fresh water which is evaporated in cooling towers." ---Avatar317^(talk) 21:58, 23 April 2025 (UTC)

It seems to me that the real issue is waste heat, then, not water as such, because it is not "used", just evaporated, whose relevance remains unclear. --Florian Blaschke (talk) 22:00, 22 April 2026 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 30 January 2025 and 23 May 2025. Further details are available on the course page. Peer reviewers: SD2125!.

— Assignment last updated by Sadferret (talk) 17:32, 22 May 2025 (UTC)

Can this section be reviewed please? AI climate solutions listed are hypothetical and from this wiki and sources listed, it's quite contradictory to use AI to solve climate change or natural disasters when AI is contributing to it. Love-And-Gratitude-To-The-Earth (talk) 19:55, 15 July 2025 (UTC)

be the WP:BOLD change you wanna see.

make sure to use reliable sourcing Bluethricecreamman (talk) 20:05, 15 July 2025 (UTC)

Would you say it's also contradictory for humans to solve climate change when we're contributing to it? Of course not, because humans can have a negative and positive impact on climate change. It's the same for AI. Paditor (talk) 12:27, 9 August 2025 (UTC)

I don’t know whether it was correct when it was written but now that the global increase in electricity demand is being met by clean generation (as I just added to this article) I contend that AI does not have a significant carbon footprint. Chidgk1 (talk) 06:46, 7 October 2025 (UTC)

Your addition looks like WP:OR to me; my (occasionally simplistic) standard statement is: "Unless a source mentions the subject of this article "XXXXXXX" than it is disallowed WP:SYNTH to relate it to this topic."

For the record, I heard some interview with people talking about how tech companies were investing in solar and other renewables to CLAIM that their data centers are powered by clean energy, but the fact is that (since we have very little storage capacity-batteries or pumped hydro storage) electricity needs to be generated when it is used, and large pulses of solar being dumped on the grid (so much that we have solar Curtailment (electricity)) in the middle of the day doesn't negate the need for fossil-fueled power to generate when these renewables are not generating. Essentially: data centers cannot (and do not) run (economically) on exclusively renewable energy. ---Avatar317^(talk) 00:43, 14 October 2025 (UTC)

agreed. reverted it. User:Bluethricecreamman (Talk·Contribs) 00:53, 14 October 2025 (UTC)

I have also reverted their removals. Their reason for removal was lack of "quotes" from the article, but the context was clearly in the source's 4th paragraph. Anyone have time to look at the rest of their edits? DN (talk) 02:31, 14 October 2025 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 28 August 2025 and 17 December 2025. Further details are available on the course page. Student editor(s): MrYeticabra (article contribs).

— Assignment last updated by MrYeticabra (talk) 03:33, 4 November 2025 (UTC)

Can someone fix it please? ThrowawayEpic1000 (talk) 00:01, 5 November 2025 (UTC)

Regarding this sentence:

"A request made via ChatGPT, an AI-based virtual assistant, uses 10 times as much electricity as a Google Search"

The original source for this claim is back from 2024 and now outdated, since Google now automatically implements AI in its searches. Frieden.innit (talk) 11:18, 2 January 2026 (UTC)

Please see WP:NOR; if you can find a newer source which says differently, than feel free to update that. ---Avatar317^(talk) 01:45, 7 January 2026 (UTC)

I'll change sentence. User:Bluethricecreamman (Talk·Contribs) 17:20, 8 January 2026 (UTC)

The following Wikimedia Commons file used on this page or its Wikidata item has been nominated for deletion:

• SinemGörücü-SiliconLandscapes-2560x1810.jpg

Participate in the deletion discussion at the nomination page. —Community Tech bot (talk) 23:21, 29 January 2026 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 26 January 2026 and 29 April 2026. Further details are available on the course page. Student editor(s): ND902353418 (article contribs).

— Assignment last updated by Nschafe3atND (talk) 02:49, 17 February 2026 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 5 January 2026 and 20 March 2026. Further details are available on the course page. Student editor(s): MerliaSummer$07 (article contribs). Peer reviewers: Slothie732.

— Assignment last updated by Juniper37 (talk) 21:04, 25 February 2026 (UTC)

I added to the water usage section to try to improve the coverage of AI's water usage because it is becoming a hot topic of discussion. Specifically, I added finding from a 2025 peer reviewed study on public awareness of AI's water footprint. Added Bloomberg and Reuters reports on daily water usage and regional drought impacts, which includes projections in levels of water stress. As well as adding a part about the lack of disclosure from these AI companies on the amounts of water that they are using. All of these edits use reliable secondary sources or peer reviewed articles that follow guidelines. ND902353418 (talk) 17:16, 3 March 2026 (UTC)

Some recent revisions were undone in "good faith" but it seems doubtful whether the arguments for undoing them are plausible.

https://en.wikipedia.org/w/index.php?title=Environmental_impact_of_artificial_intelligence&oldid=1343201978

1) the source "beyondfossilfuels" might not be high on the quality list, but the report was commissioned by an alliance of environmental NGOs and can also be found on other domains. Given the report was done by an analyst with references to IEA data etc it a meaningful contribution to the article much better than — as it current stands throughout the article — blurring the lines between genAI and other AI applications without acknowledging that these are plausibly different. This is important to note; but perhaps someone has better sources for this; and perhaps the article needs a makeover to cover that.

2) The revision suggests that "The section on "Systemic environmental implications of generative AI use" is not an "environmental effect of AI" and therefore doesn't belong in this article." The statement was contextualised with a high quality source on GHG emissions related to AI, which includes physical/material infrastructure, deployment, but also environmental effects of intended use and systemic effects. (https://www.nature.com/articles/s41558-022-01377-7)

The wiki article in fact already acknowledges that systemic effects of AI are environmental impacts, otherwise the section currently titled "Use of AI in environmental research" does not make sense either: it assumes that there are potential positive environmental effects of intended use of AI. This is by and large as uncertain as potential negative effects of intended use (except maybe that positive narratives appear more trustworthy because aligned with current hype and zeitgeist), which systemic impacts are part of. Misinformation/disinformation is definitely one such systemic impact that is not limited to climate impacts. However, disinformation was also only a small part of the systemic impacts section, together with other sources that e.g. argue how and why systemic impacts due to the ordinary function of gen AI software should be considered part of gen AI's Scope 3 emissions (https://journals.sagepub.com/doi/10.1177/20539517251365226) — this is not limited to genAI capabilities of generating plausible-sounding text without factual basis, i.e. misinformation, and their environmental consequences. Anyway, the source is not limited to misinformation but touches upon other environmental effects of "intended use".

So perhaps the text should have been improved rather than deleted based on a glance that does not hold up to the relevance and relatedness of the content and the trustworthiness of most if not all sources. But perhaps the whole article needs a complete makeover first. ~2026-16057-66 (talk) 10:17, 14 March 2026 (UTC)

1) the beyondfossil fuels link really isn't suitable, and NGOs need to WP:ATTRIBUTED if they are even WP:DUE. best to look for another source or drop it.
2) I think that's easier to argue and I'd support that inclusion. @Avatar317, thoughts? User:Bluethricecreamman (Talk·Contribs) 15:44, 15 March 2026 (UTC)

I still oppose inclusion of the "Systemic environmental implications of generative AI use" that I removed. The statements are basically WP:CRYSTALBALL conjecture along the lines of "this new tech COULD do bad things". Almost identical statements could have been made decades ago with a find/replace of "AI systems" with "the internet", and then subsequently with "social media".

Text from my reversion/removal: "Generative AI technologies can generate..." Title: "AI could create..." "As generative AI is increasingly used to collect or gather information by users, it can present them..." "Outputs might also affect..."

It brings up the DUE question, who is speculating this and are these opinions/conjectures widely held/believed by those researching AI as potential problems. I think the misinformation definitely is seen as a potential problem, but it would cover any type of misinformation, not restricted to climate misinformation. ---Avatar317^(talk) 00:00, 17 March 2026 (UTC)

its not really crystalball, as according to that section anticipated events must be verifiable, and the subject matter must be of sufficiently wide interest that it would merit an article if the event had already occurred..
also, can and could is not about the future, its about the potential of the tech, which is inherent to any article we have about AI. User:Bluethricecreamman (Talk·Contribs) 01:09, 17 March 2026 (UTC)

I agree with this thought. It seems to me that terminology in concern is about writing style. The section on AI in environmental research is similarly empty and full of speculative promises, just perhaps hung up on more concrete examples. I am happy to add more concrete examples if that is what is needed to recover that section.

"Beyond generating climate insights, AI technologies are being deployed in emission reduction efforts. Google has developed Project Green Light, which uses AI to optimize traffic light timing to reduce stop-and-go traffic, potentially decreasing vehicle emissions at intersections." On that note, Google and others also improve oil drilling with AI, which is a potential negative environmental impact.

It continues: "Some climate scientists^[who?] have suggested that AI could improve the efficiency of renewable energy systems, though concerns remain about AI's own substantial energy requirements."

The next paragraph employs similar broad brush assumptions as the misinformation point brought up: "AI is also being applied to genetic engineering. An AI tool called Social LEAP Estimates Animal Poses (SLEAP) is being used to improve the carbon sequestration of plant root systems." Here we might wonder, if this should not go into the plant root systems article if the misinfo should not be also placed here? Additionally, to fulfill its potential, the gene edited plants would need to be taken up at scale which given regulatory contexts and the inertia of farming systems is quite unlikely. Is that not crystal ball then?

And then later: "AI is also being explored to improve sustainability in environmentally intensive sectors such as food production and fast fashion by optimizing supply chains and reducing waste. However, comprehensive frameworks for evaluating the net climate impact of AI systems, accounting for both their energy costs and potential environmental benefits, have yet to be widely adopted." — Here, one could add that sectors like fast fashion and other consumer facing sectors also use try AI tools to sell more stuff to people. Similarly speculative, but definitely part of a "net climate impact".

And for many of these (like the traffic light example above), the effect is similarly systemic, i.e. a plausible effect on societal level if the technology is deployed on large scale. I maintain that these plausible effects cannot be only to include both positive examples, but must also include negative ones.

So I suggest to recover that section on systemic impacts. One could intersperse academic literature on these (which I know is very sparse) with examples from news reporting if that makes it less "crystal ball". Although I don't think the structure quite works, so the article could be rearranged to separate direct use impacts (Scope 1), upstream supply chain (Scope 2), and downstream supply chain (scope 3) which would include both intended use and systemic effects of intended or unintended use, the latter then including all that research text. ~2026-16057-66 (talk) 14:25, 17 March 2026 (UTC)

1) I don't disagree that the article CURRENTLY has a lot of conjecture similar to your addition. Maybe that stuff doesn't belong either, some looks like overly promotional fluff.

2) I still think your addition works better in the Disinformation/Misinformation articles; couldn't AI also be used to generate misinformation about war (which it is being used to do as recently reported about the Iran war) to create more war which also has huge environmental impacts?

3) Personally, I think this article should ONLY talk about DIRECT impacts (energy and water usage), though others may disagree, and if so, then I like your suggested structure to separate the article into: first direct impacts, then upstream supply chain, and finally downstream THEORETICAL impacts, while making clear that the downstream negatives involve a lot of "if sufficient guardrails are not in place" - as really can be said for any and all new tech. ---Avatar317^(talk) 18:02, 17 March 2026 (UTC)

For point 1, If we are gonna cut all “conjecture”, submit an AfD for Artificial general intelligence and Technological singularity^[sarcasm]

CRYSTALLBALL includes the statement that reliable sourcing allows for inclusion .

for 2 and 3 i might have more sympathy for some of those arguments. User:Bluethricecreamman (Talk·Contribs) 18:47, 17 March 2026 (UTC)

The graph based on Epoch AI was removed on the grounds that it is not a reliable source:

However, there are several indications that Epoch AI meets reliability criteria under Wikipedia:USEBYOTHERS. Epoch AI’s data are used by reputable secondary sources such as Our World in Data () and Statista (). In addition, major media outlets cite their work; for example, Time Magazine explicitly describes Epoch AI as a research organization: Their data also appear in academic publications, such as: Tibrewala, Alok (2025)

Finally, the estimate of ~0.3 Wh per request is broadly consistent with other published estimates, which are already included in the article:

• Google Gemini (2025): ~0.24 Wh per median text query
• University of Michigan (Llama 3.1): 0.03–1.86 Wh per response depending on model size
• Luccioni et al. (2024): ~0.05 Wh per request (text generation)

PJ Geest (talk) 19:25, 19 March 2026 (UTC)

There are a few issues with how this article is set up that might be worth looking at:

• There are opinions and conjectures throughout the article stated in wikivoice or only attributed using weasel words.
• WP:STRUCTURE issues are created when "Climate solutions" is given special attention in its own section, essentially creating a "counterargument" against the rest of the article, instead of incorporating the content in the relevant sections.
• A lot of the article is an indiscriminate list of statements as opposed to a survey of the literature. This is especially the case with opinions; an opinion is probably WP:UNDUE and a WP:MINORASPECT if the only source is the person or group who expressed it. This often happens when many editors come along each adding the thing they consider important rather than mining an overview source, but editors sometimes fall into the trap of compiling a collection of isolated facts and opinions.
• The section I checked in more detail, water usage, doesn't accurately describe the current understanding of water usage. Namely that there isn't a single clear understanding because it can be measured in different ways that produce different results. See IEEE Spectrum and Resources, Conservation and Recycling for examples of sources that go into this specific issue. This sort of thing is a consequence of the aforementioned "indiscriminate list" problem.

Thebiguglyalien (talk) 02:47, 27 March 2026 (UTC)

This rewrite improves neutrality and sourcing, which is welcome.

However, I’m concerned that the reduction in size may have gone too far (removing roughly 70% of the text) and that this has led to the loss of useful empirical detail and variation present in the literature. In particular, the current version seems to underrepresent differences across models, tasks, and measurement approaches, which were well documented in the previous version. This may raise a due weight issue, as the topic is characterised by substantial variation depending on assumptions.

The text has also become relatively abstract, whereas concrete examples in the previous version often helped illustrate the scale and variability of impacts.

It would be useful to find a way to retain the improvements in neutrality, while also preserving the most informative examples from the previous version. PJ Geest (talk) 09:15, 3 April 2026 (UTC)

Hi PJ Geest, I think that is a fair point. My aim in the rewrite was to remove weak sourcing, promotional tone, and overly news-driven material, but I agree that the article may now be too abstract in places, and underrepresents the degree of variation in the literature.

I think the best next step is selective re-expansion of certain sections. In particular, I plan to restore a few representative quantitative examples in the “Individual level”, “System level”, and “Water usage” sections, so that the article better reflects variation across tasks, models, and accounting methods, while keeping the newer sourcing and neutral tone.

I also agree that examples are useful where they illustrate methodological differences, so I will focus on adding back well-sourced cases that clarify those differences rather than simply increasing length. Andromedy (talk) 14:14, 3 April 2026 (UTC)

Thanks for your work on the article so far, and I agree with you both about how to proceed. Nuance is informative, since statistics like averages don't give a true understanding if you have a bi-modal distribution, for example. ---Avatar317^(talk) 01:07, 4 April 2026 (UTC)

I doubt I will write any more myself as I would like to edit other articles which fewer editors are interested in. But I thought I would spend some minutes to write a comment here in case others are interested.

The sentence “Location can materially affect the significance of data-centre water use.” is very important. Presumably location is also very significant for carbon footprint? Article could perhaps explain why we don’t put most data centers in secure, cold, wet countries with green power like Iceland. Is that because governments insist on data staying in their territory? Chidgk1 (talk) 08:25, 4 April 2026 (UTC)

• Meta data center water ... closed-loop cooling system purge spread rare metal-resistant bacteria in Cheyenne’s water system
• Law breaking and bending plays

This I think is important and should be added, because it's what allows this negative environmental impact. AI companies are immensely rich and consequently poweful and obviously they get away with that. So I think it's important to note that.

• xAI Colossus's installed 30+ natural gas power plants ("portable" - on tracks) but they only had a permit for 15, and later again when caught, played a fool and left 20+. Used a loophole of "364 days stationary relating to portable equipment" - not an accident but a deliberate play by xAI - they applied for 15 and took advantage of the loophole. did apply for a permit for 15 generators – even though it had been running up to 35 generators on-site, violating the Clean Air Act.
  
• "permit by rule" pre-construction authorization play that avoids public notice requirement and public input so they can build things out of nowhere and when they have a foothold they ask for higher level permits and then it's very difficult even impossible to stop them if the construction has already begun.
  
• politicians signing non-disclosure with private (AI) companies (it seems it's not illegal but it's obvious it should be) - so residents find out something is being built when it's too late.~2026-39396-77 (talk) 22:56, 12 July 2026 (UTC)