[gknoy]

Training AI models uses a large amount of energy (according to what I've read / headlines I've seen /etc), and increases water usage. [0] I don't have a lot to offer as proof, merely that this is an idea that I have encountered enough that I was suprised you hadn't heard of it. I did a very cursory bit of googling, so the quality + dodginess distribution is a bit wild, but there appear to be indiustry reports [2, page 20] that support this:

""" [G]lobal data centre electricity use reached 415 TWh in 2024, or 1.5 per cent of global electricity consumption.... While these figures include all types of data centres, the growing subset of data centres focused on AI are particularly energy intensive. AI-focused data centres can consume as much electricity as aluminium smelters but are more geographically concentrated. The rapid expansion of AI is driving a significant surge in global electricity demand, posing new challenges for sustainability. Data centre electricity consumption has been growing at 12 per cent per year since 2017, outpacing total electricity consumption by a factor of four. """

The numbers are about data center power use in total, but AI seems to be one of the bigger driving forces behind that growth, so it seems plausible that there is some harm.

0: https://news.mit.edu/2025/explained-generative-ai-environmen... 1: https://www.itu.int/en/mediacentre/Pages/PR-2025-06-05-green... 2: (cf. page 20) https://www.itu.int/en/ITU-D/Environment/Pages/Publications/...

[coryrc]

USA uses 21.3 TWh of petroleum per day for transportation. Even if AI was fully responsible for all data center usage (it is not even close) we're quibbling over 20 days of US transportation oil usage, which actually has devastating effects on the environment.

Data centers are already significant users of renewable electricity. They do not contaminate water in any appreciable amount.

[astrange]

There's an "AI is using all the water" meme online currently (especially on Bluesky, home of anti-AI scolds), which turns out to come from a study that counted hydroelectric power as using water.

[baggy_trough]

I agree that there is some incremental electricity usage. I do not think it can be characterized fairly as "massive environmental harm".

[autoexec]

As an example, Ren and his colleagues calculated the emissions from training a large language model, or LLM, at the scale of Meta’s Llama-3.1, an advanced open-weight LLM released by the owner of Facebook in July to compete with leading proprietary models like OpenAI's GPT-4. The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car between Los Angeles and New York City. (https://news.ucr.edu/articles/2024/12/09/ais-deadly-air-poll...)

see also:

https://www.techrepublic.com/article/news-ai-data-centers-dr...

https://www.scientificamerican.com/article/a-computer-scient...

[ahepp]

> The study found that producing the electricity to train this model produced an air pollution equivalent of more than 10,000 round trips by car between Los Angeles and New York City.

I am totally on board with making sure data center energy usage is rational and aligned with climate policy, but "10k trips between LA and NY" doesn't seem like something that is just on its face outrageous to me.

Isn't the goal that these LLMs provide so much utility they're worth the cost? I think it's pretty plausible that efficiency gains from LLMs could add up to 10k cross USA trips worth of air pollution.

Of course this excludes the cost of actually running the model, which I suspect could be far higher

[root_axis]

> 10,000 round trips by car between Los Angeles and New York City.

That seems like very low impact, especially considering training only happens once. I have to imagine that the ongoing cost of inference is the real energy sink.

[autoexec]

It doesn't happen only once. It happened once, for one version of one model, but every model (and there are others much larger) has its own cost and that cost is repeated with each version as models are continuously being retrained