Florin HaslerMar 30
Rainhard

Wow:

arxiv.org preprint: "The data heat island effect: quantifying the impact of AI data centers in a warming world".

Excerpt: "We estimate that the land
surface temperature increases by 2°C on average after the start of operations of an AI data centre, inducing local microclimate zones, which we call the data heat island effect."

https://arxiv.org/pdf/2603.20897

"AI data centres can warm surrounding areas by up to 9.1°C. Hundreds of millions of people live close enough to data centres used to power AI to feel warmer average temperatures in their local area."
https://www.newscientist.com/article/2521256-ai-data-centres-can-warm-surrounding-areas-by-up-to-9-1c/

Mar 30 at 1:24pmWeb
Show threadChristine Lemmer-WebberMar 30
@rainhard Holy moly! I absolutely believe it (esp having worked in some of the first mega-datacenters early in my career). But wow
Show threadBee O'Problem Mar 30

@rainhard I wonder how that compares to something like a foundry or blast furnace. Not as a whataboutism but as a point of comparison since those numbers look pretty bad and being able to say "hey, these places are as bad as literal hell on earth" might drive home the point a bit.

My own cursory searching didn't turn up any specific studies about that sort of industry's effect on microclimate though.

Show threadGalactic StoneMar 31
@rainhard I'm all for warming the data centers as much as possible. 🔥 🔥 🔥
Show threadMarc Leobet5d ago

RE: https://infosec.exchange/@rainhard/116318344437253051

@rainhard @ced
I read the article because I disbelieved one datacenter could do it. So, to add some keys for people :
- it is arXiv (no review) but with famous universities/research centers;
- it is similar to the urban heat island, with an +2 impact locally with a -30% at 7 km (and that is still locally huge!)
- it is focused on areas with data centers concentration (especially in warm countries? Mexico, Brazil, south Espana quoted).
Note : I think the authors calls "AI DC" what were mostly "DC", because AI had quite no impact before, say, 2020-21, and hyperscalers came really after 2012 (following https://ocient.com/blog/the-history-of-hyperscale-in-computing-from-data-centers-to-software-workloads/) (the study watch 2004-2024).

Show threadMarc Leobet5d ago
@rainhard @ced About the second quote : the author, Andrea Marinoni, is the first author of the first article. However, she dramatically stressed what could be happen while the first article warns : "The fluctuations on forecast scenarios that could be drawn are further amplified by a lack of both consistency in methodologies and comprehensive data, even in historic estimates of
environmental footprint from the ICT sector" and propose both software and hardware potential solutions to alleviate the data heat island effect.
The value is quite different.