> general environmental doom-mongering

I don't think this criticism is very well thought out. There's ample evidence at this point that we are sliding precipitously toward environmental catastrophe and, no, the massive data center build out to support people generating slop illustrations and build toy theorem provers is not helping.

▲

kragen 21 hours ago | parent [-]

We're not just sliding towards environmental catastrophe — we're deep into it! Ocean acidification is killing corals, we just had the hottest month in recorded history, mass refugee movements from climate disasters are fueling xenophobia and warfare in many parts of the world, and the rate of species extinction keeps increasing, as it has for decades. Extinct species are irreparable environmental damage; they're gone forever.

However, this is not because of massive data center buildouts. Goldman Sachs's projection in https://www.goldmansachs.com/insights/articles/us-data-cente... is that by the end of next year US data center power demand will have increased from 31GW to 66GW, roughly 17GW per year. The EIA says https://www.eia.gov/tools/faqs/faq.php?id=87&t=1 that the US uses about 95 quadrillion Btu per year (Americans will do anything to avoid using metric units). That's 3200 gigawatts. So the US's data centers are currently using 1% of the US's energy and adding ½% per year, so that in 02027 that number increases to 2%. 99% of the US's energy use is due to non-data-center causes.

And direct energy use is a way in which data centers have an especially large environmental impact. Their other potential impacts, such as cement production, sewage, parking lots, commuters driving to them, and freshwater consumption, are much less than that 1%.

Energy use, by the way, doesn't damage the environment. Burning fossil fuels does. All of China's net new energy production is coming from renewable energy — they're shutting down coal plants faster than they're building new ones, and they're decommissioning gasoline-powered vehicles faster than they're selling new ones. A friend of mine from the US told me that one of the most surprising things about visiting Shenzhen was how silent the downtown was, because almost all the cars are electric there — a much higher rate than for China overall, but even for the whole country, it's over 50% of new cars.

AI might turn out to be a wonderful thing or a terrible thing, but at least for the next decade or two, that won't be because of datacenter buildouts, which are an utterly insignificant fraction of its effect.

▲

voidhorse 20 hours ago | parent [-]

> So the US's data centers are currently using 1% of the US's energy and adding ½% per year, so that in 02027 that number increases to 2%.

1% of an entire nation's energy production for a single, non-essential use-case is a lot. For comparison, circa 2016, US farming, which we generally need to live, takes up 1.9% of energy production (https://www.ers.usda.gov/data-products/charts-of-note/87964). Hopefully this makes it more obvious that consuming 2% of all energy production, even though the raw magnitude is small, is a lot for a single, non-essential service currently controlled by a handful of companies.

> Energy use, by the way, doesn't damage the environment. Burning fossil fuels does.

I'm aware. The current reality we occupy in the US involves mostly burning fossil fuels to power these tools, and some plants deliberately skirt rulings around environmental safety. Management of resources and the economy is about trade-offs. Yes, in principle a certain activity may not depend on a bad trade, but if your only option right now is a bad tradeoff, it's "in theory" independence from this doesn't mean you should go ahead and do it.

> However, this is not because of massive data center buildouts.

It's because of a number of factors, obviously, but the data center buildouts do not help stem the bleeding, they (currently) exacerbate it. This idea of "well, it contributes to <bad thing> but less than <other bad things> therefore we should do it" is myopic and hopefully it's obvious to you why such a line of reasoning is not going to lead to any mitigation of any kind of crisis.

It's about trade-offs. If the overall ROI is worth the absurd costs, then, yes, it's all worth it. But "hey I (N=1) personally like this tool and get value out of it" does not justify global costs that contribute further to, as you said yourself, irreparable damage.

This isn't even touching on a host of other problems involved in the buildouts, namely making districts decidedly less attractive places to live and displacing current residents by allowing a handful of companies with massive amounts of power to directly disrupt the lives of citizens, for no direct recompense. (https://www.texastribune.org/2026/07/09/texas-data-centers-a...) (https://www.tpr.org/2026-09-17/residents-file-lawsuit-challe...)

> AI might turn out to be a wonderful thing or a terrible thing, but at least for the next decade or two, that won't be because of datacenter buildouts, which are an utterly insignificant fraction of its effect.

If the 2% energy consumption ends up being the final tipping point for total climate collapse, you might feel differently about this statement.

▲

jakzurr 17 hours ago | parent [-]

Sure, it's possible a certain 2% could be a tipping point, but seriously, look at US energy use as a whole, and it sounds unlikely.

https://en.wikipedia.org/wiki/Energy_in_the_United_States#Fi...

As far as sources go, reading further shows wind and solar still expanding. I don't think we can generalize anything from energy consumption of data centers.

	▲	voidhorse 7 hours ago \| parent [-]
		Yes, that kind of indeterminate conclusion I can accept. What I cannot accept are people like the parent who want to ignore the very possibility of an issue or the role of this buildout in contributing entirely or who want to claim others are telling "lies" when this is patently not the case. The amount of sophistry and absolute dogshit argumentation this forum invites now thanks to general boosterism is pathetic.