AI Data Centers Could Erase 15 Years of America's Climate Progress by 2030

Teever

I suspect that the heated debate and protests around wateruse from datacentres is intentional misdirection by big tech companies to shunt environmental thought/discussion to a managed topic instead of the one they don't want to talk about and can't avoid which is carbon emissions. It feels like almost over night we went from not doing enough but caring about carbon emissions and the impending calamity that they will cause to just... not, and and completely forgetting about them and what's about to happen. No real discussion about it in the media, no government policies, or lip service to the topic, just radio silence. The only real environmental discussions that rise up are about water usage near data centres, no bigger picture stuff. It's really disturbing how a handful of people can just decide that this is the direction of the global economy and resource allocation, there's no democratic discussion about this or input from experts on climate, economics, or domestic security and self-sufficiency. I don't really feel like we live in any close to a democracy in western countries anymore, it just seems like the locus of control is firmly outside of the average person, institutions, or even the moderately wealthy. It's just a few deranged people with more dollars than sense steering the ship into the rocky shore of a desert island.

bradknowles

So, there are solutions to pretty much all of these problems. The only issue is that they are expensive. You can get zero-waste recycling cooling systems, which take in almost no water after the initial fill. The main problem that you then have to deal with is the contaminants that get into the cooling loop and which have to be filtered out — that’s called “blowdown”. There are both liquid and solid blowdown solutions, but in the end that has to be trucked offsite to be dealt with separately. And in terms of water use, if you’re getting your power off-site, then whoever is generating that power is almost certainly using 5-10x as much water as your datacenter is, and that’s assuming you’re using the most water efficient method of traditional power generation — combined cycle natural gas turbines. Coal and nuclear are even worse. Check the HARC report for more on this. For power, sure it’s easiest to just hook up to the grid. But you can also do behind-the-meter solutions based on natural gas turbines, renewables like wind or solar, etc…. If you do renewables, then you also need to connect up grid-scale batteries to handle the power draw when the renewables aren’t generating. Then you have to deal with the fire risk. If you’ve got tons of lithium-ion batteries onsite, or tons of diesel fuel onsite for your diesel generators, then those could easily start a fire. And for lithium-ion batteries, about all the fire companies can do is sit there and wait for them to burn out. That’s when you wish you had done LiFePO4 instead. Once an equipment fire gets going, that can ignite nearby fuel sources, and if you’re close to the WUI at your site, that can get a wildfire started. That could potentially take out hundreds or thousands of nearby houses, schools, other businesses, etc…. Natural gas turbines rarely catch fire, so they’re not the same level of risk as diesel. And natural gas turbines don’t have large fuel tanks onsite, unlike diesel. But you have to be careful with natural gas and ensure that you properly winterize all your pipelines, otherwise you run into the risk of a complete power outage, and the kinds of problems we saw here in Texas during the “Great IcePocalypse”, a.k.a., “Winter Storm Uri”. For the exhaust of greenhouse gasses, those can be treated before they are released into the atmosphere, but treatment is expensive. If you make data centers cover all of their costs and not externalize any of those costs onto the people, then they quickly become unprofitable in many places. Which is why all the hyper-scalars building data centers do everything they can to hide everything they’re doing and to buy off all the politicians who could cause any problems for them. Case in point — Texas.