135

Microsoft Looks to Nuclear to Fuel AI Plans (tech.co)

submitted 9 months ago by lntl@lemmy.ml to c/worldnews@lemmy.ml

48 comments fedilink hide all child comments

Data centers, the things that physically store and share applications and data, require an enormous amount of energy to run. These giant storage units, responsible for 1-1.5% of global electricity consumption, have traditionally relied on renewable sources like solar and wind but it seems as though renewable energy just won’t be able to keep up with the demand required moving forward.

you are viewing a single comment's thread
view the rest of the comments

[-] umbraroze@kbin.social 19 points 9 months ago

I'm, like, OK, nuclear power isn't necessarily a bad thing.
But power plants like that should probably serve wider municipal needs.

Building a private nuclear power plant just to power a data center? Well that's clearly stupid.
Building a private nuclear power plant just to power a data center focused on a niche application? Well you know how that goes.

Also, look up SL-1. Disturbingly few Americans I've talked to have heard about that. Generally a good argument about why not every single thing should be powered by a tiny dedicated nuclear reactor.

[-] Revan343@lemmy.ca 12 points 9 months ago

A failure in a poorly designed test reactor isn't a great argument against nuclear power

[-] Cethin@lemmy.zip 11 points 9 months ago

SL-1 can't happen in a modern reactor design. In fact, it probably could never have happened in any non-test bed reactor. The issue was cause by an actual human pulling the control rod physically, from what I can tell. That's just not a thing done with operating reactors I'm pretty certain. There are also a ton more precautions built in than they had, as well as better protocal and understanding of the mechanics.

[-] Badabinski@kbin.social 10 points 9 months ago* (last edited 9 months ago)

I'm not making this comment to disagree with your point, but the failure of the SL-1 reactor strikes me as an engineering and process failure more than anything else. The reactor was not designed in a safe fashion, probably because it was designed as a test bed for reactors that could be deployed via airplanes to the Arctic circle. The fact that an engineer was even able to fully remove a control rod, and the fact that removing that control rod lead to a fatal steam explosion make me think that they really tried too hard when they removed weight and volume from the reactor design.

In well designed safety-critical systems, human error should not be able to cause any form of bodily harm. I don't think it's a great idea for a private company to be running nuclear reactors on Earth to power something as trivial as a data center (investing in storage + local solar/wind/geothermal/hamster wheel velodrome seems like a more efficient use of resources for one thing), but I also don't think that SL-1 is the best example to cite here.

As an aside, my high school Physics teacher went on a long diatribe about how the three SL-1 casualties were the only humans ever killed as the direct result of nuclear fission in the context of a nuclear reactor. Looking back on it, I think she was splitting hairs a bit, but it is an interesting point to make.