452
In China, the AI bubble has already burst (www.reuters.com)
submitted 1 week ago by to c/technology@lemmy.world
143 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 62 points 1 week ago (11 children)

My ideal scenario:

  • The AI bubble lasts a little bit longer, they put data centers in space.

  • The bubble pops immediately after this, all the companies go broke and nobody maintains the satellites.

  • Free PC components rain from the skies, I sit out on the roof surrounded by buckets catching RAM and SSDs.

  • They're a bit scorched but I'll take what I can get.

  • permalink
  • fedilink
  • source
  • hideshow 11 child comments
    •
  • [–] 17 points 1 week ago* (6 children)
    • Data centers in space
    • Overheat as there is no way to disperse heat in a vacuum
    • AI companies go broke
    • Can't maintain the useless space centers
    • They crash and collide starting Kessler Syndrome
    • We don't know who struck first, us or them, but we know that it was us that scorched the sky
  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 5 points 1 week ago (4 children)

    Overheat as there is no way to disperse heat in a vacuum

    This is incorrect. You cannot do convection heat removal in a vacuum because there's no air. However, you can still do radiant heat dispersal.

    So you dump your heat into something that has a lot of surface area with which to radiate it away from your components.

    Space is either very hot or very cold, depending on whether you are in light or shadow. If one side is constantly exposed to light, then there must be radiators on the other side to radiate that heat back off into the cold.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 5 points 1 week ago

    Now calculate the size of the radiators a full scale data centre would need.

    About 2000 sqm per MW, according to Google. Given an average data center load of 100 MW, about 200k sqm. Add that again for the solar panels you'd need, and you're at a cool half a million sqm. Compare the ISS, at a measly 7000 sqm. The mind (and budget) boggles.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 1 week ago

    Exactly. Look at pictures of the ISS: it has radiators. It has to get rid of its heat, and it works. And those radiators aren't even as big as the solar panels. The radiators really aren't the problem.

    The problem is scale. The ISS has about enough power for a single server rack. If you want entire datacenters in space, you're going to need thousands of times the scale of the ISS.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 1 week ago

    I actually work in the space industry designing/building/testing satellites.

    The answer is big fuggin radiators and perpetual night time sun synchronous orbit while the other vaporware/grifters trying to shill orbital power beaming blast your solar arrays with the power of a thousand suns every now and then.

    There is math for it, and I have a fuck load of doubts, but people are funding this bullshit.

  • permalink
  • fedilink
  • source
  • parent