• Veltoss@lemmy.world
      link
      fedilink
      English
      arrow-up
      13
      ·
      11 months ago

      Yeah it’s a “this is cheaper and we’re greedy” problem but people will add this to their AI fearmongering and hating circlejerks.

      Apparently they’re looking into nuclear reactors for this which doesn’t waste as much water from what I understand.

      • lud@lemm.ee
        link
        fedilink
        English
        arrow-up
        2
        ·
        11 months ago

        I don’t understand how a nuclear reactor could cool a datacenter. Could you explain?

        • CADmonkey@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          11 months ago

          Could supply power for better A/C?

          I admit that doesn’t make sense, but then neither does it make sense to cool a data center with evaporative water cooling as if it were a hit-and-miss engine from the 1910’s, so I dunno.

          • lud@lemm.ee
            link
            fedilink
            English
            arrow-up
            2
            ·
            11 months ago

            Cooling with evaporative cooling does make some sense since it works. It’s absolutely not ideal though.

            It should totally be a closed water (or other fluid) loop and where possible build datacenters in cooler climates.

            • V H@lemmy.stad.social
              link
              fedilink
              English
              arrow-up
              1
              ·
              11 months ago

              Latency limits datacenter placement a lot, but for batch jobs like AI training it’s certainly an option.

  • AutoTL;DR@lemmings.worldB
    link
    fedilink
    English
    arrow-up
    6
    ·
    11 months ago

    This is the best summary I could come up with:


    Microsoft’s data centers in West Des Moines, Iowa guzzled massive amounts of water last year, the Associated Press reported earlier this month, to keep cool while training OpenAI’s ChatGPT-4, the Microsoft-backed company’s most advanced publicly available large language model.

    Critics point out a further inconvenient detail: this happened in the midst of a more than three-year drought, further taxing a stressed water system that’s been so dry this summer that nature lovers couldn’t even paddle canoes in local rivers.

    But what will the future hold, both in those regions and in relatively water-flush areas like Iowa, as a result of the sudden increased demand for AI combined with the growing impact of climate change?

    That’s all even more worrisome in the context of a 2021 study, spearheaded by Virginia Tech researchers, which found many American data centers depend on stressed water systems.

    Even states like lush Florida face a predicament, because the water footprint of data centers is so large that they have the potential to put pressure on local watersheds, Virginia Tech researcher Md Abu Bakar Siddik told Futurism.

    In a response to our questions, a Microsoft spokesperson said that the company is monitoring the environmental impact of data centers and aims to be "carbon negative, water positive and zero waste by 2030.”


    The original article contains 1,047 words, the summary contains 213 words. Saved 80%. I’m a bot and I’m open source!