• 0 Posts
  • 563 Comments
Joined 1 year ago
Cake day: January 2nd, 2025




  • These are not AI data centers - two very different things.

    A traditional data center is largely a data storage facility with service-hosting/data processing capability.

    An AI data center is built with a focus on maximizing processing performance per sq ft.

    Nearly all traditional data centers probably can’t even get a significant increase in power to upgrade to significant AI capability.

    As I said elsewhere, there are roughly 3x as many McD’s as data centers, so for every 3 McD’s you pass means one DC.

    DC’s are everywhere - the basement of a large office building, in the central core of a single-floor large building, etc.

    I’ve worked for/with small companies that had DC’s rooms in their building, because they had sufficient power capability and cooling was trivial to reconfigure/upgrade.

    When people use online backup services (e.g. CrashPlan, Backblaze, etc), or running their own web server/VPS, it’s in these kinds of generic data centers (though I’m sure CrashPlan and Backblaze have some very large DCs, the concept is the same - it’s largely about storage capability which is a very different power and cooling profile than AI).


  • 10x fewer

    Most data centers have nothing at all to do with AI.

    Hell, most data centers would struggle to add AI to their environment - they likely lack the spare power and cooling to increase workload that much.

    Most data centers are data-storage facilities, with some processing workload capability (as much as is needed for the systems they host for their clients). But really, that’s largely just traditional database processing, etc.

    You’re probably within spitting distance of one.

    There are nearly 3x as many McDonalds locations in the US than data centers. So for every third McD’s you know of, there’s likely a data center nearby.