369
'Exactly how the dot-com bubble burst': A market research firm says keep an eye on this AI warning sign (www.businessinsider.com)
submitted 4 months ago by to c/technology@lemmy.world
89 comments fedilink hide all child comments
🀞 1
you are viewing a single comment's thread
view the rest of the comments
[–] 25 points 4 months ago (9 children)

But graphics cards, CPUs, RAM, and other components needed in data centers are a far cry from the same components used in home and office desktops and laptops. It’s like trying to sell parts used in F1 race cars on the consumer car market. Technically, there will be buyers, but the vast majority of these parts will remain unsold because demand for such specialized components is negligible in the general consumer market.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 18 points 4 months ago* (last edited 4 months ago) (3 children)

    I'm not sure if it's fully a bubble, or if the bubble is partly being used as a smoke screen to hide the upfront cost of redesigning computing infrastructure.

    A lot of the time, I think AI is just the branding layer. The real goal is top-to-bottom SaaS.

    Like, they're letting these AI companies hold the bag for building datacenters which will then get scooped by various companies like microsoft and google to offer virtualized home computing through a client.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 2 points 4 months ago

    There might.be a little of that going around as well, but the datacenters haven't actually been built even close to the scale claimed, and the racks in there are kinda useless at serving SaaS webpage stuff, too many GPUs to make it make sense financially.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 4 months ago

    The capacity will be freed.
    I am sure nvidia can repurpose the dies for consumer focused stuff and call them whatever or put them into business/workstation SKUs instead of enterprise/DC applications.
    Same for other products they sell.
    They arent as hyper specialized as an F1 (except maybe for those extreme 5U gpu compute nodes. They can hardly repurpose those lol

  • permalink
  • fedilink
  • source
  • parent