• kent_eh@lemmy.ca
    link
    fedilink
    English
    arrow-up
    13
    ·
    22 days ago

    It’s going to take years (and truckloads of money) to build domestic capacity to rival what the US companies currently have. And then you will have to pay that money back with a much smaller total potential customer base, while competing against established international competitors with deeper pockets.

    It’s not impossible, and some players will absolutely be able to chip away at it, but it’s not going to be as easy or fast as people what it to be.

      • kent_eh@lemmy.ca
        link
        fedilink
        English
        arrow-up
        1
        ·
        21 days ago

        Sure, we can work with some other countries, but the reasons we want a “made in Canada” solution are the same reasons other countries want their own locally controlled data infrastructure.

    • AGM@lemmy.ca
      link
      fedilink
      arrow-up
      2
      ·
      22 days ago

      Canada’s small population to serve inference to is an advantage to some extent in such a capex heavy space, but the money being committed federally is pretty tiny. I look at $3.3 billion committed for AI over 5 years when compared with $60B for new subs, $3B for Arctic patrol ships, and up to $27B on F-35s if we go through with that order and don’t think we’re being serious about sovereign AI at all. We don’t have to be an OpenAI spending half a trillion on data centers for training and inference to reach a market of a billion+ people, but $3.3B over 5 years is not being serious about carving out our own independence from hyperscalers.