Excerpt:

LaRocque said the decision is baffling, citing concerns over rising electricity demand, massive water consumption and air pollution linked to AI data centres.

“Vancouver is in the middle of a housing crisis and water shortage,” he said. “These centres will use more heat and water — it seems counterintuitive to me.”

  • wampus@lemmy.ca
    link
    fedilink
    English
    arrow-up
    3
    ·
    5 个月前

    We already have a bunch of data centers in Vancouver. The electricity demand, water consumption and air pollution are generally not valid concerns – the current set rarely, if ever, use their diesel generators. Most of those are things like Tier1 data centers, where anyone can go in and buy rackspace to mount servers – a bunch of local companies use them, if they want more ‘sovereign’ network stacks while still having the hefty physical security controls. Having “No” datacenters at all, means putting all your data into US Cloud service providers (at the moment at least, for the most part). I doubt “AI” datacenters are all that different than regular datacenters, unless someone knows different and can clue me in? I’m fairly sure the push to build more “Canadian company owned” datacenters is part of trying to build more ‘sovereign’ tech stacks in Canada, though it’s a bizarre step since DCs have existed for a long time in Canada – it’s the software that’s the bigger issue in general.

    The better arguments, I’d say, are that the Job creation estimates look misleading, and that the real estate cost in Vancouver makes it a silly location for multiple new projects of this sort. In terms of job creation, the current data centers in Vancouver do not have a ton of staff at all – they’re generally built like mini-fortresses, with K-Rated fences (enough to stop a semi-truck goin like 100km/h), biometric entry points, mantraps, tons of physical security features… but in terms of actual staff, there’s like maybe 10-30 per site. There are security guards that’re on site 24/7, though it’s usually a relatively small team supported by cameras etc, working out of one primary security office near the main entry point. There are electricians/sparkies, but they usually work at numerous sites since you don’t have constant demand for ‘new’ installations. There’re a couple manager sorts to meet potential clients and do general overhead junk, and a couple to provide ‘remote hands’ troubleshooting if customers need it. Pretending like you’re going to create 525 permanent jobs is a generous estimate – it’s also a piddly number for the investment of public money.

    Datacenters are also typically just space/power, with companies needing to install their own compute hardware. Access to those systems is virtual, with the specific physical location not meaning much. Assuming most major cities in BC have decent power availability, and decent telecommunications infrastructure, it makes more sense to build these things in other communities with lower land cost. Even just pushing the vancouver DC’s out towards the valley, or up the coast a ways, would make more sense.

    • brianpeiris@lemmy.caOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      edit-2
      5 个月前

      With high-density computing, like the data centers that run artificial intelligence, comes immense heat that cannot be cooled with a conventional air-cooling system.

      The typical cabinet loads have doubled and tripled with the deployment of AI. An air-cooling system simply cannot capture the heat generated by the high KW/cabinet loads generated by AI cabinet clusters.

      Water cooling can be done in a smaller space with less power, but it requires enormous amount of water. A recent study determined that a single hyper-scaled facility would need 1.5 million liters of water per day to provide cooling and humidification.

      AI is typically deployed in 20-30 cabinet clusters at or above 40 KW per cabinet. This represents a fourfold increase in KW/cabinet with the deployment of AI. The difference is staggering.

      A typical Chat-GPT query uses about 10 times more energy than a Google search – and that’s just for a basic generative AI function. More advanced queries require substantially more power

      https://www.techradar.com/pro/security/i-sat-down-with-two-cooling-experts-to-find-out-what-ais-biggest-problem-is-in-the-data-center

      • wampus@lemmy.ca
        link
        fedilink
        English
        arrow-up
        1
        ·
        5 个月前

        Hm, interesting – though them requiring specific server setups to take advantage of the water-based cooling systems in the datacenter makes me pause a bit. I’ve seen what I imagine are those sorts of rigs at the DC I use for work, typically in segregated areas of the datacenter. Building a DC with that capability though, doesn’t necessarily equate to having customers to fill the racks.