155
The AI data centre boom built on a mountain of hidden debt — Big Tech is spending more than space-race levels on AI infrastructure, but hidden loans and commitments mask the true financial risks (www.thetimes.com)
submitted 3 months ago by to c/technology@lemmy.world
21 comments fedilink hide all child comments

all 21 comments

sorted by: hot top controversial new old
[–] 20 points 3 months ago

Hidden debt? As in on top of all the obvious debt?

  • permalink
  • fedilink
  • source
    •
    [–] 12 points 3 months ago (8 children)

    So what happens to all this hardware when this comes crashing down due to unsustainability? I mean they are already losing so much money if anyone fully uses their subscription. And if they raise prices, people will stop using it completely. I’ll pay $20 a month. I will not pay $1000 a month.

    I mean if there is a fire sale I could put a few of these NVIDIA DGX B200 in a lab 🥹

  • permalink
  • fedilink
  • source
  • hideshow 8 child comments
    •
  • [–] 7 points 3 months ago (4 children)

    The cost will come down in 3 generations a lot, but the 1to3 trillion spent now to then will be worthless.

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

    I just don't see the cost coming down. It appears the expenses go up fairly linearly with their revenue, and if they lose subs, they'll have to recoup that difference.

    There was an article about open AIs finances the other day, and they lost 20 billion just last year, which is up about 4x from the previous year. Yes their revenue also when up 4x, but that the dollar amount for the loses grew more. If they keep on a similar trajectory, even investors are going to step away.

    Prices have already gone up for at least Copilot and Gemini to token use pricing from borderline unlimited use. It's going to, if it hasn't already for many, reach a point that just hiring someone to do they AI work will be cheaper and more reliable.

    There will be some sectors that benefit, like IT security monitoring, but I think optimized local models will take that spot. It already can for most personal users, it's easy to setup, and new hardware is coming out with local LLMs in mind.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 3 points 3 months ago* (last edited 3 months ago) (2 children)

    There was just a subquadratic model (SubQ)released that performs about as good as the last round of frontier models, with a 12 million token window, but with 3 orders of magnitude less resource usage. Google also has their compressed models which greatly reduce processing. I think at some point the stuff we use frontier models for today will be running locally on our PC's and all this buildout will have been for naught or for people who have such good business cases that paying top dollar for a better model makes sense.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 5 points 3 months ago (1 child)

    My question is this model is it open source or is it open weight because there is a difference. Open source gives you access to the training data that was used for the model. It gives you the code used to make the model. It gives you checkpoints during training of the model and methodologies that were used to train the model, etc.

    Open weight models, however, don't give you access to the training data or snapshots during training. Mainly, they just give the methodology and what they weighted the data as.

    The only fully open source model I am aware of, personally, is OLMO from AI2.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 3 points 3 months ago*

    Sorry I am mixing up my news. SubQ is not actually open source. I was confusing with the open source GLM5.2 which is also near frontier level but not lightweight. I edited my original post to reflect that.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 3 months ago* (last edited 3 months ago)

    It won't crash. You have to understand the background here. This is critically important to win for the US since they believe it's the future of the world. Whoever wins this wins the world, basically.

    I dont think that's true, personally. But they believe it. So it's not allowed to crash, with the US government behind it. Too big to fail.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 9 points 3 months ago

    Not hidden. Sitting there openly. That is always true for bubbles.

  • permalink
  • fedilink
  • source
    •
    [–] 9 points 3 months ago

    Hope everybody's as jazzed as I am for Lazer tag to make a comeback 🔮

  • permalink
  • fedilink
  • source
    •
    [–] 7 points 3 months ago

    Whenever you see correlated debt and conditional loans like this it’s time to panic.

    The impact of these things are almost never modelled for failure correctly, and so their impact is much much higher than a regular default.

  • permalink
  • fedilink
  • source
    •
    [–] 7 points 3 months ago (4 children)

    Like the whole world economy's foundation is the belief in AGI.

    Happen or not, but if it does not happen, this whole thing would be very funny and painful in retrospective

  • permalink
  • fedilink
  • source
  • hideshow 4 child comments
    •
  • [–] 6 points 3 months ago (3 children)

    The moment AGI is achieved it’s going to be a glorious destruction of society. Companies will lay off people and replace them with AGI and completely ignore the fact that no one will have money to buy anything. What we’re seeing now is just a fractional microcosm of what will happen.

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

    The moment AGI is achieved it’s going to be a glorious destruction of society.

    Specifically, if ultra affordable AGI that also happens to be acceptably reliable.

    Which is certainly a pipe dream, today. It might or might not ever get any more possible.

    To be economically viable, AGI needs to cost less than the people it supposedly replaces. Sure, we can all think of a couple of over-priced Word document slingers.

    But ask someone who wanted to buy a conveyor belt or a barcode scanner for a factory in an emerging market...until the technology costs dramatically less than local seasoned professionals, the technology just doesn't get bought.

    Now, I do imagine there's a few specific use cases that billionaires are very interested in replacing humans - because humans as morally empty and worthless as Epstein and Thompson (rest in piss) may actually be hard to find.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 0 points 3 months ago

    Still it is very dangerous. You need to engineer war or something like that in order to guarantee the safety.

    It is highly unlikely that amazon drones will be repurposed to kill all humans or guard data centers.

  • permalink
  • fedilink
  • source
  • parent
    • [–] 6 points 3 months ago

    They’ll just convert the data centers into prisons.

  • permalink
  • fedilink
  • source
    •
    [–] 2 points 3 months ago

    Paywall :(

  • permalink
  • fedilink
  • source
    •
    [–] 2 points 3 months ago

    ...oh no

  • permalink
  • fedilink
  • source
    •