Recent searches
Search options
Is data center AI scam bubble bursting?
"Microsoft denied everything. But TD Cowen kept investigating and found another two gigawatts of cancelled leases in the US and Europe. Bloomberg has now confirmed that Microsoft has halted new data centres in Indonesia, the UK, Australia and the US."
https://pivot-to-ai.com/2025/04/07/confirmed-microsoft-stops-new-data-centres-worldwide/
Turns out the AI race was to see who would wise up the fastest and stop throwing money into the incinerator first.
@lonespelunker @gerrymcgovern please let it be true!
@lonespelunker @gerrymcgovern Microsoft knew this at the highest levels, their “AGI” definition being profit was probably the most savvy thing I’ve seen out of Nadella.
They roped Sam with his own language and hyperbole, it’s really brilliant if you read the contract details because Microsoft basically only had to throw some money at it to seem incredibly relevant while giving OpenAI an “out” that could never be reached.
My POV on this is Meta and maybe Google are the only players WRT research
@lonespelunker @gerrymcgovern what’s particularly galling is Gates making a ton of AI predictions without regard to the underlying technology, I think he either drank some kool aid or was outright lying because he’s not stupid and it’s been abundantly clear for years now that scale won’t solve the fundamental problems with LLMs etc.
Meta is taking another approach and publicly acknowledging how flawed current technology is while still supporting research for it and crucially future novel work.
@dotsie @lonespelunker @gerrymcgovern
Bill Gates famously made a ton of predictions about whatever the latest buzz word for "Stuff with Computers" was and never cared for any of it beyond the profits for all his life...
@KarlHeinzHasliP @lonespelunker @gerrymcgovern it’s a shame because I’d love to know what he actually thinks and the way he researches topics so deeply leads me to believe he knew better about this.
The internet he famously missed but he was distracted by a lot of other things, I’ve heard him speak on many topics and his knowledge is deep not surface level and he was a brilliant programmer.
My suspicion is he aligns with Yann LeCun but could never say it publicly. That may be my bias though.
@dotsie @lonespelunker @gerrymcgovern Well, idk, he tried to hard code the internet explorer into the OS and lost all that potential profit only thanks to the courts at the time... So in my read he did not miss the internet, but almost got away with monopolizing it under Windows for max profit.
@KarlHeinzHasliP really good point
@gerrymcgovern been calling this out in meetings with our MS reps almost weekly and they're still in complete denial about it. we'll see how it looks this week 
@gerrymcgovern Microsoft shall die
"uhh guys, WHAT IF we are blowing these billions of dollars on something that does not work, that makes everything it touches measurably worse, that people dont want and are telling us they dont want, and that, even if it did serve a magical function that has not even been concieved of yet would cost more climate to run for one year than will ever exist again?"
[finishes burning 400,000 year old rainforest for data centre space]
"guys ...are we the baddies?"
@gerrymcgovern straight up imbecile shit.
I know not everyone can think but microsoft has invented the opposite of thinking. these clowns are gnikniht.
@falcennial @gerrymcgovern maybe Microsoft will dedicate some of the billions saved to solving the utterly broken search and star system in Outlook instead? Or the Mail merge on Mac? Just that would improve my life a lot more than any LLM.
@falcennial @gerrymcgovern oh and if they could stop using their AI to help the genocide in Palestine that would be good too
@Nicovel0 @gerrymcgovern SECONDED. MOTION CARRIED. GET ON IT MICROSOFT YOU NINNIES
@falcennial @gerrymcgovern Microsoft seems to be reading the same signals as the rest of us who are paying (too) close attention... To me the uncertainty now is whether enough of a critical mass of human beings who have experienced the enshittification brought by AI rebel against its general embeddedness across the monopolized tech spaces we rely on and push their governments to constrain big tech. I'm not so sure which way that will go.
@falcennial @gerrymcgovern this is a thing I'm working on right now, as we speak.. playing with critical uncertainty - the issues we are least sure about how they will play out over time. Technologically gen-AI *should* reduce the hallucinations and mistakes and get more efficient (but not less resource intensive/climate damaging), so it's still possible they make problem solving better, but emerging as well is that profitable use cases are hard to find to justify the cost of them..
@falcennial @gerrymcgovern ...four extreme scenarios will emerge from this exercise & then we can generate possible implications in any scenario, and prepare for them.
@Meznor @falcennial @gerrymcgovern I found this thread fascinating. An open source browser reaching out semi-responsibly to its (very niche) tech audience and floating the idea of AI use, and getting almost universally panned in the responses.
Perhaps the position isn't surprising considering the audience, but I feel the degree of hostility is notable. Real Clippy energy to it.
@philbetts @falcennial @gerrymcgovern ooo thank you for flagging this!! A real example of the pushback signal very prevalent on mastodon (but may be a very particular, close-paying-attention audience too). I'm not confident it translates to non tech people yet..
@philbetts
That is some negative reaction to AI! Great to see! Let's hope the trend catches on.
@Meznor @falcennial @gerrymcgovern
The baseline is: they invested billions to train the models, invested more billions in the data centres and then are forced to give it for free because not that many people wants it.
Worse still, they probably did the maths and discovered that even with not so widespread adoption, the scrapers and lost traffic is killing the websites from which the data is collected in the first place.
@Meznor @falcennial @gerrymcgovern
And if they paid attention, they probably realised that a resource like Wikipedia, built over 25 years, probably took less person/hour and a tiny fraction of storage and energy than their bloated models.
It's not sustainable.
@Lily_and_frog @falcennial @gerrymcgovern I think this is ultimately right. They're running up against some basic limits to growth on the tech side.
The big bet is the openness of the internet and ability to steal/use data at no legal or economic costs because regulations/law cannot possibly respond to the onslaught. That and the fact that a handful of billion-dollar tech companies already have all our data and simply have to change their terms of service on us so we can't escape it.
@Lily_and_frog @falcennial @gerrymcgovern my uncertainty remains more on that side of the equation - whether people will demand government/policy intervention to regulate what big tech can do with our data and make some meaningful enforcement mechanisms to deter this bad/unethical behaviour. The data centre problem seems like an overinvestment, and they'll ultimately recover from that.
@Meznor @falcennial @gerrymcgovern
Something will have to give.
It will burst eventually, but hard to tell which way it will go...
@Lily_and_frog
That's true. However, for Big Tech the issue is not to be accurate or even genuinely helpful. They make lots of money from waste--planned obsolescence. They also make money from addiction and to keep us addicted they needed to keep us scrolling. They needs lots of content and an AI that's our 'friend', always with us, the answer to everything we could imagine. Something like Wikipedia is not in any way monetisable.
What they are referring to these days as "AI" ain't Artificial Intelligence - it's is fancy auto-complete.
Another angle comes immediately to mind; if the Big Pipes are big enough, replacing one big data center with half-a-dozen smaller ones would make sense.