![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
thewayneMicrosoft has cancelled or revised data center plans to the tune of $13,000,000,000 recently, projects that were mainly for AI centers. The reason? AI/LLM is not panning out as projected. As newer models are coming out, hallucination rates are rising rather than falling. This bodes ill.
In some cases lease options are being kept and the sites will continue being used as farmland until if/when MS decides to actually build the data centers.
Meta has recently likewise started cancelling data center plans.
Article may be paywalled:
https://www.bloomberg.com/news/articles/2025-03-26/microsoft-abandons-more-data-center-projects-td-cowen-says
The Slashdot summary:
"Microsoft has walked away from new data center projects in the US and Europe that would have amounted to a capacity of about 2 gigawatts of electricity, according to TD Cowen analysts, who attributed the pullback to an oversupply of the clusters of computers that power artificial intelligence. From a report:
The analysts, who rattled investors with a February note highlighting leases Microsoft had abandoned in the US, said the latest move also reflected the company's choice to forgo some new business from ChatGPT maker OpenAI, which it has backed with some $13 billion. Microsoft and the startup earlier this year said they had altered their multiyear agreement, letting OpenAI use cloud-computing services from other companies, provided Microsoft didn't want the business itself.
Microsoft's retrenchment in the last six months included lease cancellations and deferrals, the TD Cowen analysts said in their latest research note, dated Wednesday. Alphabet's Google had stepped in to grab some leases Microsoft abandoned in Europe, the analysts wrote, while Meta Platforms had scooped up some of the freed capacity in Europe."
https://slashdot.org/story/25/03/26/1832216/microsoft-abandons-data-center-projects-td-cowen-says
Meanwhile in China, two years ago a huge data center construction boom took place in an attempt to catch up in the AI/LLM race. And then the Chinese had a breakthrough and found a way around the GPU chip embargo and discovered that there wasn't nearly as much need for huge numbers of data centers and GPU farms.
And 80% of these data centers are sitting around unused!
From the article: “The growing pain China’s AI industry is going through is largely a result of inexperienced players—corporations and local governments—jumping on the hype train, building facilities that aren’t optimal for today’s need,” says Jimmy Goodrich, senior advisor for technology to the RAND Corporation.
The upshot is that projects are failing, energy is being wasted, and data centers have become “distressed assets” whose investors are keen to unload them at below-market rates. The situation may eventually prompt government intervention, he says: “The Chinese government is likely to step in, take over, and hand them off to more capable operators.”
Something on the order of over 500 were announced in 2023/2024, which means only 100 or so are in use?! The problem was that nobody knew what they were doing with AI, but by damn, we've got to get on that bandwagon!
https://www.technologyreview.com/2025/03/26/1113802/china-ai-data-centers-unused/
In some cases lease options are being kept and the sites will continue being used as farmland until if/when MS decides to actually build the data centers.
Meta has recently likewise started cancelling data center plans.
Article may be paywalled:
https://www.bloomberg.com/news/articles/2025-03-26/microsoft-abandons-more-data-center-projects-td-cowen-says
The Slashdot summary:
"Microsoft has walked away from new data center projects in the US and Europe that would have amounted to a capacity of about 2 gigawatts of electricity, according to TD Cowen analysts, who attributed the pullback to an oversupply of the clusters of computers that power artificial intelligence. From a report:
The analysts, who rattled investors with a February note highlighting leases Microsoft had abandoned in the US, said the latest move also reflected the company's choice to forgo some new business from ChatGPT maker OpenAI, which it has backed with some $13 billion. Microsoft and the startup earlier this year said they had altered their multiyear agreement, letting OpenAI use cloud-computing services from other companies, provided Microsoft didn't want the business itself.
Microsoft's retrenchment in the last six months included lease cancellations and deferrals, the TD Cowen analysts said in their latest research note, dated Wednesday. Alphabet's Google had stepped in to grab some leases Microsoft abandoned in Europe, the analysts wrote, while Meta Platforms had scooped up some of the freed capacity in Europe."
https://slashdot.org/story/25/03/26/1832216/microsoft-abandons-data-center-projects-td-cowen-says
Meanwhile in China, two years ago a huge data center construction boom took place in an attempt to catch up in the AI/LLM race. And then the Chinese had a breakthrough and found a way around the GPU chip embargo and discovered that there wasn't nearly as much need for huge numbers of data centers and GPU farms.
And 80% of these data centers are sitting around unused!
From the article: “The growing pain China’s AI industry is going through is largely a result of inexperienced players—corporations and local governments—jumping on the hype train, building facilities that aren’t optimal for today’s need,” says Jimmy Goodrich, senior advisor for technology to the RAND Corporation.
The upshot is that projects are failing, energy is being wasted, and data centers have become “distressed assets” whose investors are keen to unload them at below-market rates. The situation may eventually prompt government intervention, he says: “The Chinese government is likely to step in, take over, and hand them off to more capable operators.”
Something on the order of over 500 were announced in 2023/2024, which means only 100 or so are in use?! The problem was that nobody knew what they were doing with AI, but by damn, we've got to get on that bandwagon!
https://www.technologyreview.com/2025/03/26/1113802/china-ai-data-centers-unused/
Well ...
Date: 2025-04-26 04:41 am (UTC)Of course they are. One reason is Hapsburg AI -- that is, when AI content gets sucked into AI training and befouls the output of the subsequent AI. Since the AI moguls went to great effort to make it hard for people to avoid AI content, they shot themselves in the foot on that one. Now if they changed this to tag all AI content as such, somehow, then it could be excluded from training; but of course, the old untagged stuff is still everywhere.
Another aspect is not fixable. This is that humans are routinely illogical and inconsistent, and sometimes delusional. So the more sophisticated AI becomes, and the more human-created data it absorbs, the more likely it will suck up the crazy along with the rest of the content.
Really, any reader of science fiction could've told people about these issues. Our genre has been exploring the ups and downs of AI for ... what, about a century now?
Re: Well ...
Date: 2025-04-26 05:10 am (UTC)Yep. And as has been said, AI/LLMs are largely black boxes, and the people who create them really don't know how they work. I hadn't heard the term Hapsburg applied to AI, but it's an excellent usage.
no subject
Date: 2025-04-26 01:25 pm (UTC)In looking for work, I've been avoiding AI-centric employers. I'm all for knowledge-based reasoning but that fell out of fashion years ago.
My general impression of generative AI has been that it's easy to get quite good results from it but hard to control it with much reliability, which really limits its utility, it'll still sometimes tend to do unwanted or surprising things. It might actually not be bad as, say, a collaborator for writing fanfic (ignoring questions of IP) but that's not worth many data centers.
Between this and the blockchain moves from proof of work to proof of stake, I hope that GPUs become cheaper. (I assume NPUs must be some GPU variant that's a bit weak on graphics-specific primitives/units.)
no subject
Date: 2025-04-26 03:59 pm (UTC)I think focused AIs are pretty cool, like the ones searching for new antibiotics that aren't vulnerable to resistance problems. And I can see the attraction to talking to general knowledge ones as they can provide better information than just a general search engine. But they're just not worth the energy cost, and with them not being trustworthy, they're just going to dumb-down the population. Case in point: Mike Lindell, the My Pillow idiot, his attorney submitted a pleading of some sort in court this week that was partially/largely AI-generated and it was riddled with errors. The judge mercilessly hammered the atty, as well he should. This has happened many times, and apparently this atty either doesn't keep up with the news or thought he could pull a fast one.
no subject
Date: 2025-04-26 04:31 pm (UTC)There are some wins, like training for predicting deformation of non-rigid heterogeneous objects, where the alternative (e.g., 3d finite element) may be even more computationally expensive. Just not enough wins to be worth all the data centers!
no subject
Date: 2025-04-27 12:49 pm (UTC)Oh darn if it shuts down. :)
Hugs, Jon
no subject
Date: 2025-04-27 05:44 pm (UTC)I like arable land remaining arable!
no subject
Date: 2025-04-28 03:53 am (UTC)no subject
Date: 2025-04-28 03:49 pm (UTC)no subject
Date: 2025-04-29 04:26 pm (UTC)no subject
Date: 2025-04-29 06:05 pm (UTC)Unfortunately with LLMs (which have applications, but not as toys), crypto and block chain, I think it's a case of the genie is out of the bottle and will not be going back in, environment and energy needs be damned.