Microsoft Abandons Data Center Projects, TD Cowen Says (bloomberg.com) 25
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.
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.
Well, they probably saw no ROI (Score:5, Insightful)
In particular, they will not have seen any real possibilities to improve LLMs at this time. The last update to ChatGPT did increase cost, but not performance. And FOMO seems to be reducing in the LLM field as well.
Yes, MS is crap on the tech side. But their business decisions are much better at least some times. I guess we are seeing the beginning of the end of the current AI hype. Good.
Re: (Score:3)
In particular, they will not have seen any real possibilities to improve LLMs at this time. The last update to ChatGPT did increase cost, but not performance. And FOMO seems to be reducing in the LLM field as well.
Yes, MS is crap on the tech side. But their business decisions are much better at least some times. I guess we are seeing the beginning of the end of the current AI hype. Good.
It's exteremely unlikely that Microsoft is giving up on AI research. After all, they are giving up leases on two specific future data centers, but are not stopping all data center growth. The article (mostly paywalled) mentions pulling back due to losing OpenAI cloud business. OpenAI is growing and will require even more data centers, just not as many from Microsoft. The article noted that Google and Meta are scooping up abandoned Microsoft capacity. Also in the last few days, it was reported that even
Re: (Score:2)
I'm not sure why there seems to be so many people wishing for AI to fail,
Because it's an obnoxious expenditure that consumes all sorts of investment that could go toward other endeavors.
Those other endeavors might even include more innovative approaches to AI itself (e.g. money thrown at the most marketable speaking people rather than the most able to advance the state of the art).
Because many uses of AI are obnoxious (people taking a quick concept that can be properly conveyed in a sentence and having an LLM stretch it to a page of slop because it seems more proper or impressiv
Re: Well, they probably saw no ROI (Score:2)
Because it's an obnoxious expenditure that consumes all sorts of investment that could go toward other endeavors.
You mean something that can engage a diverse group of people on a new media platform to interact synergistically with new and evolving technologies? /s
Re: (Score:2)
The story is not about research...
Re: (Score:2)
The story is not about research...
Agreed. And that's a major point. AI research, which is driving much of the current data center expansion continues forward.
As a sign of this, Microsoft [reuters.com] "added its plans to spend $80 billion on AI infrastructure this fiscal year are on track."
Re: (Score:2)
Re:Well, they probably saw no ROI (Score:5, Informative)
Running DeepSeek on my own hardware was enough for me to see that we probably don't actually need +$7,000,000,000,000 in investment as Altman suggested. Before DeepSeek I and many others took that statement at face value.
Re: (Score:2)
Read more carefully. The first sentence says the projects are for generating "2 gigawatts of electricity." This is their power plant projects.
What has happened is they now have an American regulatory environment that is favorable to giving them artificially cheap electricity, with no demands on carbon mitigation that could increase the cost, and which allows them to shift these negative externalities to people without the power to stop them.
So they no longer have to generate their own power. They can get sw
Re: (Score:2)
Add to that the recent roadmaps from NVDA on next, next, and next generation chips - capabilities and power requirements - and they may have a better idea of what they intend to buy in future expansions.
If the chips do more compute, with lower power and cooling requirements, then the power and cooling needs overall lessen. Ditch a couple of datacenter contracts that will no longer be needed in the future.
planetary destruction (Score:3, Funny)
"Yes, the planet got destroyed. But for a beautiful moment in time we thought AI would create a lot of value for shareholders."
Re: (Score:2)
Why is this modded "funny?" This is literally their plan.
It could be that they already know the planet is destroyed, however, and that the powers-that-be are keeping this from us. We can't come up with something fast enough, so we're throwing it to AI to solve. Then it's a moonshot.
But honestly, that's unlikely. They're probably doing exactly what you said. Trying to make money at the cost of the masses.
They also sense the AI (Score:1)
...bubble waves brewing.
hype over Copilot is dying (Score:2, Insightful)
Microsoft changing its keyboard layout will go down in history with Samsung's Bixby button.
Re: they already ruined the keyboard (Score:2)
I used a 101 keys AT keyboard fro the laye 1980s for the longest time. First with PS/2 adapter. Then PS/2 to USB. A big reason i held on so long was to avoid seeing the Windows logo on the keys. The keyboard eventually failed, and I no longer have it.
I got a very fancy Ometon with color LEDs and perhaps 120 keys. I haven't bothered to program the exrea keys yet. But I'm enjoying the massive size and weight of the keyboard.
The bubble is finally about to burst (Score:5, Interesting)
The bubble is finally about to burst in datacenter economics.
Another company even demolished the old AOL headquarters (three buildings and parking garages) to make room for more datacenters in Sterling, VA.
Say it ain't so! (Score:2)
Not restarting Three Mile Island? (Score:2)
Re: (Score:2)
Right. The "clean electricity" is "clean coal." Which doesn't exist.
Oh, I'm sure it's coming soon. AI will figure it out for us. /s
Ironic considering today's capacity-related outage (Score:2)
Microsoft's East US 1 region suffered a major outage today, when they had a capacity shortage of all AMD VM SKUs (which are generally Microsoft's default/recommended SKUs). This caused widespread service failures and degradations when services that use automatic scaling and provisioning all stopped scaling and provisioning. They then suffered from shortages of equivalent Intel VM SKUs as everybody tried to migrate their workloads to VM SKUs that weren't impacted.
What next for those buildings? (Score:2)
Above, gweihir wrote: "I guess we are seeing the beginning of the end of the current AI hype. Good."
Who knows? - no one can predict the future with certainty - maybe just wishful thinking.
But, that was my first thought as well on reading the article.
Then, kriston wrote similarly: "The bubble is finally about to burst in datacenter economics."
- and then added: "Another company even demolished the old AOL headquarters (three buildings and parking garages) to make room for more datacenters in Sterling, VA."
I
Re: (Score:1)
Mod parent useless.