Microsoft, OpenAI Plan $100 Billlion 'Stargate' AI Supercomputer (reuters.com) 41
According to The Information (paywalled), Microsoft and OpenAI are planning a $100 billion datacenter project that will include an artificial intelligence supercomputer called "Stargate." Reuters reports: The Information reported that Microsoft would likely be responsible for financing the project, which would be 100 times more costly than some of the biggest current data centers, citing people involved in private conversations about the proposal. OpenAI's next major AI upgrade is expected to land by early next year, the report said, adding that Microsoft executives are looking to launch Stargate as soon as 2028. The proposed U.S.-based supercomputer would be the biggest in a series of installations the companies are looking to build over the next six years, the report added.
The Information attributed the tentative cost of $100 billion to a person who spoke to OpenAI CEO Sam Altman about it and a person who has viewed some of Microsoft's initial cost estimates. It did not identify those sources. Altman and Microsoft employees have spread supercomputers across five phases, with Stargate as the fifth phase. Microsoft is working on a smaller, fourth-phase supercomputer for OpenAI that it aims to launch around 2026, according to the report. Microsoft and OpenAI are in the middle of the third phase of the five-phase plan, with much of the cost of the next two phases involving procuring the AI chips that are needed, the report said. The proposed efforts could cost in excess of $115 billion, more than three times what Microsoft spent last year on capital expenditures for servers, buildings and other equipment, the report stated.
The Information attributed the tentative cost of $100 billion to a person who spoke to OpenAI CEO Sam Altman about it and a person who has viewed some of Microsoft's initial cost estimates. It did not identify those sources. Altman and Microsoft employees have spread supercomputers across five phases, with Stargate as the fifth phase. Microsoft is working on a smaller, fourth-phase supercomputer for OpenAI that it aims to launch around 2026, according to the report. Microsoft and OpenAI are in the middle of the third phase of the five-phase plan, with much of the cost of the next two phases involving procuring the AI chips that are needed, the report said. The proposed efforts could cost in excess of $115 billion, more than three times what Microsoft spent last year on capital expenditures for servers, buildings and other equipment, the report stated.
Can I come, too? (Score:1)
Re: Can I come, too? (Score:2)
Re: (Score:3)
Jaffa, kree!
Power? (Score:3, Interesting)
All this AI stuff is great and all, but we don't have nearly enough power to run a fraction of what people are proposing with AI, let alone the climate issues.
They need to start building nuclear power plants now to power all of this; renewables can't be as reliable as needed to operate something like this 24/7 so oil and gas is it unless we get some better power options.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
How are they going to power this?
A Naquadah generator, duh. Naquadria is too unstable.
Re: (Score:2)
We're probably reaching the point where a ZPM (charged) will be needed
Re: (Score:2)
Just be sure to check it for Goa'uld tampering first.
Re: (Score:2)
Just be sure to check it for Goa'uld tampering first.
Indeed. We don't want to blow up the whole solar system.
Re: (Score:2)
Yes: https://www.ans.org/news/article-5842/amazon-buys-nuclearpowered-data-center-from-talen/
MSFT not far behind AMZN.
Re: (Score:2)
Re: Power? (Score:2)
Translates to: We come in peace. Thousands, nay millions of light years to bring you this message: Girl power!
Re: (Score:2)
just need one to dial an other local (galaxy) Stargate to go long distance you need more.
Re:Power? (Score:5, Interesting)
Spoke to a Microsoft engineer on the GPT-6 training cluster project. He kvetched about the pain they're having provisioning infiniband-class links between GPUs in different regions.
Me: "why not just colocate the cluster in one region?"
Him: "Oh yeah we tried that first. We can't put more than 100K H100s in a single state without bringing down the power grid."
I know how this ends (Score:3)
The AI creates replicators and then they'll escape. Better stock up on your P90 [fnamerica.com] ammo now, because energy weapons aren't effective.
Re: (Score:2)
The AI creates replicators and then they'll escape. Better stock up on your P90 [fnamerica.com] ammo now, because energy weapons aren't effective.
You mean these guys [archive.org]?
Re: (Score:1)
Don't bet on it. We're well-beyond the point of diminishing returns. Scaling up by a factor of 10, 100, or even 1000 isn't going to noticeably improve performance. Neither is magical "emergence" going to fundamentally change the nature of the technology, no matter how badly you want that to be true.
If you're still worried about terminators or whatever, remember that these things aren't just expensive to train, they're expensive to operate. The robot revolution won't last 20 minutes. Oh, and 'energy weap
Re: (Score:2)
^ Found someone who hasn't watched SG1. ^
Re: (Score:2)
LOL, yeah, it's so tempting to call April fools.
Re: (Score:2)
Wow! We really need this! (Score:1)
https://shop.blackclinic.net/
How to keep up on AGI progress (Score:2, Informative)
Mathhew Berman https://www.youtube.com/@matth... [youtube.com]
TheAIGRID https://www.youtube.com/@TheAi... [youtube.com]
David Shapiro https://www.youtube.com/@DaveS... [youtube.com]
Re: (Score:3, Insightful)
Bullshit. There is not even a hint of AGI on the distant horizon. In fact, nobody sane knows whether it is even possible. Obviously the usual insane fanbois spout nonsense, as usual.
Re: (Score:2)
What are things that were said about new technologies not being viable such as the automobile that turned out to be world changing?
Here are a few notable examples:
Automobiles: When cars were first introduced, many believed they were impractical and unreliable. Skeptics questioned their safety, feasibility, and the need for such a mode of transportation. However, automobiles eventually transformed society, enabling faster travel, economic
Re: (Score:2)
Of course we should ask AI what it thinks about the argument?
That would require AI to be able to "think". It cannot. All it can do is regurgitate some arrangement of the thoughts of actual people. These actual people are often not very smart and easy to impress and manipulate. Hence the results of such a question would be meaningless. Science is not a democracy. You need to have actual facts.
It is really fascinating though how easy it is to convince so many people that _finally_ the AI revolution is here, when the AI field has pulled this scam so many times before. A
Re: (Score:2)
“People say, It’s just glorified autocomplete,” he told me, standing in his kitchen. “Now, let’s analyze that. Suppose you want to be really good at predicting the next word. If you want to be really good, you have to understand what’s being said. That’s the only way. So by training something to be really good at predicting the next word, you’re a
Re: (Score:2)
https://theconversation.com/ai... [theconversation.com]
Great gamble, but... (Score:2)
What happens if future tech does not require giant datacenters?
Placing huge bets, based on current tech, in a rapidly changing landscape is risky
Re: (Score:3)
Microsoft has made mountains of money on bad technology. They are just betting they can make even more money on even more bad technology.
Re: (Score:2)
This is really a multi-year plan and incremental ramping up. Even if full AGI is a ways off, AI is advancing and seeing more and more use cases, so this seems as much keeping up with demand as making any major bet.
Amazon also recently announced a $100B+ multi-year datacenter expansion plan (and are the compute provider for Anthropic).
The only aspect of this that seems a bit risky is perhaps the fact that it's somewhat tied to training costs of future "AI" (LLM) models. GPT-4 cost $100M+, and future generati
Who, me worry? (Score:1)
Or we could just call it SkyNet.
Re: (Score:2)
Maybe this is a Heuristically-programmed ALgorithmic system? Doing one better than anything IBM ever developed?
Alternate site ... (Score:3)
On MSN [msn.com]
Aren't supercomputers regulated as a munition? (Score:2)
What a waste of time (Score:2)
and money. Silicon Valley just keeps jumping sharks.