Microsoft and OpenAI are planning a massive $100 billion AI supercomputer dubbed "Stargate"

[zohaibahd]

Why it matters: The AI arms race between tech titans is heating up, and Microsoft and OpenAI are gearing up for an absolute behemoth of a project. The partners plan to build a massive US-based data center codenamed "Stargate" that could have an 12-figure price tag.

The Information reports that Microsoft and OpenAI are in the early planning stages for an enormously ambitious new AI supercomputer installation. Unnamed sources state executives peg the budget at $100 billion for the US-based facility, which could house millions of cutting-edge AI chips.

The proposed Stargate data center would be the largest of five AI installations that Microsoft and OpenAI aim to develop over the next six years as part of their multi-year partnership. If the enormous project goes smoothly, Stargate should become operational by 2028.

While the plan remains fluid, Microsoft would likely finance the lion's share of the $100 billion project--a staggering sum around 100 times more than some of the most sprawling data centers today. The partners are evaluating potential chip suppliers, including Nvidia and AMD. Microsoft is even considering self-supplying Stargate's hardware needs. The facility's design may require innovations like denser server racks and newer cooling solutions to handle immense computational workloads and heat output.

Microsoft's partnership with OpenAI dates back to 2019. The companies have already jointly built multiple large AI supercomputer systems as they race to develop the best ChatGPT model and other generative AI technologies. Their current "Phase 3" installation is already operational.

In the short term, Microsoft already approved a $1 billion data center expansion in Mount Pleasant, Wisconsin, that could grow into a $10 billion "Phase 4" system expected to go live in 2026. OpenAI's next AI model upgrade is also reportedly slated for early 2025 release.

The potential $115 billion price tag for all of Microsoft and OpenAI's proposed future AI projects dwarfs Microsoft's current $50 billion annual spending on its global Azure data center infrastructure. However, the deepening ties between the two companies and the rapid escalation of their AI ambitions have drawn increasing regulatory scrutiny in the US, UK, and Europe over competition concerns.

Should it come to fruition, the unprecedented scale of the Stargate facility highlights how an arms race mentality is now driving firms to make economy-shaping investments in the field.

Permalink to story.

https://www.techspot.com/news/102452-microsoft-openai-massive-100b-targate-supercomputer-could-reshape.html

 

[Squid Surprise]

I suppose I may as well give the obligatory comment of, “when does it nuke Russia and bring out the Terminators”

 

[human7]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

 

[Squid Surprise]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

Well, the fiction is that AI will be able to do all that in the future…

 

[Axeia]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

TBF a lot of money is spend on weather prediction and for good reason it can greatly affects harvests (which can save a lot of money) and you can bet your *** on it that AI is used for better weather models as well.

But yeah money is spend where there's money to be made. Nothing new there

 

[brucek]

It'll be a real shame when they have to landfill that $100 billion system five years later because the Windows team decided arbitrarily to no longer support it.

 

[Thanthan]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

Well… the only guy I know who has a server rack with fully up to date computing hardware in his office is a meteorologist, and that’s on top of access to a supercomputer in Denmark for work that isn’t better processed locally (Greenland).

So well xD some money does get spent on predicting the weather xD

 

[Mr Majestyk]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

Well the Ai might just find those cures. It has already discovered two new antibiotics for testing against drug resistant strains of bacteria. We usually only hear the BS frivolous use cases, not the actual real benefits in science, engineering and medicine to name a few.

 

[wiyosaya]

I think this article should be under the article about the new image of the black hole at the center of the Milky Way due to the amount of energy it will consume.

 

[godrilla]

I think this article should be under the article about the new image of the black hole at the center of the Milky Way due to the amount of energy it will consume.

I believe Nvidia's GTC inferenced this research about ai helping in clean energy via fusion reaction.

https://engineering.princeton.edu/news/2024/02/21/engineers-use-ai-wrangle-fusion-power-grid



Data centers use about 10% of the world's energy consumption.

https://en.m.wikipedia.org/wiki/IT_energy_management

Maybe this is how they get away with it claiming we are getting closer and closer and closer to the research of clean energy while they use a significant amount of the world's energy in the process.

 

[Furriest]

"over competition concerns,"
Microsoft, OpenAI, and all of the corporate AI-bro chain is as anti-competitive as it gets in terms of hosting, licensing, and even running the scrapers, servers and telemetry; because nobody owns the full content they're scraping and are doing so inconsequentially due to the same 'be the first to the finish' or 'AI race' mindset previous innovation wars had.

 

[Theinsanegamer]

If only this kind of funding was being spent on supercomputers for weather prediction, finding cures for diseases, and other problems which could save lives. There's not as much money to be made there, so it goes to generative AI instead.

This reminds me of the people protesting the Apollo 11 mission because "some people are hungry".

That kind of money IS spent on computing for cures for diseases and weather prediction already. The computing resources being put together here could also be used for both those purposes.

 

[Squid Surprise]

This reminds me of the people protesting the Apollo 11 mission because "some people are hungry".

That kind of money IS spent on computing for cures for diseases and weather prediction already. The computing resources being put together here could also be used for both those purposes.

The difference was that it was a government trying to get to the moon - so you could argue it was in the national interest.

These are private companies - do we really believe they are interested in anyone but themselves?

 

[wiyosaya]

I believe Nvidia's GTC inferenced this research about ai helping in clean energy via fusion reaction.

https://engineering.princeton.edu/news/2024/02/21/engineers-use-ai-wrangle-fusion-power-grid



Data centers use about 10% of the world's energy consumption.

https://en.m.wikipedia.org/wiki/IT_energy_management

Maybe this is how they get away with it claiming we are getting closer and closer and closer to the research of clean energy while they use a significant amount of the world's energy in the process.

While the avoidance of this specific type of plasma seems to be a step forward in Fusion research, a commercial Fusion reactor is likely still something in the future, and we don't know when it will arrive. This makes me think that this development, WRT AI, is a marketing ploy.

The development reminds me of other AI developments where scientists are looking for the proverbial needle in a haystack. While AI may have reduced the size of the haystack, the AI results still require scientists to test each item in AI's list of possibilities to see if any of the possibilities pan out and they might not.

There is no guarantee that AI will give, or produce, meaningful results. If the results are not meaningful, perhaps they can be used to improve the model, but still, companies selling AI have every reason, especially in the name of profit, to try to market AI as a solution to problems that are not yet known to exist.

 

[wiyosaya]

This reminds me of the people protesting the Apollo 11 mission because "some people are hungry".

That kind of money IS spent on computing for cures for diseases and weather prediction already. The computing resources being put together here could also be used for both those purposes.

Its that old argument: Would the money be better used elsewhere?

Clearly, AI has shown in some circumstances, that it is useful. However, I think it is tainted by the fact that AI, no matter who is behind it, is competitive, and those in the field are trying to out compete each other to reap the profits before others get the chance.

 

[Zsoltblabla]

Well, the fiction is that AI will be able to do all that in the future…

Nope…. The only reason they invest in AI is to replace humans workers with machines. Whichever company achieves that on a higher level will be worth hundreds of trillions of dollars.
Simple as that.