Advanced Computing in the Age of AI | Monday, April 22, 2024

OpenAI and Microsoft Plan $100 Billion ‘Stargate’ Data Center in the U.S. 

(Gorodenkoff/Shutterstock)

The surge in demand for GenAI has created a pressing need for AI-powered data centers that have the capabilities to handle complex AI models and process large data volumes. Data centers must be able to scale seamlessly to accommodate increasing workloads. 

According to a report by The Information, Microsoft and OpenAI are discussing a project called “Stargate” that would see Microsoft spend as much as $100 billion to build a supercomputer data center to support OpenAI’s advanced AI models. The project is set to be based in the U.S. and could launch as soon as 2028. 

“Microsoft has demonstrated its ability to build pioneering AI infrastructure used to train and deploy the world’s leading AI models. We are always planning for the next generation of infrastructure innovations needed to continue pushing the frontier of AI capability,” Microsoft spokesperson Frank Shaw said in a statement to The Information.

Microsoft and OpenAI are reportedly planning to build several supercomputers throughout the U.S. in five phases. Stargate is part of the fifth and final phase of the project. Microsoft and OpenAI are currently believed to be in the middle of the third phase. A significant portion of the cost of the next two phases will go to acquiring AI chips. 

Since March 2023, Microsoft has been using Nvidia chips to power OpenAI. There have been reports of Microsoft introducing its own customized AI chips. The Azure Maia 100 and Cobalt 100 chips are the first two custom silicon chips designed by Microsoft for its cloud infrastructure. It would be interesting to see what chips are used for Stargate.  

The project is expected to be the largest in the series of supercomputing data centers the two tech companies hope to build. It is expected to be built on hundreds of acres of land and with over 5 gigawatts of power and millions of processors geared toward AI workloads. 

The $100 billion price tag dwarfs even the largest data centers around the globe. Just to provide some context to the staggering figure, the projected amount is triple what Microsoft spent last year in total on buildings, servers, and equipment. The project is expected to have extremely high power needs, and that may prompt Microsoft and OpenAI to explore alternative energy sources. 

OpenAI already uses Microsoft infrastructure to train its AI models. Both companies share a close partnership.  Microsoft’s support for OpenAI began in 2019 with a $1 billion investment. To date, Microsoft has invested over $13 billion in OpenAI. 

In 2020, Microsoft revealed it had built an Azure-hosted supercomputer to support OpenAI. More recently, Microsoft upgraded its infrastructure for OpenAI to include tens of thousands of A100 chips. 

It is worth noting that as the Stargate project is expected to be completed by 2028, there could be several changes along the way. The entire project could depend on OpenAI’s ability to deliver a significantly superior GPT-5 that requires advanced computing power. Is reported that OpenAI has already created a working prototype of GPT-5 that is “materially better” than the current GPT-4.  

Related Items 

Microsoft Solidifies Multi-year, Multi-billion Investment in OpenAI 

Inside Microsoft’s AI Announcements at Ignite 

Microsoft Seeks $10B Investment in OpenAI: Report

EnterpriseAI