Sources close to the idea said that Microsoft would probably fund the project, which is projected to cost 100 times as much as some of the largest data centers now in operation.
According to a report published in The Information on Friday, Microsoft and OpenAI are developing plans for a data center project that may cost as much as $100 billion and include the debut of the artificial intelligence supercomputer “Stargate” in 2028.
Requests for a response from Reuters were not immediately answered by OpenAI.
The need for AI data centers—which can do more complex jobs than standard data centers—has skyrocketed because of the rapid adoption of generative AI technology.
According to sources participating in private discussions about the idea, The Information said that Microsoft will probably finance the project, which is anticipated to be 100 times more expensive than some of the largest current data centers.
According to the report, the corporations plan to develop a series of supercomputers over the following six years, the largest of which would be the U.S.-based model.
The Information cited two sources for the estimated $100 billion cost: one who discussed it with Altman and the other who saw some of Microsoft’s first cost projections. It failed to cite their sources.
The supercomputers have been divided into five phases by Altman and Microsoft, with Stargate representing the fifth phase. According to the report, Microsoft is developing a fourth-phase supercomputer that will be smaller and intended for release in 2026 for OpenAI.
The report stated that Microsoft and OpenAI are currently in the middle of the third phase of their five-phase plan, with the acquisition of the necessary AI chips accounting for a substantial amount of the costs for the next two stages.
AI chips are frequently priced highly. The CEO of Nvidia, a chip firm, Jensen Huang, stated to CNBC earlier in March that the pricing range for the newest “Blackwell” B200 artificial intelligence chip will be $30,000 to $40,000.
A pair of specially created CPU chips were also announced by Microsoft in November of last year. According to the article, the new project will be made to function with chips from several vendors.
In an email to Reuters, a Microsoft representative stated, “We are always planning for the next generation of infrastructure innovations needed to continue pushing the frontier of AI capability.” Regarding the report concerning the scheduled launch of the Stargate supercomputer, the spokesperson refrained from making any immediate comments.
According to the source, the plan’s costs might surpass $115 billion, which would be more than three times what Microsoft spent on capital expenditures on servers, buildings, and other equipment last year.
