Microsoft and OpenAI are reportedly working on a massive data center that will house AI-focused supercomputers with millions of GPUs. information The project could cost “in excess of $115 billion” and the supercomputer, currently referred to as “Stargate” within OpenAI, will be based in the United States, the report said.
Microsoft will pay for the data center, which could be “100 times more expensive” than some of today’s largest operating centers, the report said. Stargate will be the largest in a series of data center projects the companies plan to build over the next six years, with executives hoping to have it operational by 2028.
According to the report, OpenAI and Microsoft are building these supercomputers in stages, and Stargate will be a Phase 5 system. The Phase 4 system will be cheaper and could start as early as 2026. informationSources say he may be considering starting in Mount Pleasant, Wisconsin. The system could require power from several units Because the Stargate could require so much power (“at least several gigawatts”), Microsoft and OpenAI are considering alternative power sources such as nuclear power. We are considering.
Officials say this is in part because existing designs require “far more GPUs in a single rack than Microsoft has historically used to improve chip efficiency and performance.” He suggested that a data center of this scale would be difficult. That also means coming up with new ways to keep everything cool.
It also appears that the companies may use this design phase to move away from dependence on Nvidia. The report claims that OpenAI wants to avoid using Nvidia’s InfiniBand cables in Stargate, even though Microsoft is using Nvidia’s InfiniBand cables in its current projects. OpenAI claims that it would rather use an ethernet cable.
Many things haven’t been decided yet, prices and plans are all subject to change, and it’s unclear when details will be finalized. information The company also said it has not yet been determined where the computers will be located and whether they will be built in a single data center or “multiple data centers in close proximity.”
Reports earlier this year suggested that OpenAI CEO Sam Altman has ambitions to develop AI chips and is seeking to raise as much as $7 trillion to build foundries to produce them. Last year, Microsoft announced a 128-core Arm data center CPU and Maia 100 GPU dedicated to his AI projects. There are also reports that Microsoft is developing its own networking equipment for AI data centers. Nvidia’s GPUs are in high demand as AI becomes more popular, so it’s no wonder companies like Microsoft and OpenAI want to have other options.
“We are constantly planning for the next generation of infrastructure innovations needed to continue pushing the frontiers of AI capabilities,” said Frank Shaw, chief communications officer at Microsoft. information, However, he did not appear to comment directly on supercomputing plans.
Microsoft has poured billions of dollars into its partnership with OpenAI, much of it in the form of computing power to run its models. If things like Stargate come to fruition, we’ll likely see deeper partnerships as the investments get bigger and more complex.
