Microsoft, OpenAI plan $100 billion supercomputer called ‘Stargate’: Report

Microsoft and Open AI are set to build a $100 billion AI supercomputer called Stargate, according to a new report.

Satya Nadella and Sam Atman posing together. (Image credit: Sam Altman)

Microsoft and OpenAI are working together on building an artificial intelligence supercomputer called “Stargate” that could cost as much as $100 billion, according to a report by The Information on Friday.

The publication reported the tentative cost of $100 billion, citing a person who spoke to Sam Altman about it and a person who has viewed some initial cost estimates. It also seems likely that Microsoft would finance the project that could cost more 100 times more than some of the biggest data and would only come around by the year 2028.

The report says that Stargate will be the biggest in a series of supercomputers that Microsoft and OpenAI plan to build over the next six years. The supercomputer plan is currently in the middle of the third phase with Stargate being part of the fifth and final phase. A significant part of the cost for the upcoming phases will come from acquiring AI chips, according to the report.

There is still a bit of a bottleneck in the AI development process — the shortage of graphics processing units or GPUs, which are used to crunch data for AI models. The generative AI boom caused the demand for these chips to skyrocket. On top of that, one manufacturer, Nvidia, almost has a monopoly on the market because it produces some of the most capable chips for AI. And it is struggling to deliver orders with ever-increasing demand. And its chips are only getting more expensive.

Also Read | OpenAI unveils artificial intelligence technology that re-creates human voices

Nvidia CEO Jensen Huang, right after the company’s GTC 2024 event, confirmed that the company’s latest AI chip called Blackwell B200 will cost anywhere between $30,000 and $40,000 per unit. The company has spent over $10 billion on research and development of the chip that outperforms all others on the market by a huge margin.

But now, a host of technology companies including Intel, Qualcomm, Google Cloud, Arm and Samsung have come together to create the “United Acceleration Foundation,” an open standard accelerator programming model that aims to challenge Nvidia’s software and hardware dominance in AI.