Microsoft and OpenAI are planning to build an artificial intelligence (AI) supercomputer data center that is estimated to cost more than $100 billion and begin running in 2028.

The data center project, dubbed Stargate, will be headquartered in the United States to meet the growing demand for high performance computing (HPC) that supports generative AI workloads and applications – a tall order not necessarily suited for traditional data centers.

According to an initial media report, Microsoft will likely foot the bill for the project. Stargate is expected to be 100-times more expensive than today’s largest data centers.

Microsoft has a storied history of supporting OpenAI’s endeavors, particularly by supplying computing resources for OpenAI to run its models. The Stargate project represents another avenue by which the two companies are deepening the complexity of their partnership.

With five design phases for this series of supercomputers, Stargate will represent more than three-times the amount of money Microsoft spent on data center servers, buildings, infrastructure and equipment in 2023.

The two partners are in the third phase of their project, and Microsoft is working on the fourth phase – a smaller supercomputer for OpenAI that will launch in 2026.

It’s unclear where in the U.S. Stargate will be located and if the project will be housed in one data center or in multiple data centers within close physical proximity.

AI chips like GPU accelerators will account for a majority of the costs for the final two phases. Market leader Nvidia, with 97% of revenue share, announced last month its latest B200 AI chips will cost between $30,000 and $40,000.

The media report notes that OpenAI, however, hopes to avoid using Nvidia InfiniBand cables in the Stargate project in favor of Ethernet cables. Microsoft currently uses Infiniband technology in its data centers.

Global data center spending is projected to grow at an annual rate of 18% and reach $200 billion by 2028, according to research from Dell'Oro Group. AI workloads are projected to represent 25% of annual data center capex by 2028.

In 2023, data center providers overwhelmingly shifted their infrastructure investment priorities from general purpose computing and storage to (AI) and accelerated computing infrastructure, Dell’Oro Senior Research Director Baron Fung told SDxCentral

“The shift has been pronounced given the high cost of the equipment and GPUs relative to general purpose computing, and also the sudden shift in AI-related infrastructure spending,” Fung said.