Microsoft, OpenAI plan data center and ‘Stargate’ supercomputer


Microsoft and OpenAI are joining forces to build a $100 billion dollar data-center complex, media reports say. And inside that canter will sit an AI supercomputer named Stargate.

Microsoft is looking to have the center and its Stargate supercomputer operational by 2028, The Information reported, after speaking with two sources familiar with the matter on Friday.

One of those sources said to be connected directly to OpenAI CEO, Sam Altman, the other said to have seen Microsoft’s initial estimates on the center.

The Silicon Valley-based fintech publication said Microsoft would likely act as a financier for the project – one of six proposed supercomputer and data center installations on which the two firms are collaborating.

It was not revealed where the centers may be located.

The Microsoft-OpenAI joint venture would cost 100 times more than the largest data centers running today and contain "millions of GPUs," the sources told the outlet.

Additionally, the report also stated that OpenAI's is expecting to release its next major upgrade by early next year.

Future need for data and cloud storage

The report stated that Microsoft is also working on a smaller, fourth-phase supercomputer for OpenAI its planning to launch in 2026.

Microsoft and OpenAI are alleged to be in the third phase of the five-phase plan, with phases four and five focused solely on acquiring the AI chips needed for the projects.

"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.

The entire multi-faceted project is expected to cost Microsoft a cool $115 billion – more than three times what Microsoft spent last year on capital expenditures for servers, buildings, and other equipment, the report stated.

Still, there are more than 10,000 data centers are operating around the world today, with about half located in the US.

Washington, DC, is considered one of the most dense data center markets on record, with over 300 centers responsible for about one-third of the world’s online traffic, supporting companies such as Google, Amazon, Meta, and Microsoft.

Currently, Microsoft's largest data center is said to be the Columbia Data Center, which opened in Washington State back in 2007. Meantime, China Telecom's Inner Mongolia Information Park is said to be the largest in the world.

In January, Alphabet’s Google announced it would spend $1 million to help build a data-center about 15 miles outside of central London.

Microsoft data center aerial Netherlands
Aerial view of Microsoft data center construction site in 2019. Image by Aerovista Luchtfotografie | Shutterstck

Data center cost factors

The global data center market over the next decade is forecast to grow by more than 10% per year, with data center construction at 7.5% growth.

Data center design companies say the cost to build a data center can run anywhere from half a million to tens of millions of dollars depending on the building’s footprint, site, and purpose, and is often measured by the megawatts the facility will hold.

Major factors contributing to the price tag include the infrastructure and construction costs, number of servers and storage units, software licenses and applications needed, power consumption, network connectivity, and the HVAC cooling systems, said tech company Intel Granulate.

Intel granulate said the average mid-sized enterprise data centers costs can range from $2 million to $5 million, with a typical capacity of up to 100 racks (or around 1,400 servers).

The growth is expected to put a strain on already maxed-out electric grids.

Altman, at the World Economic Forum this January, said it was imperative to find new and innovative ways to support the power consumption required to sustain future artificial intelligence.

The CEO has pushed the ideas of nuclear fusion and solar power as two climate friendly alternatives to electric power.