Microsoft is reportedly ramping up its investment in data centers and now has more than 5GW of capacity at its disposal.
Details of the tech giant’s spending on digital infrastructure were revealed in leaked documents seen by Business Insider.
Microsoft has been at the forefront of the AI boom through its partnership with ChatGPT-maker OpenAI and has added AI-powered functions to many of its products. This in turn has required significant additional spending on server capacity.
Microsoft’s data center plans revealed
The leaked document reportedly shows that Microsoft has obtained more than 500MW of additional data center space since July 2023.
This means the company now has more than 5GW of IT capacity available, according to the paper, which is said to be part of a confidential slide deck prepared by the company’s cloud and innovation team earlier this year.
Microsoft is looking to accelerate the growth of its data center footprint, according to the document, and is planning to double the amount of new data center capacity it builds in the first half of the 2024 fiscal year, adding 1GW of server power over the next six months.
For the first half of 2025, its target will be 1.5GW of new data center capacity, the document seen by Business Insider says.
Recent developments announced publicly by Microsoft include a $2.9 billion investment in data center infrastructure in Japan and a $3.16 billion spending spree in the UK. The latter will include a new data center campus being planned for the site of the Eggborough power station in North Yorkshire.
Building GPU capacity
The leaked document also reportedly shows that Microsoft has been busy acquiring GPUs, the chips needed to train and run AI systems.
The company made a multi-billion investment in OpenAI last year, with much of this thought to have come via cloud credits that allow OpenAI to use compute power from Microsoft’s Azure cloud platform.
Business Insider reports that Microsoft claims to have delivered “record-level GPU capacity” in the second half of last year, more than doubling the number of installed chips, though it does not put a figure on this.
The company’s GPU footprint has grown by an additional 39 data centers, the report says, meaning Microsoft now has live “AI clusters” in 98 locations around the world.
As well as being a major customer of Nvidia, the company that makes the most popular and powerful AI GPUs on the market, Microsoft has also developed its own AI accelerator chip in-house. The Azure Maia 100, built on the Arm architecture, is used in the company’s cloud data centers.