Microsoft’s AI ambitions fuel $3.3 billion bet on Wisconsin data center

Microsoft is betting big on AI, investing $3.3 billion in a new AI data center in Wisconsin as part of a growing wave of investment in the technology.

US President Joe Biden visited the site in Mount Pleasant, Racine County, on Wednesday to announce the news. The data center is set to come online by 2026. As part of the project, Microsoft said it is co-funding a new solar energy project that will generate 250MW of power.

“The announcement reflects an investment in AI’s broader potential to transform businesses and manufacturing,” said University of Pennsylvania engineering professor Benjamin C. Lee. “Such transformation needs more than just data centers. It needs people who are skilled in operating those data centers and people who are skilled in connecting AI capabilities to unique challenges and opportunities in existing businesses and communities. Much of this investment focuses on the people.”

Expanding cloud and AI infrastructure

Microsoft broke ground on the facility in September 2022 and said at the time that the project would cost $1 billion. However, the company now plans to invest $3.3 billion in the site to “expand its national cloud and AI infrastructure capacity.”

It has not released details of the hardware it is installing at the data center but said it will “help enable companies in Wisconsin and across the country to develop, deploy and use the world’s most advanced cloud services and AI applications to grow, modernize and improve their products and enterprises.”

A Microsoft spokesperson said the company could not comment further on the announcement.

Microsoft said in a news release that the project will create 2,300 construction jobs. The company is also partnering with Wisconsin’s Gateway Technical College to build a data center academy, which it said will “train and certify more than 1,000 students in five years to work in the new data center and IT sector jobs created in the area.”

To offset the site’s power consumption, Microsoft said it is working with National Grid to co-fund a 250MW solar energy project in Wisconsin. This part of the project is expected to be up and running by 2027.

“Any user or developer looking to build large-scale data centers needed to accommodate AI is scouring the country for extensive power infrastructure that can accommodate up to — and sometimes more than — a gigawatt of power,” said Andy Cvengros, managing director and US data center markets co-lead at JLL. Wisconsin has substantial power infrastructure due to previous investments made within the state.

The Mount Pleasant site was initially earmarked for a manufacturing plant operated by electronics giant Foxconn. However, a planned $10 billion investment, announced in 2017 and championed by former President Donald Trump, never fully materialized. Foxconn did open a data center on the site in 2021, but plans to manufacture LCD screens there were shelved, and now Microsoft is building on the land instead.

Unlocking Advanced AI Applications

Microsoft’s new investments will greatly improve AI applications, moving from systems that simply find and show existing information to ones that can create new content, said Andrés Diana, chief innovation officer for Accrete AI. He said that the additional capacity will enable more sophisticated cloud services, machine learning models, and real-time AI analytics.

“Specific technologies that could be developed include more advanced generative AI applications that could transform content creation, programming, design, and other creative fields,” he added. “Perhaps the most exciting area additional capacity will unlock is the unfettered use and insights that can stem from AI agents running 24/7 conducting research, synthesizing data, and producing predictive insights, recommendations, and work-product at a rate and quality that we cannot fathom today.”

Hyperscalers are excited about emerging capabilities, but there is significant uncertainty about their computational demands, Lee said.

“Data center capacity is growing to support training for larger models and datasets and to support serving these models as businesses and users discover new applications for these models,” he added. “Investments in capacity reflect the belief that more and more people and organizations will use larger and larger models.”

Skill gaps and access to economical labor are vital to scaling AI capabilities, said Jason Carolan, chief innovation officer at data center provider Flexential.

“With their proximity to Chicago, hubs like the greater Milwaukee area and Madison make this a logical choice,” he added. “This also leverages the history of Wisconsin and Illinois in innovation — such as Cray computing in Northern Wisconsin and Minnesota, and Mosaic at Urbana-Champaign.”

The state-of-the-art data center campus Microsoft is building will provide the computing power and high-speed connectivity required to develop and train cutting-edge AI systems, said Bill Long, chief product officer of Zayo, a global telecom infrastructure company.

“Microsoft’s new Co-Innovation Lab partnering with businesses will also enable companies to directly tap its AI expertise to design custom AI solutions to enhance their products and operations,” he added. “Other tech companies need to follow in Microsoft’s footsteps — these huge giants are already planning for the future, and other organizations need to as well to ride the AI demand wave.”

Adnan Masood, UST’s chief AI architect said hyperscalers are moving fast because analysts project the AI infrastructure market will take off in the coming years. “Companies that can provide the best AI services will dominate the cloud market and shape the course of technological progress,” he added. “It goes beyond raw capacity. Cloud providers are also racing to make their operations sustainable. Microsoft’s partnership with National Grid on solar energy and its focus on water conservation reflect this. Going green is a necessity in the age of climate change.”