Microsoft takes over a Texas AI data center expansion after OpenAI backs away

Microsoft Takes Over Texas Data Center Project Abandoned by OpenAI, Becoming Neighbors in Massive AI Complex

Microsoft is stepping in to take over a data center construction project in Texas that OpenAI has decided not to pursue, positioning the two tech giants as neighbors in one of America’s largest artificial intelligence computing complexes.

Data center developer Crusoe announced Friday that it has partnered with Microsoft to construct two new “AI factory” buildings along with an on-site power plant in Abilene, Texas. The new facilities will be located adjacent to a larger computing campus that Crusoe has already been developing for OpenAI and Oracle.

The existing OpenAI project, known as Stargate, gained national attention when President Donald Trump officially announced it after his inauguration last year, framing it as a “resounding declaration of confidence in America’s potential” and highlighting the country’s growing investments in artificial intelligence infrastructure.

Although Microsoft maintains approximately a 27% stake in OpenAI and was once the ChatGPT maker’s exclusive cloud computing provider, the companies are increasingly charting separate paths in AI development, even as they now prepare to operate side-by-side on the same tract of Texas land.

Crusoe has already completed two buildings for OpenAI and Oracle, providing substantial computing power that supports the development and operation of technologies like ChatGPT. SoftBank has also participated as an investment partner in the project. The developer continues work on six additional buildings for OpenAI and Oracle, with completion expected by year’s end.

Earlier this month, OpenAI announced it was abandoning plans for further expansion of its Abilene facilities. “Our flagship Stargate site is one of the largest AI data center campuses in the United States,” explained Sachin Katti, OpenAI’s head of compute infrastructure, in a post on X. “We considered expanding it further, but ultimately chose to put that additional capacity in other locations.”

Katti noted that OpenAI is developing more than half a dozen sites across the United States, including a facility in Wisconsin being built in partnership with Oracle.

The addition of Microsoft’s two new facilities in Abilene will bring the total number of data center buildings at the complex to ten. Together, these facilities are expected to deliver an extraordinary 2.1 gigawatts of computing capacity. The site, once vast stretches of mesquite shrublands home to coyotes and roadrunners, was initially planned as a cryptocurrency mining operation before developers pivoted and expanded their designs following the AI boom sparked by ChatGPT.

Crusoe co-founder and CEO Chase Lochmiller highlighted the scale of the power infrastructure supporting the new Microsoft project. “A new power plant attached to the Microsoft project will be able to generate 900 megawatts to continue building the industrial foundation for American AI — at a velocity the industry has never seen,” he stated.

This new power plant will significantly exceed the capacity of the existing 350-megawatt, gas-fired power plant that supports the OpenAI and Oracle facilities. Oracle has previously described that on-site plant as a backup power source, with the data centers primarily drawing electricity from the regional grid, which includes power generated by nearby wind farms.

The rapid expansion of AI infrastructure has created tensions with tech companies’ environmental commitments to reduce greenhouse gas emissions. Most emissions come from burning fossil fuels, which drives climate change. During a visit to Abilene last year, OpenAI CEO Sam Altman acknowledged this challenge, stating, “We’re burning gas to run this data center,” while expressing hope that “in the long trajectory of Stargate” they would transition to more diverse power sources.

This development comes amid a period of intense competition in the artificial intelligence sector, with major tech companies racing to build the computing infrastructure needed to train and run increasingly powerful AI models that require enormous amounts of energy and computational resources.