Microsoft reportedly pulls back on global data center expansion

Sources familiar with the matter shared the details with Bloomberg. While Microsoft acknowledges making changes to its data center plans, the company hasn't provided specifics about individual projects.

The pullback could stem from several factors: lower-than-expected AI service adoption, power supply constraints, or graphics card availability issues. However, Microsoft CEO Satya Nadella recently suggested a different explanation - he expects an oversupply of computing capacity in the coming years due to the ongoing race for energy and computing resources.

His comments point to a potential cooling of AI enthusiasm. Specifically, Nadella stated: "I am thrilled that I'm going to be leasing a lot of capacity in '27, '28 because I look at the builds, and I'm saying, 'This is fantastic.' The only thing that's going to happen with all the compute builds is the prices are going to come down."

Recent technological developments might also factor into Microsoft's decision. In January, Chinese startup Deepseek demonstrated that competitive AI services can run on fewer resources than previously thought, and moreover, Deepseek published its model as open source. This aligns with Nadella's recent assessment that AI models are becoming commodities.

Microsoft still plans to invest approximately $80 billion in data centers during the current fiscal year ending in June, but the company signaled slower investment for the following year.

OpenAI continues compute expansion while Microsoft shows caution

Microsoft has yet to demonstrate significant revenue from its AI products like Copilot, beyond its growing cloud business. While OpenAI generates billions in revenue, it continues to operate at a loss.

The cooling relationship between Microsoft and OpenAI suggests Microsoft may be taking a more measured approach to AI investment. Meanwhile, OpenAI has sought partnerships with Oracle and Softbank to expand its computing capabilities, with reports indicating the company is operating at full capacity.

The company also expects its new test-time compute reasoning models to create additional demand for computing resources. This expectation helps explain OpenAI's massive Stargate project - a planned $500 billion AI infrastructure investment that would serve OpenAI exclusively. According to reports, OpenAI aims to source approximately 75% of its computing power from these proprietary cloud computers by 2030.