Microsoft has pulled back from leasing data centers across the U.S. and Europe, abandoning projects that were set to use 2 gigawatts of electricity over the past six months. This move, according to analysts from TD Cowen, is a strategic response to an oversupply of infrastructure relative to Microsoft’s revised demand forecasts.
A key factor behind the decision is the tech giant’s shift in AI infrastructure strategy, particularly regarding OpenAI’s model training needs. Microsoft has been a significant backer of OpenAI, but analysts led by Michael Elias noted that the company is reassessing its AI-related investments, choosing not to allocate additional resources for OpenAI’s expanding workloads.
The broader AI landscape has seen increased investor scrutiny over the high costs associated with AI model training and deployment. The emergence of Chinese startup DeepSeek, which introduced a cost-efficient AI technology earlier this year, has further pressured U.S. tech giants to reevaluate their AI spending. DeepSeek’s advancements demonstrated that high-performance AI capabilities could be developed at a lower cost, raising concerns over the sustainability of massive AI infrastructure investments.
Despite Microsoft’s pullback, other major players are stepping in to fill the gap. TD Cowen’s supply chain analysis indicates that Google has been securing capacity in international markets, while Meta Platforms has taken similar steps within the U.S. This shift highlights how demand for data center capacity remains strong, even as individual companies adjust their strategies.
Microsoft’s stock dropped by more than 1% on Wednesday following the news. However, the company reassured stakeholders that while it may strategically pace its infrastructure expansion in select areas, its overall growth remains robust across all regions.
This recalibration underscores the evolving nature of AI infrastructure investments. While the AI arms race continues, companies like Microsoft are demonstrating a more measured approach to scaling their cloud and data center operations. With AI development costs under increasing scrutiny, strategic adjustments in spending are likely to define the next phase of AI infrastructure growth.