Microsoft Corp. has reportedly cancelled some leases for data center capacity in the United States, according to analysis from Toronto Dominion Cowen (TD Cowen). This development has prompted broader speculation about whether the tech giant is overestimating its long-term requirements for artificial intelligence (AI) computing power.
TD Cowen’s findings suggest Microsoft, a significant player in the AI sector, has voided leases representing “a couple of hundred megawatts” of capacity – roughly equivalent to two data centers. These cancellations involve agreements with at least a couple of private operators, based on “channel checks” with supply chain providers, the U.S. brokerage wrote in a recent report. Additionally, TD Cowen indicated that Microsoft has reduced the conversion of statements of qualifications, which usually lead to formal leases.
A Microsoft spokesperson, in a statement, reiterated the company’s existing spending target for the fiscal year ending June but refrained from commenting on TD Cowen’s analysis. The exact reasons behind the lease pullbacks remain unclear. TD Cowen’s Monday report also suggested that OpenAI is shifting workloads from Microsoft to Oracle Corp., as part of a recently formed partnership. Furthermore, Microsoft is a substantial owner and operator of its own data centers, investing billions in its infrastructure. TD Cowen separately proposed that Microsoft may be reallocating some of its internal investment from foreign locations to the U.S.
“While we have yet to get the level of colour via our channel checks that we would like into why this is occurring, our initial reaction is that this can be tied to Microsoft potentially being in an oversupply position,” wrote TD Cowen analysts Michael Elias, Cooper Belanger, and Gregory Williams.
Following the report’s release, Microsoft shares saw minimal movement in pre-market trading.
The potential pullback in leasing decisions by Microsoft raises critical questions regarding the company’s outlook on overall demand, especially as one of the leading companies in the AI space. The company has outlined plans to spend US$80 billion during this fiscal year on AI data centers. During a late January earnings call, CEO Satya Nadella emphasized the importance of sustaining spending levels to manage “exponentially more demand.”
“While we may strategically pace or adjust our infrastructure in some areas, we will continue to grow strongly in all regions,” stated the Microsoft spokesperson. “Our plans to spend over US$80B on infrastructure this fiscal year remains on track as we continue to grow at a record pace to meet customer demand.”
European stocks related to the energy sector experienced a decline in response to the report, possibly indicating an anticipated reduced need for power to operate data centers among tech giants. Schneider Electric SE and Siemens Energy AG both saw their stock prices fall.
Critics have consistently expressed skepticism regarding the scarcity of practical, real-world applications for AI, despite the billions of dollars that Microsoft, Meta, and Amazon.com Inc. have pledged to invest in the necessary data centers. Recently, Wall Street analysts have begun to question the magnitude of these expenditures following the release of a new open-source AI model by DeepSeek, a Chinese company, which claims that its technology rivals U.S. technology at a significantly lower cost.
Microsoft executives have downplayed concerns around potential AI overcapacity. The company is committing more capital than ever before, with the majority of the spending allocated towards procuring chips and establishing data centers that support power-intensive AI services.
TD Cowen’s report highlighted that its channel checks uncovered several signs of Microsoft’s gradual retreat from certain agreements. The firm found that Microsoft had allowed over a gigawatt of agreements at larger sites to expire and had withdrawn from several deals involving approximately 100 megawatts each. TD Cowen noted that Microsoft used facility and power delays as justification for terminating the leases, a strategy previously employed by competitors like Meta Platforms Inc. to curb capital spending.
