Microsoft's drawback on datacenter investment may signal AI demand concerns

Investment bank claims software giant ditched 'at least' 5 land parcels due to potential 'oversupply'

Microsoft has reportedly cancelled leases on datacenter capacity in the US, raising questions about whether the company may have overestimated demand for AI services and the compute power it needs to drive them.

Financial services biz TD Cowen published a report claiming the Redmond megacorp was pulling back on datacenter spending as it may have found itself "in an oversupply position."

It says Microsoft has cancelled leases in America that add up to a couple of hundred MW of capacity with at least two datacenter operators, is not proceeding with signing other negotiated leases, and has reallocated a considerable portion of its international spend to the United States.

TD Cowen is a New York-based investment bank and financial services division of TD Securities, which operates as a broker-dealer and investment manager. It claims in its industry update that Microsoft is using facility/power delays as a justification for cancelling the leases.

This is the same tactic Facebook owner Meta previously used to cancel multiple bit barn leases in the US when it cut back on a $48 billion capex program related to the metaverse, the market watcher says, implying that Microsoft is also likely decreasing capex investment.

Microsoft has also pulled back on converting negotiated Statement of Qualification documents (sometimes referred to as a "500") into signed lease agreements, the investment bankers claim.

TD Cowen says it is unclear if this is simply a delay in 500-to-lease conversion or if it is an outright termination of the agreement. However, it notes that the conversion rate of "500 docs" into a signed lease is usually "close to 100 percent," with the datacenter providers using that as the signal to start construction.

At the start of this year, Microsoft said it was on track to pour approximately $80 billion into AI-enabled datacenters during its financial year 2025, to train and deploy AI models and cloud-based applications based on them around the world.

This was to meet expected demand for AI development and cloud services based on it. However, in the Windows-maker's most recent financial report, it talked of not being able to keep up with the clamor for AI, while there were signs of waning cloud demand.

So if the industry giant isn't growing cautious about the prospects for business interest in AI, why the cutbacks in bit barn capacity leasing?

According to TD Cowen, Microsoft recently walked away from at least five land parcels it had secured in multiple Tier 1 markets, and says this combined with the datacenter capacity it is giving up "in our view indicates the loss of a major demand signal that Microsoft was originally responding to". It believes this is because of OpenAI.

Microsoft was procuring capacity based on a forecast that included greater OpenAI workloads, but there is talk of late that the AI developer is switching to another compute provider. As a result "there is capacity that it has likely procured… where the company may have excess datacenter capacity relative to its new forecast," TD Cowen states.

So it may simply be the case that Microsoft had banked on needing datacenter capacity to serve OpenAI that is now surplus to requirement, or it may be that Redmond is concerned by the uncertainty that actions by the Trump administration are causing in the global marketplace.

A Microsoft spokesperson told us: "Thanks to the significant investments we have made up to this point, we are well positioned to meet our current and increasing customer demand. Last year alone, we added more capacity than any prior year in history.

"While we may strategically pace or adjust our infrastructure in some areas, we will continue to grow strongly in all regions. This allows us to invest and allocate resources to growth areas for our future. Our plan to spend over $80 billion on infrastructure this FY remains on track as we continue to grow at a record pace to meet customer demand."

Analysts are rather more skeptical about the level of end user demand and business value, with some saying investment in AI datacenters continues to heavily outweigh the amount enterprise are forking out on AI software licenses. ®