Microsoft Reportedly Canceled Some Planned Datacenter Capacity

Microsoft still plans to spend at least $80 billion on AI infrastructure in its current fiscal year. But a new report suggests that a spending shift may be underway.

“Our recent channel checks indicate that Microsoft has terminated select leases with at least two private data center operators across multiple U.S. markets to the tune of ‘a couple of hundred [megawatts of electricity capacity]’,” TD Cowen analyst Michael Elias said. “Our initial reaction is that this is tied to Microsoft potentially being in an oversupply position.” (Bloomberg notes that “a couple of hundred megawatts” is roughly the equivalent of two datacenters.)

Separately, TD Cowen said that Microsoft may be reallocating some of its international spending budget to the U.S., while noting that the tactics seemed to mirror the steps Meta took when it lowered its metaverse-related investments.

Microsoft had no comment on the report, though it reiterated its previous pledge to spend over $80 billion on AI infrastructure during its current fiscal year, which ends June 30.

It’s easy to see this as the inevitable beginning of the end of Microsoft’s madcap spending splurge on AI. But it’s more likely related to the recent restructuring of its deal with OpenAI, which allows the latter firm to seek outside AI infrastructure for specific needs if Microsoft passes on the opportunity. In effect, Microsoft may have been over-committing on capacity to meet the expected needs of OpenAI. And it is now essentially offloading some of that capacity, and its expense, to others, including Oracle.

Whether this is happening in response to investor concerns or is simply a pragmatic internal decision is unclear, and almost beside the point. Microsoft, after all, will benefit from OpenAI’s work, and its financial upside if and when that happens, regardless of where it gets its datacenter capacity. Given the costs associated with this build-out, one might even argue that Microsoft will benefit even more if less of OpenAI’s workloads are clogging up Azure capacity.