Microsoft has abandoned a planned 2GW expansion of its datacentre footprint across the US and Europe due to its softening relationship with OpenAI, an analyst note from TD Cowen has suggested.
The document states that Microsoft has cancelled and deferred datacentre lease agreements in both the US and Europe in the past month, and “walked away from +2GW” of capacity over the course of the past six months that it was in the process of leasing.
“In our view, the pullback on new capacity leasing by Microsoft was largely driven by the decision to not support incremental OpenAI training workloads,” said the TD Cowen analyst note, dated 26 March 2025. “We [also] continue to believe the lease cancellations and deferrals of capacity points to datacentre oversupply relative to the current demand forecast.
“As such, we believe the lease deferrals are intended to provide Microsoft with a medium-term runway of capacity in major markets to support cloud/inference workloads, with Microsoft cancelling leases for capacity that exceeds its updated medium-term capacity needs.”
The news comes several months after details emerged that Microsoft tweaked the terms of an exclusivity contract it had with OpenAI pertaining to its use of the Azure public cloud when deciding where to host new workloads.
The details of this reworked agreement were published in a Microsoft blog post, dated 21 January 2025, that confirmed OpenAI had recently made a “large Azure commitment” that included “changes to the exclusivity on new capacity, moving to a model where Microsoft has a right of first refusal (ROFR)”.
This means Microsoft gets first refusal on whether or not it wants to host OpenAI workloads, but OpenAI also reserves the right to build its own capacity with other partners if Microsoft cannot meet its needs. “To further support OpenAI, Microsoft has approved OpenAI’s ability to build additional capacity, primarily for research and training of models,” the blog post added.
Around the same time, in January 2025, OpenAI went public with The Stargate Project initiative, a $500bn effort to build out OpenAI’s infrastructure over the next four years, with Softbank, Oracle, MGX and OpenAI as equity funders for the project.
Microsoft was listed as a technology partner for The Stargate Project, but its lack of financial involvement came at a time of growing press speculation about – what TD Cowen described in its 24 January analyst note – the “fraying relationship” between the two companies.
“We believe the announcement of The Stargate Project with Oracle is a further proof point of a shift in the allocation of incremental OpenAI business between Microsoft and Oracle,” said the January 2025 analyst note from TD Cowen. “OpenAI leaders [have been] highlighting that Microsoft could not keep up with its datacentre needs as it was moving too slowly.”
And this is why Microsoft has been retreating from its leasing commitments in certain geographies, the company said in its March 2025 analyst note.
However, the capacity that Microsoft is no longer interested in is being snapped up by its competitors, continued TD Cowen.
“Our checks point to Google stepping in to backfill capacity that Microsoft walked away from in international markets … [and] Meta backfilling capacity in the US, as both of these hyperscalers are in the midst of a material ramp in datacentre demand,” it added.
Computer Weekly contacted Microsoft for comment, but no response had been received at the time of publication.
