Microsoft has paused or delayed several data centre projects around the world, reflecting a recalibration of how and where it builds the infrastructure to support artificial intelligence and cloud services, according to Bloomberg.
Development activity has been scaled back or put on hold in countries including Indonesia, the UK, and Australia, as well as in US states Illinois, North Dakota, and Wisconsin. People familiar with the matter said the changes involve stalled negotiations or delays to construction at many sites.
Microsoft’s relationship with OpenAI has positioned it as a major host for AI services, and investors monitor the company’s infrastructure investments as a signal of future demand. However, it remains unclear whether the pullback is a response to weaker-than-expected demand, supply-side constraints, or a reallocation of resources. Some analysts believe it may indicate that projected demand for AI workloads no longer justifies earlier spending plans.
Concerns over data centre spending have had broader effects on tech markets in recent weeks, particularly for chipmakers. Nvidia, whose products are central to AI deployments, has seen a fall in stock value of 16.9% in the year to date, and Microsoft’s stock has dropped about 9% this year.
In response to inquiries, a Microsoft spokesperson said, “We plan our data centre capacity needs years in advance to ensure we have sufficient infrastructure in the right places. As AI demand continues to grow, and our data centre presence continues to expand, the changes we have made demonstrates the flexibility of our strategy.”
One site affected is located between London and Cambridge in the UK, where Microsoft reportedly withdrew from talks to lease a property promoted for its suitability for Nvidia-powered infrastructure. A similar pause took place at a proposed site near Chicago. In both cases, sources familiar with the negotiations declined to be named.
Elsewhere, Microsoft backed away from a deal to acquire additional capacity from cloud provider CoreWeave, according to the latter’s CEO Michael Intrator. He did not specify the number or location of impacted sites but said CoreWeave had since found another buyer for the capacity earmarked for Microsoft.
In Indonesia, construction has been paused at parts of a campus near Jakarta, while in Wisconsin, expansion efforts at a facility in Mount Pleasant have also been put on hold. The site was visited by President Joe Biden in 2023 and had already had US$262 million spent on it during early development, with roughly US$40 million allocated to concrete alone, according to Bloomberg.
Negotiations for a separate facility in North Dakota have also stalled. Applied Digital, which originally discussed leasing the site to Microsoft, said delays in negotiations caused the exclusivity window to lapse. The company has since secured funding from Macquarie Asset Management and is now in talks with other potential tenants. “Over the past year, we’ve learned that the hyperscaler contract process is extremely thorough,” said Applied Digital CEO Wes Cummins during a January earnings call.
In London in the UK, Microsoft was engaged in talks to lease space at Ada Infrastructure’s 210-megawatt Docklands data centre but has since held off, sources said. The site is now being shown to other interested companies. Ada’s parent firm, Ares Management, has declined to comment.
On the positive side, Microsoft said it remains committed to its US$3.3 billion data centre project in Wisconsin, which is expected to begin operations next year. Additionally, a spokesperson in Jakarta also confirmed that the Indonesia Central cloud region is still on track to go live in the second quarter of 2025, despite pauses in construction in some areas.
Microsoft projects capital spending of around US$80 billion for data centres in the current fiscal year, which ends in June. However, it has signalled that the pace of investment will slow in next year, shifting from new builds to equipping existing sites with servers and hardware.
Analyst scrutiny of AI infrastructure spending has intensified following a claim from Chinese company DeepSeek, which announced an AI model built with fewer resources than seen typically in the West. Some analysts say the development raises questions about whether future AI workloads will require less computing power.
In a recent note, analysts at TD Cowen estimated that Microsoft had pulled back on projects representing about two gigawatts of potential data centre capacity in the US and Europe. They attributed the move to a mismatch between current demand and available infrastructure, as well as Microsoft’s apparent decision to pass on some opportunities involving OpenAI.
OpenAI, currently valued at US$300 billion, is pursuing a large-scale AI infrastructure investment through a joint venture with SoftBank and Oracle. That effort, which could involve up to US$500 billion, may be prompting OpenAI to shift some of its compute needs away from Microsoft, the analysts suggested.
Industry sentiment about the pace of infrastructure growth has also been cautious. In March, Alibaba chairman Joe Tsai warned of a potential data centre bubble, noting that supply could exceed near-term AI demand.
CoreWeave’s Intrator described Microsoft’s approach as unique to its circumstances. “It’s pretty localised, and their relationship with OpenAI has just changed,” he said. “So it stands to reason that there would be some noise.”
Ed Socia, director at market intelligence firm datacentreHawk, said companies are revisiting their priorities amid rising costs and logistical delays. “You may have initially thought one data centre project would be the fastest speed to market, but then you realise that the labour, supply chain and power delivery wasn’t as quick as you thought,” he said. “Then you would have to shift in the short term to focus on other markets.”
(Image from Microsoft channel, YouTube)