- Bloomberg reports Microsoft won’t go ahead with multiple data center projects worldwide
- It’s the third report claiming Microsoft’s more cautious approach to its data center projects
- There are indications that we’re on the verge of seeing a global capacity glut when it comes to data centers
Global stock markets are in turmoil right now as President Donald Trump’s tariffs begin to come into effect – but that’s not the only problem facing the big tech firms.
Microsoft is now the focus of a third Bloomberg report detailing pullbacks on its AI data center plans, raising fresh questions about how committed the American tech giant remains to its current level of investment in AI infrastructure.
While Microsoft maintains it will spend around $80 billion on data centers in this fiscal year, it has acknowledged that growth will slow after that. Future focus will shift from new construction to outfitting existing infrastructure with servers and AI equipment.
Still on track to spend $80 billion
Bloomberg claims Microsoft has exited or paused data center projects in multiple regions.
In the UK, it stepped away from a potential lease near Cambridge that was marketed for hosting advanced Nvidia chips.
In Indonesia, parts of a data center campus an hour outside Jakarta have been put on hold.
In North Dakota, talks dragged on so long with a potential tenant that Microsoft lost exclusivity, and expansion has also slowed in Wisconsin, where the company had already spent $262 million on construction, including nearly $40 million on concrete.
Microsoft also recently pulled out of a plan to lease an additional $12 billion of data center capacity from CoreWeave. OpenAI stepped in to plug that gap, but since Microsoft is OpenAI’s biggest funder, it likely used Microsoft’s own money to do it.
Some of the caution may come from signs that AI workloads are becoming less compute-intensive than expected, thanks in no small part to the “DeepSeek effect,” where the Chinese AI startup demonstrated performance comparable to OpenAI with far fewer resources. This has led some investors to question whether the current rate of buildout can be justified.
Meanwhile, a separate report from MIT Technology Review shows the scale of overcapacity in China.
The government there fast-tracked more than 500 data center construction projects, with at least 150 completed by the end of 2024, but an estimated 80% of these new AI data center resources are unused.
Together, these developments suggest that big tech’s data center ambitions may be outpacing the current demand for AI services, at least for now, which is starting to prompt caution among investors and industry insiders.
