Bloomberg recently reported that Microsoft has cancelled or deferred more data center projects, this time extending to Asia Pacific, one of them being a project on a campus in Jakarta. This followed Microsoft’s earlier withdrawal in March of up to 2GW of data center projects in Europe and US.
As had happened previously after each cancellation or deferment, speculations are rife about the reasons, with some even calling this the end of the data center boom. However, a more plausible reason could be just as simple as a readjustment of capacity due to changing demand. Which company has not at some point done that – when conditions change, companies rewrite their contracts or undertake a ‘dynamic demand alignment’, as some might call it.
“The AI infrastructure cycle is entering a new phase— one defined not by hype or speed, but by disciplined deployment and dynamic demand alignment,” said Obinna Isiadinso, Global Sector Lead for Data Center Investments at the International Finance Corporation (IFC).
According to Obinna, this isn’t about Microsoft walking away from AI infrastructure but more about building smart, instead of fast. In his article published in Global Data Center Hub newsletter today, the international investor listed several factors that he believed contributed to Microsoft’s ‘unusual’ moves.
- Uneven AI adoption
Enterprise adoption is progressing at different speeds and Microsoft is basically “aligning infrastructure capacity with clear signals of demand – a strategic pacing of investment to maximize efficiency and ROI.”
- OpenAI’s Cloud Freedom
The updated agreement between Microsoft and OpenAI allows the latter to run workloads on multiple cloud platforms— not just Microsoft’s Azure. “That’s a major shift in how cloud demand is forecasted,” he reasoned. As such, Microsoft is “evolving its infrastructure strategy to prioritize flexibility across clients, broader ecosystem integration and sustained partnerships across the AI stack.”
- AI Models Are Getting More Efficient
Obinna believes instead of relying on massive GPU clusters for every workload like previously, the current situation has evolved as follows:
- Smaller language models (SLMs) are outperforming expectations.
- Open-source and fine-tuned models are reducing compute intensity.
- Quantization and optimization techniques are making inference more efficient than ever.
“This shift in model architecture is leading to a rethinking of infrastructure footprints, timing, and location,” he opined. But Microsoft isn’t the only one making strategic adjustments – Google is acquiring vacated sites across Europe, Meta is absorbing U.S. capacity that became available, and developers and data center REITs are reassessing timelines, regions, and tenant strategies. This strategic reshuffle signals the start of a transition period to phase two evolution of the data centre industry which he termed the “Optimisation Era” compared to the “Early Growth Era”.
It’s worth noting that Microsoft has consistently stated it is still committed to spending US$ 80 billion this year on infrastructure. “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 plans to spend over US$ 80 billion infrastructure this FY remains on track as we continue to grow at a record pace to meet customer demand.”
There, the tech titan said it – it’s business as usual, recent moves were just them adapting to changing demands. This view is shared by some analysts who are of the view that a company this large with US$ 80 billion of annual spend “has the right to move in and out of data center leases, many of which were never officially signed.”
Obinna further believes that the data center industry is finally exiting the “build fast, scale everywhere” phase of AI infrastructure and entering a period defined by scaling smart through purposeful partnerships, strategic pacing and targeted innovation. He is of the view that not only should big players move fast like this, smaller players like investors, operators, and founders should also rethink their strategy and do some rebalancing to stay ahead.
As with any other industry, the early growth years are always marked by hype, high valuations, and an overestimation of capacity as had happened during the Internet bubble. But companies that can spot the changing or evolving trends or inflection point, and more importantly, act on them, would be miles ahead. Microsoft’s moves show they may be way ahead of the game.