Microsoft plans, according to a Bloomberg report, to increase its global data center capacity from twelve to more than 38 gigawatts by 2032 - more than tripling it. About one third of the capacity is then to be reserved for AI chips, up from around two gigawatts today. The trigger is shortages that have already forced the company to turn away customers.
Capacity is set to more than triple by 2032
Microsoft currently has around twelve gigawatts of data center capacity worldwide. By 2032, that figure is expected to rise to more than 38 gigawatts - an increase of about 26 gigawatts in six years. The share reserved exclusively for AI chips is growing disproportionately: from around two gigawatts today to about one third of total capacity, or more than twelve gigawatts. One gigawatt is roughly equivalent to the output of a large nuclear power plant unit - Microsoft is thus planning to add as much AI computing power within a few years as its entire current global network provides today. The announcement is part of a series of similar moves across the industry: Anthropic has recently secured compute contracts worth up to $517 billion, mostly with third-party providers such as Google and Amazon. Microsoft, by contrast, relies more heavily on its own and long-term leased sites. The planned buildout is spread across six years, equivalent to an average of more than four gigawatts of additional capacity per year - a pace well above the company’s previous expansion rates.
Capacity shortages force a course correction
The expansion responds to acute shortages. According to an analysis by DCPulse, several major US regions such as Northern Virginia and parts of Texas have already had to reject or redirect new customer requests because space and servers ran short. Microsoft also restricted new Azure subscriptions in certain UK zones, officially citing supply-chain issues in expanding the UK South and UK West regions. In August 2026, an internal allocation error in the East US region additionally prevented customers from launching new virtual machines. The shortages are hitting a company whose Azure cloud unit is growing strongly at the same time - demand for high-performance GPU infrastructure has outpaced the buildout of new data center capacity since late 2023. In the short term, Microsoft is responding, among other things, with a new agreement with cloud provider Nebius for an additional data center in New Jersey. For corporate customers, the shortages mean longer waits for new cloud quotas and, in some cases, a switch to competing providers until Microsoft has expanded its own infrastructure.
Investment spending reaches record levels
The expansion comes at a steep price. Microsoft spent about $145 billion on capital investments in the past fiscal year, on top of $329 billion in long-term data center lease obligations. For the first fiscal quarter of 2027, the company expects another $50 billion, and for calendar year 2026 overall around $175 billion - figures that are not independently verified. To lower its annually reported costs, Microsoft plans to spread long-term data center leases over 25 years instead of the previous 15. That reduces the expenses booked in any single year without changing the actual investment total. Analysts read the accounting shift as a sign of how much industry-wide AI spending is now shaping tech companies’ financial metrics. Microsoft did not comment on the Bloomberg report when asked by Reuters; the figures come from people familiar with the plans and are therefore not officially confirmed.
Whether the expansion actually solves the underlying problem or merely keeps pace with growing demand remains an open question. Financial markets are meanwhile growing more skeptical about whether tech companies’ billion-dollar AI investments will pay off - in July 2026, concerns over the returns on such spending alone drove the Nasdaq down by about three percent. If demand for AI computing power keeps accelerating, even a capacity plan built for 2032 may hit its limits sooner than expected.
