Microsoft reportedly plans to triple data-center capacity to 38+ gigawatts by 2032 -- while turning away cloud customers today

Microsoft is planning to more than triple its global data-center capacity, from about 12 gigawatts today to more than 38 gigawatts by 2032, according to Bloomberg News reporting citing people familiar with the plans. Microsoft has not confirmed the figure itself -- the roadmap comes from unnamed sources, not a company announcement or filing -- but the shortage it would address is already visible in how the company is rationing capacity today.

Only about 2 gigawatts of Microsoft's current 12-gigawatt footprint is AI-specific chip capacity today; Bloomberg's reporting says that share is expected to grow to roughly a third of the full 38-gigawatt target by 2032 -- a sixfold jump in AI-dedicated capacity on top of a threefold jump in total capacity. The gap between now and then is the story: Bloomberg reports the shortage has already forced Microsoft to restrict certain cloud subscriptions, turn away some clients, and absorb service disruptions, including an eight-hour GitHub outage in August tied to server constraints.

"I don't have warm shells to plug into." -- Microsoft CEO Satya Nadella, on the immediate shortage of ready data-center infrastructure

The near-term evidence for the squeeze is concrete, not speculative: e-commerce company Temu signed with Oracle instead of Microsoft after Microsoft couldn't supply the capacity it needed in its preferred regions, and Xbox has capped cloud game-streaming time for some subscribers. A plan to triple capacity by 2032 is a bet that this shortage is temporary -- but for the next several years, at minimum, Microsoft is telling its own customers no more often than it would like to.