Microsoft sells computing capacity to the world. This year, it is buying some. The company is spending a record $190 billion to build AI infrastructure, and it is still short enough on servers that it has started evaluating Amazon and Google—its two biggest cloud rivals—as suppliers.According to a Business Insider report, the company is now evaluating Amazon and Google as sources of additional cloud capacity, with one person familiar with the discussions putting it plainly: “We are shopping for capacity everywhere.” Amazon has already stepped in once, bailing Microsoft out following a series of GitHub outages this year. Microsoft also explored leasing infrastructure from Oracle but walked away over security and compliance concerns. For a company whose Azure pitch rests on owning the stack end to end, that is an uncomfortable place to be.
Microsoft is building data centres as fast as it can, and AI demand is still outrunning them
The shortage is not for want of money. Microsoft’s capital expenditure this year is largely going into data-centre capacity for AI workloads, and executives told Business Insider the company is still constrained. Demand for computing infrastructure has simply outpaced how fast Microsoft can build. Azure remains its fastest-growing strategic business, but internally it has turned into a balancing act over where finite GPUs actually go.Curiously, the crunch has not slowed the sales side. Microsoft is raising quotas for Azure salespeople, some by as much as 30% this year, according to people familiar with the change.
Copilot first, R&D second, Azure customers with whatever is left: Microsoft’s CFO has already said it out loud
Chief Financial Officer Amy Hood explained the pecking order on the company’s January earnings call. Microsoft solves first for rising usage in sales and the accelerating pace of M365 Copilot and GitHub Copilot, its first-party apps. Then it invests in long-term R&D and product innovation. Whatever is left over goes to Azure customers, whose demand keeps climbing regardless.The numbers behind that choice are small but telling. Hood said that if Microsoft had pushed the GPUs that came online in the first half of its fiscal year to Azure instead of its own AI products, Azure growth would have crossed 40% rather than landing at 39%. Microsoft reported $75 billion in Azure revenue for its 2025 fiscal year. Investors were not soothed. That earnings report set off a post-earnings decline of more than 10%, with the market questioning a softer Azure outlook against record AI spending.Executives who spoke to Business Insider said those tradeoffs have only sharpened since.
Inside Microsoft, the compute queue is causing arguments executives can’t settle
The rationing has produced friction that is now spilling out internally. “All of the supply is gone once you solve for frontier labs and our internal businesses like M365 and Microsoft AI,” one executive told Business Insider.Another framed the frustration as a question nobody has a clean answer to: “Why would Satya prioritize growing Adobe over growing M365?” The same person added, “I have no idea how we’re going to land that message with customers.”That is the harder problem. Microsoft can buy capacity from Amazon and Google if it has to. What it cannot easily buy is a story its salespeople can tell enterprise customers about why their workloads are behind OpenAI’s and behind Copilot’s in the queue.Investors get a fresh read on Wednesday, when Microsoft reports fourth-quarter results. Azure’s growth rate and any update on capacity constraints will matter more than usual, with the stock already down more than 24% from 12 months ago.