Microsoft hits $37bn AI run rate

Microsoft’s latest quarter answered the biggest AI question in big tech — is this stuff actually turning into real revenue yet? For Microsoft, the answer looks like yes. On April 30, 2025, the company said its AI business had surpassed a $37 billion annual revenue run rate, up 123% year over year, while Azure kept growing fast enough to calm a market that has spent months worrying about overspending. ### What did Microsoft actually report? The headline numbers were strong across the board. Revenue for the March 31, 2025 quarter rose 13% to $70.1 billion, operating income rose 16% to $32.0 billion, and net income rose 18% to $25.8 billion. Microsoft framed the quarter around cloud and AI demand, not just software renewals or price increases. That matters because investors wanted proof that the AI buildout was producing sales now, not just promises later. (news.microsoft.com) ### Why is the $37 billion figure such a big deal? Because it is not a vague “AI is helping everything” claim. It is Microsoft saying the AI business, taken together, is already running at a pace above $37 billion a year. That does not mean Microsoft booked $37 billion of AI revenue in one quarter — it means the current quarterly pace annualizes to that level. The jump was 123% from a year earlier, which tells you the company is no longer talking about a side business. (microsoft.com) AI is becoming one of the core engines. ### What happened inside Azure? Azure and other cloud services revenue grew 33% year over year in the quarter. The more revealing detail is that 16 percentage points of that growth came from AI services. Basically, about half the Azure growth rate came from AI workloads. That is the cleanest read-through in the report: customers are not just experimenting with models, they are paying Microsoft to run them. (news.microsoft.com) ### So why are people still worried? Because AI revenue is arriving with a giant infrastructure bill attached. Microsoft Cloud gross margin fell to 69%, down 3 points year over year, and the company explicitly tied that pressure to scaling AI infrastructure. In plain English — GPUs, data centers, networking gear, and power are expensive, and Microsoft is buying a lot of all four. (microsoft.com) ### Is this just Azure, or broader than that? Broader. Microsoft Cloud revenue hit $42.4 billion in the quarter, up 20%. Microsoft 365 Commercial cloud kept growing, Dynamics kept growing, and the company has been threading AI features through Copilot, GitHub, security tools, and Azure infrastructure at the same time. The point is not that every AI product is equally mature. The point is that Microsoft has multiple ways to turn model demand into billable software and cloud usage. (microsoft.com) ### What does this mean for engineering teams? It changes the optimization target. For years, cloud architecture was mostly about elasticity and developer speed. Now it is also about GPU availability, inference costs, model routing, and whether an AI feature earns more than it costs to serve. When Microsoft says AI is growing this fast while margins are getting squeezed, that is the signal. Capacity planning is now a product question, not just an infrastructure question. (microsoft.com) ### Why does this matter beyond Microsoft? Because Microsoft is one of the clearest real-world tests of the AI boom. If a company with this customer base, this cloud footprint, and this product stack can convert AI demand into tens of billions of annualized revenue, the market stops asking whether enterprise AI is real. The argument shifts to a harder one — who can make the economics work best. (microsoft.com) ### Bottom line The new Microsoft story is not “AI might help someday.” It is “AI is already huge, but the cost of feeding it is huge too.” That is a much more concrete — and much more demanding — phase of the cycle. (news.microsoft.com)