Analysts say investors are pressuring Microsoft over AI margins as AI run rate nears $37B

Microsoft is no longer trying to prove it has an AI business. It has one — and it is already huge. The new fight is over what kind of business this becomes. Investors liked the April 29 earnings beat, the 40% Azure growth, and the $37 billion annualized AI run rate. But they also saw a company spending at a pace that makes the old Microsoft margin story look less automatic. ### What changed this quarter? The simple version is that Microsoft finally gave the market a cleaner answer to the “show me the money” question. In fiscal Q3 2026, it said its AI business had passed a $37 billion annual revenue run rate, up 123% year over year. Azure growth reaccelerated to 40%. That matters because a lot of the skepticism around Microsoft was never about demand — it was about whether all the GPU and data-center spending would translate into visible revenue. ### Why didn’t that settle investors down? Because revenue and margins are not the same thing. Microsoft can show explosive AI demand and still leave investors uneasy if serving that demand is brutally expensive. The company has been telling investors that AI infrastructure and AI product usage are pressuring gross margin, even as efficiency gains help offset some of the damage. Basically, AI is working — but it is not yet the kind of effortlessly high-margin software story investors got used to with Office and classic cloud services. ### Why is capex the flashpoint? Capex is where the bill shows up. Microsoft signaled a very large 2026 spending plan to keep adding cloud and AI capacity, with Wall Street focusing on a roughly $190 billion figure for the calendar year. That is not routine maintenance money. That is build-the-next-layer-of-compute money. When spending jumps that hard, investors start asking whether the return will look like software — high margin, sticky, compounding — or like infrastructure, where growth is real but economics are heavier. ### Why does Azure matter so much here? Azure is the bridge between hype and proof. If Azure accelerates while AI workloads pile in, Microsoft can argue that AI is not just a flashy add-on but a driver of core cloud growth. That is what this quarter suggested. But Azure also hides the catch — some of that growth is expensive AI compute, and expensive compute can dilute margins even while headline revenue looks fantastic. ### What spooked investors like TCI? The sharper bear case is not “AI is fake.” It is almost the opposite. AI may be real enough to threaten the old software bundle. TCI’s move to slash most of its Microsoft position seems to reflect that fear. If AI assistants change how people create documents, search for information, write code, or use productivity tools, then Microsoft’s dominant software franchises could face a weird future where the company wins on AI adoption but gives up some of the pricing power and margin structure that made those franchises so attractive in the first place. ### Is this just a Microsoft problem? Not really. This is the core tension across big tech AI right now. The winners are posting real demand, but the cost to serve that demand is still enormous. Chips are expensive. Memory is expensive. Power is expensive. Data-center buildouts are expensive. So the market is starting to separate “AI growth” from “AI economics,” which is a more demanding test. ### Why does this change how people think about engineering? Because cost-aware engineering suddenly matters more. When compute is abundant, teams can brute-force a lot. When every extra inference carries a visible cost, efficiency becomes strategy. That means better model routing, smaller models where they work, tighter retrieval, smarter caching, and product design that does not waste tokens just because it can. ### Bottom line? Microsoft just proved AI can move the top line at scale. Now it has to prove that AI can preserve the kind of margins investors expect from Microsoft — and that is the harder part.