Microsoft $190B AI capex

Microsoft’s quarter was strong. The market still flinched. That sounds contradictory, but it actually makes sense once you separate demand from delivery. Microsoft just showed that customers want a lot more AI, right now. It also showed that serving that demand will cost an absurd amount of money and take longer than investors hoped. (microsoft.com) ### What did Microsoft actually say? On April 29, Microsoft posted fiscal third-quarter revenue of $82.9 billion, up 18% year over year, with net income of $31.8 billion and diluted EPS of $4.27. Satya Nadella said the company’s AI business has now passed a $37 billion annual revenue run rate, up 123% from a year earlier. Microsoft Cloud reve(microsoft.com)lion. (microsoft.com) ### So why did investors focus on capex? Because the bill got much bigger. Microsoft said capital expenditures and finance leases were $31.9 billion in the quarter, and it now expects calendar 2026 capex to reach about $190 billion. CNBC said that is well above Wall Street expectations, and part of the jump comes from higher component costs —(microsoft.com)s revenue convert into returns when the infrastructure bill keeps rising? (cnbc.com) ### Why is the spending so high? Because AI demand is not just a software problem. It is a datacenter problem. Every new model, agent, Copilot seat, and inference-heavy workload needs GPUs, CPUs, networking, power, cooling, and memory. Microsoft said two-thirds of quarterly capex went to short-lived assets like GPUs and CPUs, which tells you this is not (cnbc.com)now, at scale, in a market where key parts are still pricey. (fool.com) ### What does “capacity constrained” really mean? Basically, Microsoft has more AI demand than ready-to-serve infrastructure. Yahoo’s coverage of the call said management expects to remain capacity-constrained through 2026 even with the higher spending plan. That means some product growth is gated less by code shipping and more (fool.com)sh — the feature can be done before the factory-sized machine that runs it is ready. (uk.finance.yahoo.com) ### Is the demand side at least real? Yes — and that is the part bulls will point to. Azure growth stayed strong, Microsoft’s AI revenue run rate crossed $37 billion, and Microsoft 365 Copilot paid seats rose past 20 million. The company is also leaning harder into usage-based and pe(uk.finance.yahoo.com) story. It is a “customers are arriving faster than infrastructure economics are settling down” story. (microsoft.com) ### Why did the stock drop anyway? Because public markets hate long payback periods when the spending curve steepens. Microsoft shares closed April 30 at $407.78, down 3.93% on the day on Microsoft’s own IR quote page, after falling as much as about 5% intraday in broader coverage. Investors were not rejecting AI demand. They were discounting(microsoft.com)ware and datacenters. (microsoft.com) ### What should engineers and product teams take from this? A useful lesson is that “code complete” and “capacity ready” are different milestones. In ordinary software, shipping mostly means the product works. In AI infrastructure, shipping can mean the model works, the UX works, and demand exists — but the shared compute pool is still the bottleneck. That changes road(microsoft.com)st engineering velocity. It is whether the platform owner can afford and install enough machines. (fool.com) ### Bottom line Microsoft just proved that AI is already a huge business. But it also proved that winning it may require spending at a scale that keeps even Microsoft looking supply-constrained. The opportunity is real. The catch is that hyperscale AI is now part software rollout, part industrial buildout. (news.microsoft.com)