Google Cloud hits $20B, capacity constrained

Cloud earnings used to be about software growth curves. Now they’re also about steel, power, cooling, and how many GPUs a company can physically install fast enough. That was the real message from Alphabet and Microsoft on April 29. Demand for AI cloud is still ripping higher. But both companies are telling investors the same thing in slightly different language — the limiting factor is no longer customer interest. It’s supply. ### What did Google actually say? Alphabet’s first-quarter results put Google Cloud above the $20 billion mark for the quarter, which is a milestone on its own. But the more interesting line was that growth was held back by capacity constraints. In plain English, customers wanted more AI infrastructure and services than Google could currently bring online. That matters because it means the business may have grown faster if enough compute had been available. (abc.xyz) ### What counts as “capacity” here? It mostly means three things. Datacenter space, electrical power, and AI accelerators — the chips and networking gear that run training and inference workloads. You can have customers ready to spend, but if the building isn’t energized, the racks aren’t installed, or the GPUs haven’t landed, revenue gets delayed. Cloud now looks a lot more like a logistics and construc(abc.xyz) to a few years ago. This is the hidden bottleneck behind a lot of AI optimism. (abc.xyz) ### Why is AI making this worse? Because AI workloads are unusually hungry. Traditional cloud demand scales with storage, databases, and ordinary compute. Generative AI adds giant bursts of demand for accelerator clusters, high-speed interconnects, and much denser power footprints. One big enterprise model deployment can soak up far more scarce hardware than a normal software migration. So even when clou(abc.xyz)n absorb the new supply almost immediately. That’s why “strong demand” and “constrained growth” can both be true at once. (abc.xyz) ### Where does Microsoft fit in? Microsoft’s quarter landed the same day and reinforced the same theme. Revenue reached $82.9 billion. Microsoft Cloud revenue was $54.5 billion. The company said its AI business passed a $37 billion annual revenue run rate, up 123% year over year. Intelligent Cloud revenue hit $34.7 billion. Those are huge numbers, and they show the demand side is not the problem. The pro(abc.xyz)online to serve it. (microsoft.com) ### Wait — was Azure up 40%? No. The current official number for the March 31, 2026 quarter is Azure and other cloud services growth of 35%, or 33% in constant currency. That’s still strong, but it’s lower than the preliminary context you gave me. The more notable figure this quarter was the scale of Microsoft’s AI revenue run rate, which shows how quickly monetization is catching up with the buildout. (microsoft.com) ### So is this good news or bad news? Mostly good, with a catch. Good, because constrained growth usually means demand is stronger than reported revenue makes it look. Bad, because fixing the constraint is expensive and slow. You don’t solve this with better sales execution. You solve it by securing land, power, transformers, networking gear, (microsoft.com)akes execution risk much more physical. (abc.xyz) ### What should readers take from this? The AI cloud race has entered an infrastructure phase. Product quality still matters. Sales still matter. But near-term winners are increasingly the companies that can procure and deploy capacity fastest. That shifts the debate from “Is there demand?” to “Who can turn capex into usable compute soonest?” ### Bottom line? The headline isn’t just that Google Cloud cr(abc.xyz)le and Microsoft are running into the same wall — not a lack of customers, but the brutal pace of building the physical backbone AI now requires.