Amazon Cloud's AI revenue surge

Amazon just gave Wall Street a cleaner number for the artificial intelligence boom: its cloud division says artificial intelligence services are now running at more than $15 billion a year in revenue as of the first quarter of 2026. That is the first time Amazon has put a direct dollar figure on the business instead of talking about “traction” and “demand.” (reuters.com) That number sits inside Amazon Web Services, the cloud arm that rented out computing power long before chatbots made it fashionable. Amazon Web Services generated $128.7 billion in sales in 2025, so a $15 billion artificial intelligence run rate means artificial intelligence is no longer a side experiment inside the company’s biggest profit engine. (aboutamazon.com) (reuters.com) A revenue run rate is not cash already booked for a full year. It is a speedometer: if the current quarter kept moving at the same pace for four quarters, the annual total would be above $15 billion. (reuters.com) Amazon’s chief executive Andy Jassy said a “substantial portion” of that artificial intelligence capacity is already backed by customer commitments. That is the detail that makes this look less like companies testing small pilots and more like companies signing long contracts for real workloads. (reuters.com) The bottleneck is not only software. Jassy said the tech industry is still hitting capacity constraints, which is a polite way of saying there are not enough data centers, power connections, networking parts, and advanced chips to satisfy every order arriving at once. (reuters.com) One part in particular keeps showing up: high-bandwidth memory, which is the stacked memory attached to artificial intelligence chips so models can pull data fast enough to keep the processors busy. Amazon’s comments tied heavy high-bandwidth memory orders directly to rising demand for artificial intelligence services, showing how a cloud contract can ripple backward into the semiconductor supply chain. (english.aawsat.com) (aws.amazon.com) Amazon is also trying to capture more of that hardware spending itself. Jassy said Amazon’s chips business, which includes Graviton processors and Trainium artificial intelligence chips, is now running at more than $20 billion in annual revenue, up from the $10 billion milestone the company cited earlier in 2026. (reuters.com) Trainium matters here because it is Amazon’s in-house alternative to buying every artificial intelligence accelerator from Nvidia. Amazon says a Trainium instance can include 512 gigabytes of high-bandwidth memory and 9.8 terabytes per second of memory bandwidth, which is why memory suppliers now sit so close to the center of the cloud race. (aws.amazon.com) The spending behind this is huge. Amazon said in February it expected about $200 billion in capital expenditures in 2026, with most of that aimed at artificial intelligence infrastructure, and Jassy said much of the Amazon Web Services portion should be monetized over 2027 and 2028. (aboutamazon.com) (reuters.com) Amazon is not alone, but the scale is getting easier to compare. Microsoft said in January 2025 that its artificial intelligence business had passed a $13 billion annual revenue run rate, and Amazon is now saying its cloud artificial intelligence business is above $15 billion in early 2026. (microsoft.com) (reuters.com) The picture this leaves is simple: companies are no longer just paying for artificial intelligence models on a screen. They are buying cloud capacity in chunks large enough to move Amazon’s revenue, justify $200 billion of infrastructure spending, and tighten the market for the memory chips that feed the machines doing the work. (reuters.com) (aws.amazon.com)