AWS AI revenue surge

Amazon just put a hard number on a business every cloud company has been hinting at: Amazon Web Services says its artificial intelligence revenue is now running at more than $15 billion a year, the first time Amazon has disclosed a direct top-line figure for its artificial intelligence push. (aboutamazon.com) That number is a run rate, not booked annual sales. It means the first-quarter pace, if repeated for four quarters, would come out above $15 billion. (channelnewsasia.com) Amazon Web Services is Amazon’s cloud-rental business, where companies pay to use computing power in the same way a factory pays for electricity instead of building its own power plant. Artificial intelligence demand shows up there when customers rent chips and software to train models and answer user prompts. (aboutamazon.com) The expensive part of this boom is not the chatbot screen people see. The expensive part is the data center full of processors doing the math behind every image, answer, and search result. (reuters.com) Amazon is making a specific bet on that math getting cheaper if it designs more of the hardware itself. Chief executive Andy Jassy said Amazon’s custom silicon business, which includes Graviton, Trainium, and Nitro chips, is now running above $20 billion a year. (finance.yahoo.com) Trainium is Amazon’s in-house artificial intelligence chip, built to do model training and model serving inside Amazon Web Services instead of relying only on Nvidia processors. Amazon says the Trainium family is designed for “cost efficiency” across training and inference workloads. (aws.amazon.com) Inference is the part after a model is built. It is the live moment when a customer asks a question and the system has to produce an answer fast, millions of times a day, which is why cloud providers obsess over cost per response. (aboutamazon.com) Jassy’s argument is that cheaper inference changes customer behavior. If the cost of each answer falls, companies can afford to put artificial intelligence into search bars, call centers, coding tools, shopping apps, and internal software that employees use every day. (cnbc.com) Amazon is also spending at a scale that shows it thinks this demand is real, not experimental. Jassy wrote that Amazon plans to invest about $200 billion in capital expenditures in 2026, with a large share aimed at artificial intelligence infrastructure. (finance.yahoo.com) That spending comes after Amazon already said in February that Amazon Web Services had reached a $142 billion annualized revenue run rate in the fourth quarter of 2025. The new artificial intelligence figure suggests the artificial intelligence slice is becoming large enough to move the whole cloud business, not just decorate it. (ir.aboutamazon.com) Amazon is not saying it has won the cloud artificial intelligence race. It is saying the race has turned into a real revenue line, and that owning the chips as well as the cloud may let it keep more of each dollar customers spend on artificial intelligence. (aboutamazon.com)