Amazon letter flags big AI spend and capex plan

Amazon’s recent shareholder letter highlights that AWS AI revenue exceeds a $15 billion annual run‑rate and that the company plans roughly $200 billion in 2026 capex focused largely on AI infrastructure. Those figures underline Amazon’s scale bets on AI compute, edge services and custom chips for cloud customers. (x.com)

Amazon just put a hard number on a business it had mostly described in broad strokes: Amazon Web Services said its artificial intelligence revenue run rate is now above $15 billion in the first quarter of 2026. Chief executive Andy Jassy put that figure in his annual shareholder letter on April 10, 2026. (aboutamazon.com) That number matters because a “run rate” is not a forecast from a slide deck. It means if the current quarter kept going at the same pace for a full year, the business would be bigger than many entire software companies. (aboutamazon.com) Amazon paired that disclosure with an even bigger one: it expects about $200 billion in capital spending in 2026. Jassy said most of that money will go to artificial intelligence infrastructure, which means data centers, networking gear, and the chips that fill those buildings. (aboutamazon.com) Amazon Web Services is Amazon’s cloud division, which rents computing power the way a utility rents electricity. In the same letter, Jassy said the current artificial intelligence wave is only about three years old, and Amazon Web Services already has a far larger revenue base than Amazon Web Services itself had at a similar age. (aboutamazon.com) The bottleneck here is not demand. Jassy wrote that Amazon would deploy more artificial intelligence capacity faster if it could get enough power, servers, and chips online, which is the corporate version of saying the factory floor is full before the next building is finished. (aboutamazon.com) Amazon is also trying to avoid paying tolls to outside chip suppliers forever. Jassy said its custom silicon, including Trainium for training models and Inferentia for running them, offers better price performance than some alternatives, which is Amazon’s way of lowering cloud costs while keeping more of the stack in-house. (aboutamazon.com) That chip push is already large enough to stand on its own. Jassy said Amazon’s custom-chip business has reached a multibillion-dollar annual revenue run rate, and outside reporting on the letter said he described customers asking to buy all available Graviton capacity for 2026, showing how chip supply and cloud demand are now tied together. (aboutamazon.com) (geekwire.com) Amazon is not selling only raw computing time. Its artificial intelligence stack now includes Amazon Bedrock for accessing foundation models and Amazon Q for workplace and developer assistants, and Amazon says Bedrock is used by more than 100,000 organizations worldwide. (aws.amazon.com 1) (aws.amazon.com 2) The scale is easier to see next to the rest of the cloud business. Reuters, via Yahoo Finance, said the $15 billion artificial intelligence run rate is roughly 10% of Amazon Web Services’ broader $142 billion revenue run rate, which means artificial intelligence is already a meaningful layer inside a giant existing machine, not a side project. (finance.yahoo.com) Investors have been watching whether all this spending would crush cash flow before the payoff arrived. GeekWire noted Amazon’s free cash flow fell from $38 billion to $11 billion as capital spending jumped, and Jassy’s answer in the letter was blunt: Amazon is not spending at this level “on a hunch.” (geekwire.com) (aboutamazon.com) So the letter was really two disclosures wrapped together. Amazon said the artificial intelligence business is already large enough to measure in tens of billions, and it said 2026 will be the year it spends like that business could get much bigger very quickly. (aboutamazon.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.