Amazon flags big AI revenue and capex plans
Amazon’s shareholder letter disclosed AWS AI revenue running at more than $15 billion annualized and outlined roughly $200 billion of 2026 capex—largely tied to AI infrastructure like chips, satellites and robots. That level of investment signals cloud providers are aggressively treating AI as a priority for both product development and capacity build‑out. (x.com)
Amazon just put a hard number on a business it had mostly described in adjectives: its artificial intelligence services inside Amazon Web Services are now running at more than $15 billion a year. Chief executive Andy Jassy disclosed that figure in his shareholder letter on April 9, 2026. (aboutamazon.com) That matters because Amazon Web Services is the part of Amazon that rents computing power to other companies, like a giant utility for software. When customers build chatbots, code assistants, or image generators, they pay for the chips, storage, and software tools sitting inside Amazon’s data centers. (aboutamazon.com) Amazon paired that revenue number with a much bigger one: roughly $200 billion of capital spending in 2026. Jassy said most of that money will go into artificial intelligence infrastructure, which means more data centers, more networking gear, and more computing chips. (aboutamazon.com) The scale jumps out when you compare it with last year. Amazon spent $131.8 billion on capital expenditures in 2025, so a $200 billion plan for 2026 would be about $68 billion higher, or roughly 52% more in a single year. (aboutamazon.com) Amazon is arguing that this is not speculative spending. Reuters reported on April 9 that Jassy said a large portion of the planned investment is backed by signed contracts and customer commitments, which is Amazon’s way of saying the warehouses are being built after orders have already shown up. (reuters.com) The company also wants investors to know it is not relying only on Nvidia’s chips. In the same letter, Jassy said Amazon’s custom chip business has reached a revenue run rate of more than $20 billion annually, led by Trainium for training models and Inferentia for running them after they are built. (aboutamazon.com) That chip push connects directly to margins. Amazon says its in-house chips can offer better price-performance than some alternatives, so every server it fills with its own silicon can lower the cost of serving artificial intelligence workloads while keeping more of the economics inside Amazon. (cnbc.com) The spending plan also reaches beyond cloud computing. Amazon’s fourth-quarter 2025 earnings release said 2026 costs would include higher spending on Project Kuiper, the company’s satellite internet network, and Jassy’s letter tied robotics and logistics automation to the same long-term infrastructure buildout. (aboutamazon.com 1) (aboutamazon.com 2) So the story is not just that Amazon has a fast-growing artificial intelligence business. It is that one of the world’s biggest cloud companies is now treating artificial intelligence the way Amazon once treated warehouses and delivery networks: as a capacity race where the winner is the one that builds enough before demand outruns supply. (aboutamazon.com)
