Amazon just told its employees AWS will hit $600 billion by 2036, a 4.6x jump driven almost entirely by AI demand.
The Signal
Andy Jassy doesn't throw numbers like this around lightly. AWS did roughly $130 billion in 2025. To reach $600 billion by 2036 means adding $470 billion in revenue over 11 years, averaging about $43 billion in new revenue annually. That's not organic growth. That's AI infrastructure eating the enterprise world.
The math tells you what Jassy sees: companies aren't just experimenting with AI anymore, they're rebuilding their entire technology stacks around it. Every enterprise agent, every autonomous workflow, every real-time decision engine needs compute, storage, and orchestration at scale. AWS is betting it can capture the infrastructure layer of the agent economy before Microsoft or Google lock it down.
But here's the deeper signal. For AWS to 4.6x in a decade, businesses have to believe AI agents deliver enough value to justify that infrastructure spend. This isn't cloud migration Part 2. This is companies paying to let software make decisions, manage workflows, and interact with customers without human checkpoints. Jassy is basically saying: by 2036, agents running on our infrastructure will be generating trillions in economic value across every industry.
The projection also reveals timing. If AWS expects this trajectory, the agent economy isn't 10 years out. It's ramping now. The companies building agent platforms, the ones figuring out orchestration and safety and reliability, they're not early. They're right on time.
The Implication
Watch where AWS invests next. They'll build toward the bottlenecks they see coming: agent-to-agent communication, persistent memory systems, compliance and auditing for autonomous decisions. If you're building in the agent space, your success depends on infrastructure that doesn't exist yet at scale. Jassy just told you it's coming, and what it'll cost.
Source: The Information