$200 billion. That's the size of the CPU market Jensen Huang says he's found inside the AI agent boom - and he's betting Nvidia can own it.
Speaking this week, the Nvidia CEO described AI agent infrastructure as a "brand new" market for the company, distinct from the GPU (graphics processing unit, the specialized chip that powers AI model training and generation) business that made Nvidia the most valuable semiconductor company on earth. AI agents are software programs that can plan and execute multi-step tasks on their own - booking a meeting, drafting a proposal, filing a report - without a human directing each action.
Why CPUs, Not Just GPUs
The logic Huang is laying out: GPUs handle the heavy computation when an AI model generates a response. But running agents continuously across enterprise infrastructure requires a different kind of chip - one that manages memory, coordinates tasks, and keeps processes running around the clock. That's CPU territory, and Nvidia's Grace CPU line is the product it wants to put in those slots.
This isn't a pivot - Nvidia's GPU business isn't going anywhere. It's an expansion bet. Huang is arguing that the shift from "query a model occasionally" to "run agents permanently" opens a hardware market that didn't meaningfully exist three years ago.
How Big Is the Real Opportunity
The $200B figure is a TAM - total addressable market - which is what analysts and executives use to describe the full theoretical size of a market if one company captured all of it. These numbers are always optimistic. Nvidia won't take $200B. But even capturing 10-15% of a market that size would represent tens of billions in new revenue for a company that posted roughly $130B in revenue over the past year.
The more useful signal here is directional: Nvidia is positioning itself for a world where AI agents become permanent infrastructure, not just a feature companies experiment with. For businesses currently evaluating how to deploy AI agents internally, the message from the hardware layer is that this is being treated as a long-term build, not a trend.