Personal AI agents such as Meta's Muse are expected to unleash a super cycle for CPUs, benefiting Intel, AMD, and Arm. Muse is a personal AI agent that executes multi-step tasks on a user's behalf, including opening a cloud-based browser, navigating websites as a human would, and managing background chores even after the app is closed. Meta is expanding the list of connectors that link Muse to third-party apps and services. The demand signal is the CPU-to-GPU ratio, with estimates ranging from 4:1 all the way to 40:1, and some forecasting the high end.

That an agent that browses, clicks, and manages tasks is a CPU-heavy workload, not a GPU-heavy one. Unlike a training run or a large inference batch, an agent that drives a browser and orchestrates chores leans on the host CPU for scheduling, I/O, and orchestration. The report notes that if Muse scales to 100 million users, it would require 1.58 million AMD Ryzen CPUs, 800 petabytes of RAM, and 10,000 petabytes of SSD under ideal conditions. That is a CPU-and-memory bill, not a GPU bill, which is why the ratio prediction matters for the CPU vendors.

This reframes the AI hardware demand story. The dominant narrative has been GPU-centric, with accelerators as the scarce resource. An agent workload that is CPU-heavy shifts part of the demand back to the CPU side, which is good news for Intel, AMD, and Arm. The 40:1 figure is the upper bound of the forecast range, and the 4:1 floor is the conservative end, so the actual ratio is uncertain but the direction is a CPU uplift.

The actual CPU-to-GPU ratio, the user count Muse will reach, and whether the 100-million-user scenario is realistic. The forecast is a prediction, not a measured result, and the 1.58-million-CPU figure is an ideal-condition estimate, not a committed build. Meta's paid tiers start around $20 per month for heavy users, and the free allowance is up to 100 million tokens per week, but the scaling trajectory is not stated.

Whether Muse's user base grows toward the forecast scale, whether the CPU-to-GPU ratio holds at the high end, and whether Intel, AMD, and Arm position their roadmaps for agent workloads. If the agent workload becomes a mainstream AI category, the CPU side of the datacenter bill grows, and the three CPU vendors gain a new demand driver beyond the traditional server and PC markets. The 100-million-user scenario is the upper bound of the forecast, and the 4:1 floor is the conservative end, so the actual ratio is uncertain but the direction is a CPU uplift for the three vendors.

An agent workload that browses and orchestrates is CPU-heavy, not GPU-heavy. The 40:1 ratio is the forecast ceiling; the direction is a CPU uplift for Intel, AMD, and Arm.