Meta’s Muse personal AI agent has intensified investor interest in the server processors needed to run software that can complete tasks rather than only answer prompts. Muse launched on Sept. 8, and its early traction helped drive a broad semiconductor rally on Sept. 21, when Meta shares rose 11. AMD’s market value moved above $1 trillion for the first time, while the Philadelphia Semiconductor Index rose 4.3% for a fifth consecutive session.
The hardware implications extend beyond the GPUs typically associated with artificial intelligence. The environments run on AMD EPYC 9D25 Turin processors, which have up to 126 cores and 254 threads, making high core density important when the same architecture is replicated across large numbers of users.
Arm could benefit by licensing the designs used in custom server processors and by developing its own server chips.
Intel entered the agent surge with more than 40% of the data-center CPU market in 2025, but its projected share is about 22% in 2030; supply constraints have nevertheless allowed it to sell older processors and raise prices. Nvidia’s CPU opportunity is smaller relative to its dominant GPU business, although its Arm-based processors can be sold separately, paired with GPUs or included in complete server systems.
The immediate market response reflects expectations about future infrastructure demand, not confirmed orders from every company named. In a separate technical examination of Muse, the agent’s CPU-only sandbox was distinguished from the GPU infrastructure used for inference, while Meta said the runtime environment was deliberately isolated from more sensitive systems. The result is a hardware market in which agent adoption could increase demand for CPUs, accelerators and complete systems at the same time, with the likely beneficiaries depending on how workloads are divided between those components.