If Meta's Muse Personal Agent Scales To Just 100 Million Users, It Would Require 1.58 Million AMD Ryzen CPUs, 800 Petabyte Of RAM, And 10,000 Petabyte Of SSD Under Ideal Conditions

Key Points:

• Meta's Muse personal AI agent, designed to serve 100 million users with dedicated virtual machines (VMs), demands massive compute resources, including potentially 1.58 million AMD Ryzen EPYC CPUs if each user gets 2 vCPUs, 8GB RAM, and 100GB SSD.
• To manage CPU load, Meta may share CPU cores among active users, but even with 10-50% active user load, the required CPU count remains in the hundreds of thousands, highlighting the immense scale of infrastructure needed.
• Persistent memory and storage requirements pose an even greater challenge, with estimates reaching 100 petabytes of RAM and 10,000 petabytes of SSD to support continuous availability of personal Muse VMs.
• Meta's approach includes strong privacy measures by isolating each Muse agent in dedicated VMs monitored by a Sentinel system to prevent unauthorized transactions, reinforcing user data security.
• In response to these demanding compute needs, Meta has intensified investment in advanced computing hardware, becoming a lead partner for Arm AGI CPUs and significantly expanding its use of AWS Graviton cores in 2026.