[News] Meta’s Muse Highlights AI Agents’ Higher CPU-to-GPU Needs, Could Benefit Intel, AMD, and Arm

The rise of personal AI agents could change the computing requirements behind AI workloads. According to Wccftech, personal AI agents such as Meta’s Muse, which interact with websites and apps much like humans do, could represent the next major step in AI-driven productivity. The report estimates that CPU-to-GPU ratios for some agent workloads could range from around 4:1 to as high as 40:1, creating a potential new demand driver for Intel, AMD, and Arm.

A key factor is their reliance on cloud-based virtual machines (VMs). Meta’s Muse, for example, uses multiple VMs that can remain active even after users close the app, requiring continuous CPU resources to manage operating systems, memory, and background processes, as noted by Wccftech.

AI agents also rely on CPUs to coordinate increasingly complex workflows. TrendForce notes that agentic AI requires an orchestration layer to schedule tasks, route tool calls, move data between sub-agents, and evaluate whether a request has been completed. These coordination workloads are largely handled by CPUs, adding another source of CPU demand.

The potential impact could grow as personal AI agents gain users. According to TechCrunch, citing Apptopia estimates, Muse recorded 1.8 million iOS downloads in the U.S. and Canada during its first 12 days, compared with 1.3 million for ChatGPT over the same post-launch period.

The rise in CPU demand comes as supply is already under pressure. Intel CEO Lip-Bu Tan said that the company can currently meet only about 50% of customer demand, Barron’s notes. Meanwhile, AMD expects the server CPU market to grow by more than 35% annually to over US$120 billion by 2030, up from its previous forecast of 18% annual growth, as agentic AI drives higher compute requirements.

Personal AI Agent Competition Heats Up

Beyond Meta’s Muse, other personal AI agents are emerging. Wccftech notes that Spear Street Technology’s Instinct uses a persistent cloud computer to carry out real-world tasks, but takes a different approach by letting users interact with the agent entirely through messages and calls rather than a traditional app or dashboard.

Competition could intensify further. Wccftech notes that OpenAI is rumored to be preparing a Muse competitor that could launch as soon as this week, following Sam Altman’s recent hint at a series of upcoming product launches. OpenAI has already taken steps in this direction. As noted by Digital Today, the company earlier this year hired a developer behind OpenClaw, a popular open-source AI agent, to work on a next-generation personal agent.

Read more

• [News] Meta Targets 1H27 Deployment for New TSMC-Made AI Chip, Claims Efficiency Edge Over NVIDIA
• The Great Rebalance: How Agentic AI Is Reshaping the CPU:GPU Ratio

(Photo credit: Meta)