Sinolink: MUSE accelerates a new computing power narrative, with multiple points of benefit across the industry chain.

Sinolink released a research report stating that Meta launched Muse, a standalone AI agent app. After going live, downloads rose rapidly, topping the App Store charts in the U.S. and Canada, driving Meta's stock price up 21% in two weeks and marking the Agent narrative's transition from concept to actual product deployment. Muse targets billion-scale consumer users, expanding the potential demand pool by 20-100x, pushing the industry into the inference era, with token consumption expected to see substantial growth. Muse restructures the computing hardware ratio, with the CPU-to-GPU ratio shifting from the previous 1:4-1:8 to 1:1-1:2. The overall expansion of token volume amplifies computing demand across the entire chain, with GPUs, memory, SSDs, and network infrastructure benefiting simultaneously, and the entire computing power industry chain experiencing upward momentum. Sinolink's main viewpoints are as follows: MUSE ignites consumer-grade personal Agents, opening massive new space for computing power demand Meta launched Muse, a standalone AI agent app. As a consumer-grade Agent product aimed at the general public, downloads rose rapidly after launch, topping the App Store charts in the U.S. and Canada, driving Meta's stock price up 21% in two weeks and marking the Agent narrative's transition from concept to actual product deployment. Previously, AI computing demand mainly came from the tens-of-millions-scale developer community, with growth gradually plateauing. Muse, however, targets billion-scale consumer users, expanding the potential demand pool by 20-100x, pushing the industry into the inference era, with token consumption expected to see substantial growth. Tech giants such as Microsoft quickly followed suit, launching competing agent products, elevating personal Agents to an industry paradigm. The core of product competition lies not in the underlying model, but in lowering the barrier to use for ordinary users. Meta, relying on its massive user traffic and free token quotas, holds a first-mover advantage in customer acquisition. Agents are no longer limited to code creation scenarios; they can complete diverse real-world tasks such as online shopping, schedule management, and email processing. The consumer market combined with the knowledge work market opens a brand-new computing power narrative, laying the demand foundation for subsequent hardware demand expansion in CPUs, storage, GPUs, and more. Muse restructures the computing hardware ratio, CPUs usher in a structural revival, and the industry chain benefits at multiple points Muse configures a dedicated cloud virtual machine for each user. Computing loads shift from one-time inference to long-term always-on operation, with CPU and storage demand erupting first. Agent tasks are primarily serial-logic workflows involving task decomposition, process planning, and tool scheduling, significantly elevating the importance of CPUs. The CPU-to-GPU ratio is shifting from the previous 1:4-1:8 to 1:1-1:2, and the CPU market size is expected to expand rapidly. CPU supply is tightening, with Intel and AMD raising product prices, Nvidia and Arm entering the server CPU arena, and cloud providers accelerating in-house CPU development, driving incremental demand for back-end design services. CPU demand transmits upward, with FCBGA-type ABF IC substrates achieving both volume and price increases. Overseas leading substrate manufacturers are capacity-constrained and beginning expansion. Meanwhile, Agent inference traffic is spilling over to edge and distributed cloud, a trend corroborated by the large-scale computing partnership between Anthropic and Akamai. The rising CPU share does not squeeze GPU demand; the overall expansion of token volume amplifies computing demand across the entire chain, with GPUs, memory, SSDs, and network infrastructure benefiting simultaneously, and the entire computing power industry chain experiencing upward momentum. MUSE empowers on-device hardware, with AI glasses and Muse Charm building diverse Agent hardware entry points Meta integrates MUSE into AI glasses, leveraging cameras for first-person perspective environmental perception, building a "perception-understanding-planning-execution" device-cloud collaborative loop, upgrading AI glasses from a Q&A tool to a personal Agent carrier capable of completing real-world tasks, unlocking hands-free high-frequency scenarios such as shopping, office work, and health monitoring. The Meta AI Connectors ecosystem is key to capability deployment. By connecting to third-party services in e-commerce, office, and payment, it determines the upper limit of the Agent's actual execution capability, with open development tools attracting third-party developers to enrich the application ecosystem. Beyond glasses, Meta released Muse Charm, a portable voice hardware device, creating a pocket-sized Agent interaction terminal. Together with AI glasses, it forms a differentiated hardware matrix. Hardware design focus is gradually shifting from screen apps to sensors, on-device processing, and cloud services. Currently, global display-free AI glasses are shipping at high speed, with shipments growing substantially year-over-year. Muse fills the killer app gap, driving short-term demand for components such as cameras, CIS, microphones, and low-power SoCs; long-term, as display requirements evolve, the value of components such as optics and micro-displays is expected to rise further. Risk warnings Risk of Muse user growth and commercialization falling short of expectations; risk of security and privacy incidents escalating; risk of CPU supply and price fluctuations; risk of platform ecosystem resistance; risk of trade friction and export controls.