The CPU Rally Muse Woke Up: Why Arm Really Jumped 17%

This rally is bigger than one product called Muse

On September 21, Arm rose 17.1%, Intel gained 12.1%, and AMD climbed about 10%. The same day, Meta jumped more than 11% on excitement over Muse, its AI agent product. The story the market connected was simple: as AI shifts from chatbots to agents, computing demand grows, and that demand may not stay confined to GPUs.

The causality deserves a closer look, though. There is no evidence that enthusiasm for Muse directly boosted Arm's revenue. Muse looks more like a catalyst that made an investment thesis already in motion visible to the whole market: agentic AI can reignite demand for CPUs.

Here is the chain investors are pricing in:

1. Agents plan. Instead of answering a single question, they run multi-step tasks in a loop.
2. Agents call tools and databases. Search, memory, APIs, and network operations pile up.
3. CPU work increases. Orchestration, general-purpose computation, and I/O handling grow.
4. Power efficiency turns into money. The more an agent runs continuously, the more performance-per-watt and rack density matter.

The more important shift at Arm: an IP company now sells CPUs directly

The most important fact in this story is not the stock move. It is the change in Arm's business model. In March 2026, Arm unveiled the Arm AGI CPU, which packs up to 136 Neoverse V3 cores. It is the first time in the company's history that Arm is directly supplying production silicon for data centers, rather than licensing designs for others to build. Meta is the lead partner and co-developer on this product.

Three numbers frame where things stand:

• 136 — maximum core count on the AGI CPU
• $2 billion+ — customer demand Arm has disclosed for fiscal years 2027 through 2028
• 2x+ — year-over-year growth in Arm's data center royalty revenue last quarter

The $2 billion figure should not be read as a confirmed order backlog or as revenue already booked. Arm's own language for it is customer demand. The company said it delivered initial units to multiple customers in late July and that it has secured the manufacturing capacity to support the $1 billion opportunity it had flagged earlier. In other words, the market's attention is shifting from whether the demand exists to how much of that $2 billion in interest converts into actual shipments and revenue.

One more point worth correcting: the AGI CPU is not a product waiting for its first shipment in the fourth quarter of 2026. Arm unveiled it in March, and by late July it says it had already delivered initial units to multiple customers.

Why agentic AI is making CPUs matter again

Large-model matrix math still runs overwhelmingly on GPUs and AI accelerators. But agents do more work between model calls. They look up data, read user state, call APIs, distribute tasks across multiple sub-agents, and shuttle data between network and storage.

Arm's own materials argue that as agentic applications spread, CPU requirements could grow at least fourfold from current levels. That number is a company projection, not confirmed industry data. Still, the investment logic is clear: as the number of inference calls and tool calls rises, general-purpose computing outside the accelerator grows along with it.

This is where Arm's strength lies: power efficiency and core density. In AI data centers, power has already become a hard constraint on capital spending. If a CPU can handle more work within the same power and rack footprint, that efficiency stops being a spec sheet detail and becomes part of the data center's economics.

But don't read this as "CPUs win, full stop"

On September 21, Arm, Intel, and AMD all rallied hard, but not every semiconductor stock moved in the same direction. So it is more accurate to read that day's move not as capital rotating completely "from GPU to CPU," but as AI investors' field of view expanding beyond GPUs.

Insight Times Editorial Desk