Arm Holdings CEO Rene Haas said the rise of agentic AI will create a CPU market opportunity exceeding $100 billion by 2030 [1].
This projection underscores a shift in AI hardware requirements. While GPUs have dominated the current AI era, the next phase of autonomous agents will require significant CPU power to manage complex operations.
Haas said that agentic AI workloads require CPUs to coordinate tasks, manage memory, enforce security, and orchestrate accelerators [1]. This necessity creates a massive demand for CPU capacity to support the infrastructure of autonomous AI systems.
During a Bloomberg interview on Tuesday, Haas said the role of regional partnerships is critical in meeting this demand. He said, "Taiwan's partner ecosystem is critical for AI development" [2]. The mature semiconductor and design ecosystem in Taiwan is positioned to supply the components necessary for this growth [2].
Arm is also evolving its own business model to capture this growth. For the first time in its 35-year history, the company will produce its own silicon [3]. This strategic pivot comes as the company looks toward long-term financial expansion, with projected total revenue reaching $25 billion by 2031 [4].
The company's focus on agentic AI marks a transition from static models to systems that can act independently. These agents rely on the CPU to serve as the central coordinator, ensuring that the right data reaches the right accelerator at the right time.
“Taiwan's partner ecosystem is critical for AI development.”
The shift toward agentic AI suggests that the hardware bottleneck may move from raw compute power—typically handled by GPUs—to the orchestration and management layer handled by CPUs. By integrating in-house silicon production and leaning on Taiwan's manufacturing density, Arm is attempting to transition from a licensing firm to a more integrated hardware player to capture the projected growth in autonomous AI infrastructure.
