Arm sits at the center of the semiconductor industry while remaining largely invisible to end users, because designs from the British-based, American-listed, Japanese-controlled firm underpin almost all of the world’s smartphones and most other connected devices. Instead of manufacturing hardware, Arm licenses its processor architectures to customers, who can modify the blueprints and either produce the chips in their own facilities or contract chip foundries to do so. The company collects an upfront licence fee for access to its designs and then earns a slim per-chip royalty on every unit shipped that is based on its intellectual property.
The scale of Arm’s reach is vast, as more than 300bn chips built on its designs have been shipped, and over 30bn of them last year alone. That ubiquity has historically been driven by a focus on low power consumption and flexibility, which made Arm the default choice for mobile devices and a wide range of embedded systems. The company’s business model has allowed it to remain relatively asset light, avoiding the capital intensity of chip fabrication while still participating in the explosive growth of computing hardware. Its success rests on persuading manufacturers that a standardized, licensable architecture offers better economics and faster time to market than designing processors entirely in house.
The emergence of artificial intelligence is beginning to strain that winning formula, because demand is shifting toward more powerful, specialized chips that can accelerate complex machine learning workloads in data centers and at the network edge. Arm’s traditional approach of collecting modest royalties on huge volumes of relatively simple chips may leave value on the table in markets where customers are willing to pay far higher prices for cutting edge performance. As artificial intelligence workloads proliferate, Arm faces strategic choices over how aggressively to move up the value chain, how much to adapt its licensing terms, and how tightly to align with manufacturers that are racing to build new classes of processors on top of its core designs.
