ARM: The Economics of Licensing Instead of Manufacturing

THE BUSINESS IN ONE SYSTEM

Arm generated $4.92 billion of revenue in fiscal 2026 while reporting a 98% gross margin. Royalty revenue reached $2.61 billion, up 21%, and license revenue reached $2.31 billion, up 25%.

A deliberate boundary produces that combination. Arm pays for the architecture and reusable intellectual property; customers pay most of the chip-design, manufacturing, inventory, and distribution costs required to put it into devices.

The boundary lets one engineering base reach smartphones, vehicles, embedded systems, cloud servers, and AI infrastructure without Arm owning a fabrication plant. Every licensee can adapt the architecture to a different market. Wider deployment then attracts software support and produces a larger royalty stream, which funds the next generation of designs.

Thesis: Arm concentrates investment in reusable compute designs, then lets licensees spend the capital required to turn those designs into specialized chips. Device adoption expands the software ecosystem and the royalty base that finances the next architecture.

SYSTEM MAP

SYSTEM BREAKDOWN

MECHANISM 01

A modern processor begins with years of work before a customer can manufacture anything. Instruction sets, CPU cores, interconnects, memory systems, security features, verification tools, and software support all have to function together. Arm packages much of that work as licensable intellectual property. Customers can start from a tested foundation instead of building an entire compute architecture alone.

Arm’s spending shows where it concentrates capital. Its fiscal 2026 Form 20-F reports $2.78 billion of research and development expense, equal to 56% of revenue. Arm employed 8,058 engineers at year-end, about 84% of its workforce. Each reusable design gives that common engineering bill more than one chance to earn a return.

Reuse changes the unit of competition. A chip company still needs its own system design, differentiation, and manufacturing plan, but it can direct more of that work toward a target market. The purchased foundation shortens the distance between an idea and a verified design. Arm earns license revenue for access to the technology and royalties when customers ship products containing it.

MECHANISM 02

The same architecture can sit underneath products with very different constraints. A phone processor must balance performance, battery life, radios, cameras, and thermal limits. An automotive chip may prioritize safety, long product cycles, and deterministic behavior. A cloud CPU cares about throughput, memory bandwidth, and the economics of an entire server fleet.

Arm does not have to choose one finished chip for all those markets. Licensees combine Arm technology with their own accelerators, memory controllers, packaging choices, and system software. They also pay for tape-out, fabrication, inventory, and customer support. The capital required to discover demand stays largely with the company closest to that demand.

Licensees carry the cost of market experiments. A failed device program can hurt the licensee without placing Arm’s full balance sheet behind the inventory; a successful one creates license income and, after shipment, a recurring royalty stream. Arm’s fiscal 2026 results show the two clocks working together: $2.31 billion of license and other revenue reflected current agreements, while $2.61 billion of royalties reflected chips already entering the market.

The delay between signing and shipment makes the revenue streams different signals. A license can indicate design activity years before its royalty appears. Royalties reveal which past designs reached commercial volume. Operators should evaluate both: licenses describe the development pipeline, while royalties reveal market adoption.

The model also preserves customer choice at the implementation layer. Two licensees can use related Arm technology and produce chips with different performance, cost, and power profiles. Competition among those chips expands the architecture without requiring Arm to predict one winning product.

MECHANISM 03

Licensing would be a smaller business if every customer produced a close substitute for the same handset processor. The economic prize comes from allowing the architecture to travel. Arm technology now appears across mobile, consumer electronics, industrial systems, vehicles, networking equipment, and data centers. Each category brings different buyers and replacement cycles.

The spread also lowers dependence on one product generation. Smartphone volumes remain important, yet automotive, infrastructure, and cloud add royalty sources with different pricing and duration. In its May 2026 shareholder letter, Arm said data-center royalty revenue more than doubled year over year. Management also highlighted demand across smartphones, edge AI, physical AI, and cloud AI.

Broader reach can increase the royalty captured per device. Newer Arm architectures and compute subsystems carry more content than a basic core, while server and automotive designs can command higher royalty rates than low-cost embedded chips. The company therefore benefits from both unit growth and a richer mix. Licensees keep the freedom to differentiate, and Arm keeps a claim on the common layer.

MECHANISM 04

A processor architecture becomes useful when software can target it reliably. Compilers, operating systems, cloud services, developer tools, security libraries, and application frameworks form a second layer of infrastructure around the chip. Developers invest in that layer when the installed base is large enough to justify the work.

Arm reported an ecosystem of more than 22 million software developers in May 2026. That installed community addresses a coordination problem every new architecture faces: hardware makers wait for software support, while software teams wait for enough devices to justify their work. Arm can point each side to participation that already exists on the other.

Compatibility reduces adoption friction for the next licensee. A company designing a new device inherits mature toolchains and familiar programming models. Cloud operators can offer Arm-based compute without asking customers to rebuild every application from scratch. Every compatible deployment strengthens the case for more software investment, which makes another deployment easier.

MECHANISM 05

Royalty revenue arrives after customers ship chips, so it turns past design adoption into funding for future architecture. The recurring stream supports engineering before the next generation has produced its own units. Fiscal 2026 royalty revenue grew 21%, while Arm increased R&D spending 34% to $2.78 billion.

One architectural improvement can travel through several licensees and millions of devices, improving performance, efficiency, security, or software compatibility across the installed base. Customers receive a larger shared R&D program than most could fund alone, while Arm spreads the cost over a broad royalty stream.

License agreements add a second source of forward funding. Arm reported 56 extant Total Access licenses and 329 Flexible Access licenses at the end of fiscal 2026. Those programs let customers evaluate or use portfolios of technology under broader commercial arrangements. They can increase the number of teams experimenting with Arm IP before an individual chip reaches production.

MECHANISM 06

A rival needs much more than a competitive core. It needs verified IP, tools, documentation, operating-system support, trained engineers, semiconductor partners, and confidence that the roadmap will remain available for years. Customers make architecture choices early in a product program; changing later can force software work and repeat verification across the system.

The installed base creates a coordination advantage. Developers support Arm because devices use it, and chip designers choose Arm partly because developers support it. An alternative can win a focused workload through superior economics or openness. It still has to assemble enough demand on both sides to become a dependable platform.

Arm also benefits from neutrality among manufacturers. Because it does not operate as a conventional merchant chip vendor for most of its history, multiple customers can build competing products on the same foundation. The launch of Arm-designed data-center silicon in 2026 expands its role and could test that perception. The company must show that deeper integration improves the platform without turning licensees into disadvantaged customers.

FAILURE MODES

Licensee conflict. More complete Arm products may capture additional value, yet they can also overlap with work performed by customers. If major licensees expect Arm to compete directly, they have a stronger reason to fund alternative architectures.

Royalty concentration. A broad architecture can still depend heavily on a few device categories and large customers. Weak unit demand, delayed product cycles, or customer concentration can reduce royalties before newer markets become material.

Software fragmentation. Compatibility is an asset only when implementations behave predictably. Divergent extensions, inconsistent tooling, or poor support for emerging workloads can raise the cost of moving software across Arm-based systems.

OPERATOR RULE

License the shared layer when its reuse saves customers more time and risk than proprietary control would create. Keep investing where one improvement can reach many downstream products. If every customer requires a separate architecture, separate tools, and separate support, the business is selling projects rather than compounding intellectual property.

HELP SHAPE THE FIRST SIMPLIFYMBA TOOL

If you are working through a real decision in the next 90 days, tell us what is getting in the way. We will use the responses to choose one practical tool to build first.

Two minutes. No sales pitch.

SOURCE NOTES

Reporting window: information available through July 27, 2026. Arm’s fiscal first-quarter 2027 results scheduled for July 29, 2026 are outside the cutoff.