It Designs Every Chip. Makes None.

LIGHT might fix it.

What’s bringing top AI companies together? Photonics — the tech that sends data through light instead of old copper wires. It’s why Nvidia, AMD, Broadcom, Microsoft, Meta and OpenAI have joined forces on an open standard for optical AI connections. 

In a new video briefing, former Wall Street exec Jason Bodner explains what it means for investors, and his top stock to buy now.

[WATCH NOW AND GET YOUR FREE STOCK PICK]

In 1990, a group of engineers in Cambridge, England, spun a small semiconductor design team out of Acorn Computers — the British company that had built the BBC Micro, one of the most popular home computers of the 1980s. The engineers — led by Robin Saxby, the first CEO — had designed a processor architecture called the Acorn RISC Machine. It was fast. It was power-efficient. And it was unlike anything Intel or Motorola was building at the time. They called the new company Advanced RISC Machines. The world would come to know it as ARM.

The founding insight was counterintuitive. ARM would not manufacture chips. It would not build factories. It would not compete with the companies it served. It would do one thing: design the processor architecture — the instruction set, the core logic, the blueprint — and license it to chipmakers who would build their own versions. Every chipmaker that used ARM’s design would pay twice: an upfront license fee to access the architecture, and a royalty on every chip shipped.

I picked up my phone this morning. Inside it — underneath the screen, behind the camera, below the glass — sits a processor designed by ARM. The chip was manufactured by TSMC in Taiwan. The design was customized by Apple. But the architecture — the fundamental instruction set that tells the processor how to process — was licensed from a company in Cambridge that employs roughly 7,500 people and has never manufactured anything.

Your phone runs on ARM. So does the one in your friend’s pocket, your child’s tablet, your car’s infotainment system, and the server that delivered this newsletter to your inbox. Approximately 99% of the world’s smartphones use ARM-based processors. More than 280 billion ARM-based chips have shipped since the company’s founding. Apple, Qualcomm, Samsung, MediaTek, NVIDIA, Amazon — all licensees. They design their own chips. They compete fiercely with each other. But they all build on ARM’s architecture. And they all pay ARM a royalty for every unit that ships.

ARM went public in 1998 on both the London Stock Exchange and the Nasdaq. In 2016, SoftBank acquired the company for $32 billion and took it private. On September 14, 2023, ARM returned to the public markets in the largest IPO of the year — listing on the Nasdaq at $51 per share. SoftBank retained approximately 90% ownership.

In the fiscal year ended March 2026 — ARM’s most recently completed year — the company reported $4.92 billion in revenue, up 23% from the year before. Royalty revenue reached $2.61 billion, up 21%, driven by the continued adoption of the Armv9 architecture, which commands a higher royalty per chip than previous generations. License and other revenue was $2.31 billion, up 25%, supported by demand for next-generation AI-enabled chip designs. Rene Haas, CEO, said it plainly: “Arm delivered another year of strong performance, reflecting continued momentum across our key end markets and increasing adoption of our latest technologies.”

That 280 billion cumulative chip number is the installed base that no competitor can replicate. Every ARM-based chip that has ever shipped — in every phone, every tablet, every smart TV, every IoT sensor — runs software written for the ARM instruction set. Developers code for ARM. Operating systems are optimized for ARM. The entire mobile software ecosystem is built on ARM’s architecture. A competing instruction set would need to convince every developer, every operating system maker, and every chipmaker to start over. Nobody does. The ecosystem is the moat.

The pricing power is generational — literally. Each new ARM architecture generation commands a higher royalty per chip than the last. Armv9, the current generation, carries a significantly higher royalty rate than Armv8. And ARM’s Compute Subsystems — pre-designed, ready-to-implement processor modules — command even higher rates because the chipmaker is licensing not just the instruction set but a complete, optimized core design. In the quarter ended March 2025, smartphone unit shipments grew less than 2% — but ARM’s smartphone royalty revenue grew 30%. The volume barely moved. The royalty per chip did the work.

The data center is where the growth accelerates. For decades, ARM was synonymous with mobile — small, power-efficient processors for phones and tablets. But Amazon built its own ARM-based server chip (Graviton). Microsoft runs ARM-based Azure servers. Google deploys ARM in its data centers. NVIDIA’s Grace CPU — the processor paired with its AI accelerators — is ARM-based. Every major cloud provider is now building or deploying ARM processors for data center workloads. And every one of them pays ARM a royalty.

The flywheel runs on the two-payment model. A chipmaker pays ARM a license fee to access the architecture — an upfront payment that funds ARM’s research and development. Then, for every chip that ships using that architecture, the chipmaker pays a per-unit royalty. The license fee funds the next generation. The royalty funds the current generation. And every new generation of the architecture raises the royalty rate — because the design is more complex, more capable, and more valuable. The chipmaker pays more. But it also sells more, because ARM-based chips now power not just phones but servers, cars, laptops, and AI accelerators. The market expands. The royalty compounds. And ARM — which employs 7,500 people and has never built a single chip — collects on every one.

WHY THIS WORKS

In 1990, a small team in Cambridge, England, spun out of a British computer company with a processor architecture that was fast, power-efficient, and unlike anything the American chipmakers were building. They decided not to manufacture. They decided not to compete with their customers. They decided to license the design and collect a royalty on every chip that used it. Thirty-six years later, more than 280 billion chips have shipped with ARM’s architecture inside them — in every phone, every tablet, every smart TV, and now in the servers that train AI models and run the cloud. ARM has never built a factory. It has never assembled a product. It has only ever done one thing: design the blueprint. And the royalty on that blueprint — paid by every chipmaker on earth, on every chip they ship — has been compounding since the day the first one left the foundry.