Roughly 99% of the world's smartphones run on an ARM-designed instruction set. That single number does more work than most moat arguments — because it's not a market share stat, it's an architectural standard. When a chip designer at Apple, Qualcomm, MediaTek, or Samsung sits down to design a new processor, the foundational language that processor speaks was written by ARM Plc. And every time one of those chips ships, ARM collects a royalty.
That's the business in one sentence. ARM doesn't fabricate silicon. It doesn't compete with TSMC or Intel on manufacturing. It sells two things: an upfront license to use its architecture, and a per-chip royalty that runs for the life of the product. The royalty stream is the interesting part — it compounds without ARM needing to spend more.
The standard is the moat
Most "moat" discussions in tech default to network effects or switching costs. ARM's moat is a third, rarer kind: it's a de facto technical standard. An instruction set architecture (ISA — the basic vocabulary a processor uses to execute commands) is the layer everything else is built on. Operating systems compile to it. Apps compile to it. Compilers, debuggers, developer tools, drivers — an entire software ecosystem assumes ARM under the hood.
Switching off ARM isn't like switching cloud providers. It means recompiling, retesting, and in many cases rewriting performance-critical code. Apple spent years and billions migrating the Mac from x86 to ARM — and that was Apple, with infinite resources and full control of its stack. For most chip designers, that effort isn't remotely worth it.
This is why RISC-V, the open-source ISA that gets cited as ARM's existential threat, has been "about to disrupt ARM" for roughly a decade and still ships in a tiny fraction of high-performance designs. The technical merits of RISC-V matter less than the gravitational pull of an ecosystem ARM has spent 30+ years building.
Royalty economics: the part that compounds
Here's the structural beauty. ARM signs a licensee — say, a smartphone SoC (system-on-chip — the main processor that runs your phone) maker. That licensee pays an upfront fee for access. Then, for every chip they ship using ARM IP, ARM gets a royalty, typically a low single-digit percentage of the chip's selling price.
The upfront fee covers ARM's R&D. The royalties are close to pure margin. And royalty revenue keeps coming in from chips designed years ago — ARM still earns on architectures licensed in the early 2010s that are shipping in mid-range phones and embedded devices today. It's a back catalog that doesn't depreciate.
This is why ARM's gross margins sit in the 95%+ range. There is essentially no cost of goods sold — the "good" is intellectual property that's already been written. Operating margin is lower because R&D is heavy (ARM has to keep pushing the architecture forward), but the operating leverage (how much profit grows when revenue grows) is severe in the right direction. Every incremental royalty dollar drops mostly to operating profit.
Where the next leg of growth actually comes from
Mobile is mature. Smartphone units globally aren't growing meaningfully. So if you're trying to underwrite ARM, the mobile story is a maintenance story — protect the 99%, ride pricing and content gains as chips get more complex.
The real growth vectors are three:
Data centers. ARM-based server chips were a rounding error five years ago. Now AWS Graviton, Microsoft Cobalt, Google Axion, and Nvidia's Grace are all ARM-based. Hyperscalers want ARM because they can customize the design and pay less than they would to Intel or AMD. Every hyperscaler that shifts even 20% of its server fleet to ARM moves a huge royalty pool that didn't exist before.
Automotive. A modern car has dozens of processors — infotainment, ADAS (driver assistance), engine management, battery management. ARM is in most of them. As cars get more software-defined, the chip content per vehicle keeps climbing, and ARM's royalty per car climbs with it.
AI accelerators. Nvidia's Grace Hopper and Grace Blackwell systems pair Nvidia GPUs with ARM-based CPUs. Every AI server rack shipped in this configuration generates ARM royalties. ARM didn't build the AI wave, but it's positioned as the default CPU companion to it.
What can actually break this
Three real risks, in rough order of seriousness.
First, royalty rate compression. Big customers — Apple, Qualcomm, the hyperscalers — have leverage. If ARM pushes too aggressively on per-chip pricing or its newer per-device pricing models, the largest licensees have the engineering depth to invest in RISC-V alternatives as a negotiating threat. The Qualcomm-ARM legal dispute over licensing terms is a preview of how messy that can get.
Second, RISC-V in specific verticals. RISC-V probably won't dethrone ARM in smartphones or general-purpose servers anytime soon. But in narrower domains — microcontrollers, specific embedded applications, China-driven sovereign chip efforts — it's already taking share. Death by a thousand verticals is a slower threat than catastrophic displacement, but it's real.
Third, customer concentration and geopolitics. A meaningful chunk of ARM's royalty base flows through Chinese chip designers. Export controls, the ARM China governance situation (which has been its own multi-year saga), and broader US-China decoupling all sit as overhangs that don't show up in any DCF cleanly.
How to think about valuation without anchoring on a number
ARM trades at a multiple that prices in serious growth — that's not a secret, the market has clearly decided this is a structural winner. The question worth asking isn't "is it cheap." It's "what has to be true for the multiple to make sense?"
Roughly: data center royalty share has to keep climbing, AI server attach rates have to hold, automotive chip content per vehicle has to keep growing, and royalty rates on newer architectures (v9 and beyond) have to stick at the higher levels ARM has been pushing. None of those are crazy assumptions individually. All four landing well is what's priced in.
The inverse question — what would make this a genuinely defensive holding — is whether the moat survives a multi-year tech downcycle. The royalty back catalog suggests it does. Even if licensing slows, royalties on already-shipping designs keep flowing. That's the part of the business that earns the "strongest moat" label rather than "strongest growth story" label.
The two are different. A growth story can fade. A standard, once entrenched, is very hard to dislodge — and ARM is the standard.
— Anuj