Arm Is in Every Phone and Priced Like It's in Everything Else

Begin with the business, because it genuinely is one of the best ever built and the case here is about price and strategy, not quality. Arm does not make chips; it designs the fundamental architecture — the instruction set and the core designs — that chipmakers license and build upon, paying Arm a fee to license the design and then a royalty on every chip they ship using it. Because Arm's architecture became the standard for mobile computing, it collects a royalty on virtually every smartphone processor in the world, plus a growing share of chips in cars, embedded devices, and now data centers. It is an almost perfect royalty model: capital-light, enormously high-margin, and embedded so deeply in the global semiconductor supply chain that switching away from it is, for most customers, prohibitively difficult.

The numbers are excellent. In fiscal 2026 Arm reported record revenue of $4.92 billion — royalty revenue of $2.61 billion, up 21%, and licensing revenue of $2.31 billion, up 25% — its third consecutive year of 20%-plus growth since going public. Its royalty rates are rising as customers adopt the newer Armv9 architecture and its more complete Compute Subsystems packages, which carry richer take rates. And the data-center business, long the holy grail, is finally inflecting: data-center royalty revenue more than doubled year over year as AI workloads drove demand for Arm-based server chips. This is a genuinely accelerating, genuinely dominant franchise riding the biggest trend in technology.

So this essay does not argue that Arm is a bad business. It argues that the price assumes a future of near-perfect execution and expansion, that the strategic pivot Arm is making to capture more of that future introduces a real and underappreciated risk, and that the foundation of the whole model — a royalty everyone pays because there is no free alternative — faces, for the first time, a credible free alternative being built alongside it.

A royalty business priced like a hypergrowth stock

Start with the valuation, because it is genuinely one of the most extreme in the market and everything else has to be judged against it. Depending on the measure and the moment, Arm has traded at a forward price-to-earnings ratio well over 100 — at times approaching or exceeding 180 — at roughly 38 times revenue, and around 100 times its royalty revenue. These are not the multiples of a royalty business; they are the multiples the market awards to the most explosive hypergrowth stories in technology, and Arm's actual royalty growth, while excellent, is around 20% a year, not the triple-digit growth such multiples imply.

This is the central tension. Arm's royalty revenue — the recurring, high-quality, annuity-like part of the business — grew 21% in fiscal 2026. Its licensing revenue grew faster, 25%, but licensing is the lumpier, lower-quality part: it reflects customers signing up to use Arm's designs, which is encouraging as a leading indicator but is not the same as the durable, compounding royalty stream, and it can be front-loaded and uneven quarter to quarter. A business growing its core royalties at roughly 20%, even with rising royalty rates and a real AI tailwind, is a wonderful thing — but a multiple of 100-to-180 times earnings prices in not 20% growth but years of dramatic acceleration, as if the data-center and AI royalty ramp will not merely continue but explode, lifting the whole company's growth rate far above its current trajectory and sustaining it for a decade. That may happen. But the price has already assumed it happened, which means the stock can fall sharply even if Arm simply keeps growing at its current, excellent, 20% pace — because 20% growth does not justify a 150-times multiple, and the gap closes by the multiple compressing, not by the business failing.

The Switzerland that started building tanks

Now the strategic risk, which is the most important and least appreciated part of the Arm story. For its entire history, Arm's value rested on being neutral — the Switzerland of chip design. Every chipmaker, no matter how fierce their rivalry, could license Arm's architecture because Arm competed with none of them; it sold the blueprint to everyone and built nothing itself. Apple, Qualcomm, Nvidia, MediaTek, Samsung, and the cloud giants all license Arm precisely because Arm was a trusted, neutral supplier of the foundation they all built on. That neutrality was the moat. It was why customers were comfortable making Arm's architecture the indispensable core of their own products.

Arm is now abandoning that neutrality. It has begun building its own production chips — most notably an Arm AGI CPU aimed at AI data centers — moving from selling blueprints to selling finished silicon that competes directly with the chips its own customers build using Arm's designs. The strategic logic is obvious: the finished chips capture far more value than the royalty, and Arm wants a larger slice of the AI data-center boom than a small per-chip royalty provides. But the risk is equally obvious and it is the same poison that afflicts any neutral platform that becomes a competitor: the moment Arm builds chips that compete with Qualcomm's and Nvidia's and the hyperscalers' own designs, those customers have a reason to question whether they want to keep building their most important products on the architecture of a company that is now their rival. A licensor that competes with its licensees has put its royalty stream — the high-quality annuity that justifies the premium — into tension with its new chip ambitions. The Switzerland of semiconductors has started building tanks, and its customers, who are also its arms-buyers, have noticed.

The Qualcomm dispute is the evidence of the relationship souring — and the outcome is worse for Arm than mere litigation. Arm sued Qualcomm and its Nuvia subsidiary over licensing terms, and Qualcomm won a decisive court victory: a judge ruled that neither Qualcomm nor Nuvia had breached Arm's architecture license, dismissed Arm's remaining claim, and denied Arm's request for a new trial, with further matters lingering into 2026. In other words, Arm took one of its largest customers to court over the scope of its licensing rights and lost — a defeat that both signals a once-frictionless relationship turned adversarial and sets a precedent that constrains Arm's ability to extract more from licensees through legal pressure. When your royalty model depends on the goodwill and continued commitment of a handful of enormous customers, suing one of them and losing, while simultaneously launching chips that compete with all of them, is not a small risk. It is a direct threat to the durability of the royalty stream the valuation capitalizes at 100 times.

The free alternative being built alongside the toll road

There is a final threat to the foundation, and it is structural and long-term: RISC-V, the open-source, royalty-free instruction-set architecture. Arm's entire model rests on the fact that designing a competitive chip architecture from scratch is so hard that everyone pays Arm rather than build their own. RISC-V changes that calculus by offering a free, open, increasingly capable alternative architecture that anyone can use without paying a royalty to anyone. It is, in the language used elsewhere in this series, a bypass being built alongside the toll road — a parallel route that lets chip designers avoid the Arm toll entirely.

For years RISC-V was too immature to matter for high-performance computing, and for the most demanding applications it still trails Arm's ecosystem. But it is improving, and the parties most motivated to adopt it are precisely the ones whose defection would hurt Arm most: cost-conscious hyperscalers who ship chips in enormous volumes and would love to avoid per-chip royalties, and Chinese firms seeking to escape dependence on Western-controlled IP amid geopolitical tension. Every large customer that shifts even part of its roadmap to RISC-V is royalty revenue Arm never collects. RISC-V does not have to replace Arm to damage the thesis; it only has to give Arm's largest customers a credible threat to walk, which caps Arm's pricing power and erodes the assumption — baked into the 100-times multiple — that Arm's royalty rates only ever rise and its reach only ever expands. A toll road is most valuable when there is no other way around; RISC-V is the other way, and it is being paved a little more every year.

The core market that stopped growing

It is worth examining the foundation of Arm's royalty stream — smartphones — because the part of the business that is certain is also the part that has largely stopped growing, which is exactly why the valuation depends so completely on the part that is not yet certain. Arm's near-universal presence in smartphone chips is the bedrock of its royalty revenue, but the global smartphone market is mature and saturating: people are not buying dramatically more new phones each year, and in many markets they are holding their devices longer and buying used ones, which means fewer new chips shipped and fewer new royalties for Arm. The unit-volume growth engine that once powered Arm's royalties has slowed to a crawl.

This is why Arm's growth story has migrated almost entirely to two levers that are higher-quality but also less certain than unit growth: rising royalty rates (charging more per chip via Armv9 and Compute Subsystems) and new markets (data centers, automotive, AI). The rate lever is real and powerful, but it has limits — customers will only accept rising per-chip royalties up to the point where building their own architecture, or adopting RISC-V, becomes the cheaper option, which is precisely the ceiling RISC-V is lowering. And the new-market lever, chiefly the data-center and AI opportunity, is the explosive growth the valuation is really paying for — which means the entire premium rests not on Arm's certain, dominant, but slow-growing smartphone base, but on the successful capture of a data-center market it is only beginning to penetrate, against entrenched competitors, at a moment when it has chosen to compete with the very customers it needs to win there.

In other words, the safe part of Arm is mature and the growing part of Arm is the contested, not-yet-won part — and the stock is priced almost entirely on the latter. An investor paying 100 times earnings is not paying for the phone royalties, which are a slow-growing annuity worth a fraction of the price; they are paying, almost in full, for a data-center and AI royalty future that is real and promising but unproven, contested, and dependent on every strategic gamble breaking Arm's way. The certain part of the business does not support the multiple; the multiple is a bet, almost entirely, on the uncertain part.

What the bulls genuinely get right

In fairness, the bull case is powerful and Arm's quality is not in dispute — the debate is the price and the strategic gamble. Several points genuinely support the optimism. The royalty model is one of the best in all of technology: capital-light, ~95%-plus gross margin, embedded in billions of devices, with enormous switching costs that make Arm's architecture extraordinarily sticky. The AI and data-center tailwind is real and accelerating — data-center royalty more than doubling is a genuine inflection, and a $100-billion-plus data-center CPU opportunity by the end of the decade is a credible prize. Rising royalty rates from Armv9 and Compute Subsystems mean Arm captures more value per chip even without shipping more chips, a real and durable margin lever. The move into custom silicon, for all its risk, could capture vastly more value per AI system than a royalty if it works, and management points to billions of dollars of demand for it. And RISC-V, while a real long-term threat, remains years behind Arm's ecosystem for the most demanding applications. This is a dominant franchise with multiple real growth vectors, and a premium is entirely warranted.

The honest synthesis is that Arm is an extraordinary business whose valuation has run far ahead of even its extraordinary fundamentals, and whose strategy to grow into that valuation — competing with its own customers — introduces precisely the risk most likely to damage the royalty annuity the premium depends on. A bull who owns Arm as a dominant royalty compounder is on solid ground at a reasonable price. A bull who owns it at 100-to-180 times forward earnings is betting that the AI-royalty explosion arrives on schedule, that the customer relationships survive Arm's pivot into their territory, that the Qualcomm fight is an aberration rather than a preview, and that RISC-V never gains enough ground to matter — a stack of optimistic assumptions, each plausible, all required, all priced as near-certainties. The distinction matters because of how multiples this high behave: at 100-to-180 times earnings, the stock needs not merely good news but a continuous stream of better-than-already- extraordinary news simply to hold its ground, because anything short of acceleration disappoints a price built on acceleration. That is a precarious place for any stock to stand, and it is especially precarious for one whose growth depends on contested new markets and whose strategy is actively straining its most important customer relationships. Great businesses can be poor stocks, and the multiple is the mechanism by which it happens.

The SoftBank overhang

One more structural feature deserves a mention, because it shapes the stock independent of the business: Arm is roughly 90% owned by SoftBank, which means the public float is small and the share price is unusually sensitive to supply-and-demand technicalities as well as to SoftBank's own strategic and financial needs. SoftBank has a long history of using its assets, including Arm, as instruments in its own complex financing and investment maneuvers, and a controlling shareholder of that size means minority investors are along for a ride steered by an owner whose interests and time horizons may not match their own. A tiny float can amplify a stock's moves in both directions — it has contributed to Arm's spectacular rise — but it cuts the other way too, and a richly valued stock with a thin float controlled by a single strategic owner is a more volatile and less self-determined investment than its blue-chip role in the AI trade suggests. The valuation is extreme; the ownership structure makes it more so.

The kicker

Arm is in every phone, and that is a genuinely wonderful place to be — a royalty on the computing of the world, collected at the highest margins in technology, now riding the AI wave into the data center. The business deserves admiration and a premium. But the market has priced it not for the magnificent thing it is — a 20%- growing royalty franchise — but for a fantasy of the thing it might become: the indispensable, royalty-rising, ever-expanding architecture of literally all future computing, with no free alternative, no customer resentment, and no limit to what it can charge. Each piece of that fantasy is under pressure: the royalty growth is good but not explosive, the customers are increasingly also competitors, RISC-V is the free road being paved alongside, and the company itself has traded its neutrality for the chance to compete with the very customers whose royalties are its lifeblood. The toll road is real. The traffic is real. But at a hundred times earnings, the market has not bought the toll road — it has bought the assumption that every vehicle ever built, forever, will choose to pay.

Arm collects a few cents on nearly every phone on Earth, which is one of the great businesses ever devised — and the market has decided those few cents, multiplied across an imagined future of total computing dominance, are worth more than a hundred years of the company's current earnings. The phone royalty is a certainty. Everything the price assumes beyond it — the data centers, the rising rates, the absent competition, the loyal customers Arm has just started competing against — is a story, and the stock is the story, priced as though it were already the history. The royalty will keep coming, the AI wave is real, and Arm will likely remain one of the most important companies in technology for decades. But importance and price are different measurements, and at a hundred-odd times earnings the second has run a very long way ahead of even the magnificent first — far enough that being right about the company and being rewarded as a shareholder may, for a long while, turn out to be two entirely different things.