Arm (ARM): Licensing Bookings Predict Where Royalties Go

Arm collected $715 million in royalties last quarter without manufacturing a single chip, up 22% from a year earlier, and licensing revenue, the money customers pay upfront to use Arm’s designs, came in at $574 million, up 23%, according to Arm’s first-quarter fiscal 2027 results. Both were records for a first fiscal quarter. The company that supplies the instruction set inside nearly every smartphone processor on earth has never built a factory, and that is the first thing worth understanding before touching the valuation.

Arm licenses processor architectures and finished core designs to Nvidia, Apple, Qualcomm and most of the rest of the chip industry, then collects a royalty on every finished chip that ships using them. Licensing revenue gets paid and booked well before the royalties that follow it, because a customer signs a license, then spends one to three years designing a chip around it before that chip actually reaches volume production. That lag is not a flaw in the model; it is the single most useful piece of information Arm publishes, since a jump in licensing bookings today is a reasonably reliable preview of where royalty revenue heads once those designs ship.

My thesis is that the model itself is working exactly as advertised, licensing leading royalties by roughly two to three years, and the open question is not whether the business is good but whether 281.2 times trailing earnings already prices in more of that lag’s payoff than the licensing pipeline alone can deliver.

Two revenue lines, one business model

Total revenue for the quarter was $1.3 billion, up 22% year over year, though down 13% sequentially from the prior quarter, a reminder that licensing revenue in particular can be lumpy since it depends on when large deals close rather than arriving evenly. Full fiscal year 2026 revenue reached $4.9 billion, up from $4.0 billion the year before and $3.2 billion the year before that, a run of double-digit annual growth that has held for several years running. Gross margin sits at an extraordinary 97.5%, up from 97.0%, which is what a pure licensing model looks like once it scales: almost every incremental royalty dollar drops straight to gross profit, since Arm is not paying for materials or manufacturing capacity the way a chipmaker that actually fabricates silicon has to.

Operating margin, at 18%, is far lower than that gross figure, because Arm spends heavily on the engineering that produces the next generation of designs long before any of it earns a royalty. Net income of $0.9 billion against $0.8 billion a year earlier shows that spending gap directly: revenue and gross profit both grew handsomely, and a large share of the incremental gross profit went straight back into research rather than showing up as reported profit.

What licensing signed in Q1 predicts

The $574 million of licensing revenue this quarter matters more than its size suggests, because it is effectively pre-selling future royalties. A customer that licenses a new Arm core today typically ships chips built on it starting one to three years from now, and once those chips reach volume, they start paying royalties on every unit for years afterward. Licensing growing 23% while royalties grew 22% in the same quarter tells me the pipeline behind the royalty line is not shrinking relative to what is already being collected, which is the healthiest pattern this model can show: the leading indicator is not falling behind the lagging one.

Data center royalty revenue “continues to more than double” year over year, driven by Arm-based server chips ramping at major cloud providers and by networking chips such as data processing units and SmartNICs, per the same release. That is the fastest-growing piece of the royalty base by a wide margin, and it is also the newest: Arm was a phone-and-tablet story for most of its public life, and data center royalties are a distinctly different customer base with a different, typically higher, royalty rate per chip than the mobile business Arm built its reputation on.

Mobile still matters more to the overall royalty number than the data center headline suggests, simply because there are far more phones sold each year than servers. What has changed is the mix inside mobile itself: newer, higher-royalty-rate designs make up a larger share of shipments each year as older phone models cycle out of the installed base, so even flat unit volumes in smartphones can still grow royalty revenue if the chips inside those phones are newer Arm designs than the ones they replace. I read that as a second, quieter growth lever sitting underneath the more visible data center story, one that shows up gradually in the blended royalty rate rather than in a single headline number.

A multiple that assumes the lag pays off

Here is where I get more careful. That is a lot to pay. Arm trades at 281.2 times trailing earnings and 176.4 times forward earnings, against a five-year average of 359.3. Forward earnings per share of $1.56 against trailing earnings of $0.98 implies growth of 59%, an aggressive number even for a company growing revenue in the low-to-mid 20s percent. Price-to-sales runs 49.5, above its own five-year average of 38.4, which tells me the market is not just paying for the current growth rate; it is paying up for the expectation that data center royalties keep compounding at the pace they are showing right now.

Metric	Arm now	Context
Trailing P/E	281.2	five-year average 359.3
Forward P/E	176.4	implies EPS growth of 59%
Price-to-sales	49.5	five-year average 38.4
Royalty revenue, latest quarter	$715 million	up 22% year over year
Licensing revenue, latest quarter	$574 million	up 23% year over year
Gross margin	97.5%	up from 97.0% a year earlier
Analyst target, average	$315	14% above the current price

That is a materially higher multiple than most of the semiconductor names I have covered, including the fabless leaders. Nvidia’s moat in software and networking and AMD’s case as the second source for AI compute both trade on their own aggressive growth assumptions, but neither carries a trailing P/E anywhere near Arm’s. Part of the gap is structural: Arm’s near-100% gross margin and royalty-based, capital-light model deserve a premium over companies that own fabs or pay foundries directly. Part of it, I think, is the market treating the licensing-to-royalty lag as a near-certainty rather than the multi-year bet it actually is.

Twenty analysts cover the stock, 85% rate it a Buy, and the average target of $315 implies 14% upside, with a wide range between the $230 low estimate and the $500 high one. That spread, wider than most large-cap coverage I track, reflects genuine disagreement about how much of the AI-driven data center royalty growth is durable versus how much gets competed away as Arm-based server chips move from novelty to commodity.

The implied 59% earnings growth baked into the forward multiple is worth sitting with for a moment, because it is not a modest number. A company already growing revenue in the low-to-mid 20s percent range being priced for earnings growth more than double that pace means the market expects margin expansion on top of revenue growth, not just more of the same. Some of that is plausible given how much of Arm’s cost base is R&D spending that does not need to scale linearly with revenue; a dollar of incremental royalty revenue costs Arm very little to collect once the underlying design work is already done. But it leaves less room for a merely good quarter to satisfy the stock. Good, on a name priced like this, has to keep meaning record-breaking, or the multiple has further to fall than the business itself would justify.

The risk in trusting a two-year lag

The clearest risk to this thesis is that a design win recorded as licensing revenue today is a strong hint, not a guarantee. A customer’s chip program can slip, shrink in scope, or get canceled outright before it ever reaches volume production, and when that happens the royalty revenue the market priced in in advance simply never shows up. Arm does not publish a clean conversion rate from licensing bookings to eventual royalty dollars, so I am inferring the relationship from the pattern in the reported numbers rather than from a disclosed formula, and that inference could be wrong in either direction if a large customer’s plans change.

I would also flag that the stock’s own recent behavior has been unusually volatile for a company this size: the 52-week range runs from $100 to $453, meaning the stock is 39.1% below its high while sitting far above its low, a wider swing than the underlying royalty and licensing growth rates alone would explain. Short interest is a modest 1.6% of the float, so the swings look more like a high-multiple growth stock repricing on sentiment than a heavily contested short thesis playing out.

Compare that to a steadier, lower-multiple compounder like the one I described in Microsoft at 27.5 times earnings: Microsoft’s multiple asks for execution on a business the market can already see clearly, while Arm’s asks investors to trust a reporting relationship, licensing today predicting royalties years out, that has held so far but has not been tested through a full downturn in chip demand. If data center royalty growth decelerates from “more than doubling” toward the 20% range the rest of the royalty book runs at, a multiple built on the faster number would have real room to compress even if the underlying business keeps growing.

The number I would watch next is licensing revenue growth in the following two quarters. If it holds near the low-to-mid 20s percent alongside continued data center royalty strength, the lag is still working in Arm’s favor. A licensing slowdown, even while royalties stay strong on already-signed deals, would be the first sign that the multiple is pricing in more design wins than the pipeline can actually deliver two years from now.

Analysis and opinion only, not investment advice. Figures come from Arm’s quarterly results on its newsroom and its investor site; valuation multiples are approximate and were checked on September 18, 2026.