Arm Holdings (ARM): What the Fundamentals Show in Chip Licensing

Almost every smartphone on earth runs on an instruction set designed by a company that manufactures nothing. That is the starting point for any serious Arm Holdings stock analysis: Arm (ARM) sells architecture and design blueprints to the companies that actually build chips, then collects a royalty every time one of those chips ships. Apple (AAPL), Qualcomm (QCOM), Nvidia (NVDA) and the in-house silicon teams at the large cloud providers are all customers rather than competitors in the usual sense. Understanding how those two revenue lines behave, and how the cost base sits against them, matters more here than in most semiconductor names, because Arm's economics look nothing like a chipmaker's.

How Arm Holdings makes money

Arm's revenue arrives in two streams. The first is licensing: an upfront or subscription payment for access to Arm's architecture, CPU cores and, increasingly, its larger pre-integrated design packages. The second is royalties: a per-chip fee paid when a licensee's product reaches the market, typically years after the license was signed. Licensing revenue tells you about design activity today. Royalty revenue tells you about products designed several years ago that are now shipping in volume.

That lag is the single most important structural feature of the business. A license signed in 2024 may not produce meaningful royalties until 2027 or later, which means royalty revenue is unusually visible once it starts but slow to redirect. It also means the royalty rate per chip, not just unit volumes, drives growth. Arm has been pushing customers toward newer and more complete products: the Armv9 architecture, and Compute Subsystems (CSS), which bundle cores, interconnect and supporting logic into a package a customer can take closer to manufacture. Management has said the first generation of CSS delivers roughly double the royalty of an Armv9 design, which is why the mix shift matters as much as the chip count.

For the full year ended 31 March 2026, reported on 6 May 2026, royalties were $2.61bn and licensing was $2.31bn, so the two lines are close to balanced. Arm reports in US dollars despite being a UK-incorporated company listed in New York, and it runs a 31 March fiscal year end, which is worth remembering when comparing quarters against US peers.

What the latest reported numbers show

FY2026 was a record year by Arm's own description. Total revenue reached $4.92bn, up 23% year on year, the third consecutive year above 20% growth since the Nasdaq listing. Royalties rose 21% and licensing rose 25%. Non-GAAP earnings per share came in at $1.77 for the year, with non-GAAP operating income of $2,115m.

The fourth quarter of that year, covering January to March 2026, produced revenue of $1,490m, up 20%. Within it, royalties grew 11% to $671m while license and other revenue grew 29% to $819m. On a reported basis the same quarter delivered GAAP operating income of $438m, GAAP net income of $313m and GAAP fully diluted EPS of $0.29. The gap between the GAAP and non-GAAP figures is large and mostly reflects share-based compensation, which is a genuine cost to existing holders even when it is excluded from an adjusted measure.

The most recent completed quarter is Q1 FY2027, covering the three months to 30 June 2026 and reported on 29 July 2026. Revenue was $1.29bn, up 22%, ahead of the roughly $1.26bn analysts had expected. Royalties grew 22% to $715m and licensing grew 23% to $574m. Non-GAAP operating income was $531m, up 29%, for a non-GAAP operating margin of 41.2%, and adjusted EPS was $0.45, also up 29%. Free cash flow was $665m in the quarter and $1.4bn across the trailing twelve months.

For the quarter ending 30 September 2026, Arm guided to revenue of $1.38bn plus or minus $50m, which would be about 22% growth at the midpoint. Those results are scheduled for 4 November 2026, so at the time of writing the June quarter is the newest audited picture available.

Where the royalty growth is coming from

Smartphones remain the volume foundation, but the growth story has moved to the data center. In the June 2026 quarter, Arm said data center royalties roughly doubled year on year, driven by adoption of its Neoverse server cores. Alongside the FY2026 results, the company said its share of hyperscaler CPU compute had reached 50%, a figure that would have looked implausible a few years ago when x86 designs from Intel (INTC) and Advanced Micro Devices (AMD) held almost the entire market.

The mechanism is straightforward. Large cloud operators design their own server CPUs, and when they do, they increasingly start from Arm cores or an Arm compute subsystem rather than buying a merchant processor. Arm captures a royalty on each of those chips without taking on manufacturing risk or inventory. The same pattern is spreading into AI accelerators and networking silicon, where an Arm CPU often sits beside the accelerator as the control processor.

CSS adoption gives some sense of the pipeline. By the December 2025 quarter Arm reported 21 CSS licenses across 12 companies, with five customers already shipping CSS-based chips and two of those on second-generation platforms. Arm has also disclosed that demand for its new AGI CPU product exceeds $2bn, up from an initial expected order value of around $1bn. These are company statements about contracted or anticipated demand rather than booked revenue, and they convert into royalties only as the resulting chips ship.

The counterweight appeared in the same July 2026 update. Management lowered its expectation for full-year royalty growth from around 20% to the high teens, citing smartphone weakness connected to higher memory prices squeezing handset builds. The finance chief said the company expected any royalty softness to be offset by at least as much licensing strength. That is a useful illustration of how the two lines interact: licensing can fill a near-term gap, but it does so by pulling forward design activity rather than by adding shipping volume.

Research spending, cash and shareholder returns

Arm's cost base is dominated by engineering. The income statement charge for research and development in FY2026 was $2,934m, against $2,097m the year before. Set that against $4.92bn of revenue and the shape of the business becomes clear: this is a company that spends a very large share of its top line designing products whose royalties will not arrive for several years. Investors looking at margins should treat R&D as the main variable, since it is largely a choice about future royalty streams rather than a cost of current sales.

The balance sheet carries the spending comfortably. At 30 June 2026 Arm held $3,058m of cash and cash equivalents plus $830m of short-term investments, up from $2,751m of cash at 31 March 2026, with $902m of operating cash flow generated in the quarter alone. Arm has not established a dividend since its Nasdaq listing, and there is no recent payout record, so cash generated is being retained and reinvested rather than distributed.

Ownership is an unusual feature worth stating plainly. Arm's FY2026 annual filing reported that SoftBank Group beneficially owned approximately 922,733,999 shares, or 86.4% of issued and outstanding share capital, as of 21 May 2026, down slightly from 87.1% a year earlier. SoftBank subsidiaries also hold a margin loan facility initially secured by a pledge of 769,029,000 shares, a 72.0% equity interest. The practical consequences are that the free float is small relative to the market capitalization, that SoftBank controls the outcome of shareholder votes, and that any change in SoftBank's position is itself a material event for the share price.

What the valuation multiple is measuring

This is where an Arm Holdings stock analysis has to be careful about arithmetic. Market data readings through October 2026 put the trailing price-to-earnings ratio in the region of 230 times, calculated from a share price in the low $200s against trailing GAAP earnings of roughly $0.97 per share. Market capitalization readings across the same month ranged from roughly $225bn to $260bn depending on the date and source, so any single figure should be checked against a live quote.

A trailing P/E of that order tells you something specific and limited: the current price is not supported by current reported earnings, and is instead set against expected earnings several years out. The non-GAAP figures look very different, with FY2026 adjusted EPS of $1.77 producing a far lower multiple, but the difference between the two measures is mostly share-based compensation. Which number a reader uses changes the headline dramatically, and neither one answers the question of whether the royalty growth embedded in the price will materialize. What is measurable is the rate at which reported earnings are converging on the adjusted figures, and the pace at which royalty revenue per chip is rising.

Risks and what would change the picture

Several things would move the analysis materially. Smartphone unit weakness, already flagged in the July 2026 guidance change, directly reduces the volume base beneath royalties. A slowdown in custom silicon programs at the large cloud operators would hit the fastest-growing royalty line, since a handful of customers account for a substantial share of data center design activity. Arm's move further up the stack with CSS and its own CPU products also brings it closer to the territory of customers like Qualcomm and Nvidia, and Arm's licensing relationship with Qualcomm has already been the subject of a court dispute.

On the other side, the lag between licensing and royalties means today's license signings are a leading indicator. Rising CSS licenses, further second-generation platform shipments and continued increases in royalty per chip would show up in royalty revenue over subsequent years. Concentration risk in the SoftBank holding, the small free float and the gap between GAAP and adjusted earnings are the structural items that do not resolve with a single quarter.

What to watch from here

The next concrete marker is the Q2 FY2027 report on 4 November 2026, against guidance of $1.38bn plus or minus $50m. The specific lines worth reading are royalty revenue growth relative to the high-teens full-year expectation, the data center contribution within it, licensing revenue as the stated offset to smartphone softness, and R&D spending against the $2,934m FY2026 run rate. Beyond that, the longer-term questions are whether hyperscaler CPU share holds above the 50% level Arm has claimed, whether CSS royalties scale as the per-chip rate commentary implies, and whether reported GAAP earnings begin to close the gap with the adjusted figures that the market tends to quote. Those are the measurable items an investor can track quarter by quarter, rather than relying on any single valuation snapshot.