Report contents
At a glance
- Business model: Arm licenses processor designs and the instruction set behind them, then collects a per-chip royalty, producing a 92.5% gross margin in fiscal 2026. Software and tooling switching costs support the pricing lever, with Total Access agreements rising to 56 from 44. The obstacle is concentration: a few licensees now fund a company that also sells compute subsystems and its own AGI CPU.
- Financial quality: free cash flow of $3.14 billion over fiscal 2022-2026 slightly exceeded net income of $3.07 billion, though timing is lumpy, with the June 2026 quarter's $0.90 billion arriving as receivables fell $0.47 billion. Cash of $2.75 billion at March 2026 against $0.06 billion of short-term borrowings leaves room, but capex has risen to 11% of revenue.
Business model
Business model: an architecture toll with a silicon ambition
Arm licenses processor designs and the instruction set architecture they are built on. Chip companies pay an upfront license fee for the right to use a design, then a per-chip royalty once the finished chip ships. The payer is the chip designer, not the phone or server buyer, and the royalty is levied on unit volume, so it scales with units shipped and not with the price of the end product. That is how a company of 9,024 employees earns money from chips it never touches. The May 6, 2026 investor presentation describes the current phase as extending the model into compute subsystems, chiplets and complete chip solutions: an expansion-stage platform business starting to sell more of the chip.
Who pays, and for what
The customer base has widened beyond merchant chip vendors. Smartphone application processors remain the largest royalty pool, with consumer devices, infrastructure, automotive and IoT behind them, and systems companies that design their own silicon now license directly. That last group changes the negotiation: a single agreement can cover very large volumes, and the design decision precedes any royalty by years. Arm gives no dollar revenue split by end market, so the phone-versus-server mix rests on management commentary, and quarterly unit shipments are not reported at all.
Licensing and royalties behave differently. Licensing is negotiated, lumpy and driven by design starts; royalties arrive several quarters later and track units shipped. Incremental royalty dollars arrive with almost no matching cost, so operating leverage is extreme in both directions. The June 2026 quarter shows the effect: operating income was $0.10 billion against net income of $0.27 billion, with interest on $3.06 billion of cash and a tax credit doing much of the work.
Contracted license revenue provides a partial order book. Arm discloses remaining performance obligations, which cover signed agreements not yet recognized and give more visibility than the licensing line alone. The limitation is that they say nothing about royalties, the more recurring stream.
Pricing power sits in the wrappers
Bargaining power rests on the ecosystem. A royalty of a few percent of chip revenue is small next to the cost of rewriting software, tooling and validation for a different instruction set, and that gap is what lets Arm raise rates generation by generation. The evidence sits in the commercial formats. Total Access, a subscription giving a customer rights across the catalogue, reached 56 agreements in fiscal 2026 from 44; Flexible Access, the cheaper entry tier, reached 329 from 314. Both grew, so the format is spreading. The low-cost tier expanding nearly as fast means average revenue per customer depends on which tier new signings land in, and Arm discloses counts without revenue by tier.
Concentration caps that power. Arm's filings flag dependence on a limited number of customers for a significant share of revenue, and the largest licensees have both the strongest incentive and the deepest engineering bench to design around the architecture. RISC-V, a royalty-free open instruction set, is the obvious route. Substitution has so far landed in microcontrollers and embedded work, where software inertia is weakest, rather than in high-volume phone and PC sockets. Whether it stays there decides how much pricing room Arm has.
Selling the chip, not just the design
The larger change is Arm selling more of the chip. Compute subsystems, a bundle of CPU cores plus system IP bought as one unit, lift the amount of Arm content per design. The Arm AGI CPU, introduced in March 2026, goes further into finished silicon. That raises revenue per design and captures value that currently accrues to licensees. It also puts Arm in competition with the companies that pay it and converts a variable-cost licensing model into one carrying tape-out, wafer and inventory commitments. Implied capex rose to about 11% of revenue in fiscal 2026 from 3% in fiscal 2022. The capital consequences belong to a separate section; what matters here is that a licensee weighing a new design start now also weighs whether it is funding a rival.
Geography adds distance between Arm and the demand it serves. Revenue is booked where the licensee sits, and China, served through the locally owned Arm China entity that sub-licenses the technology, is a large piece of it. Chips designed in one country and shipped in another make the country line a poor proxy for end demand, and the split that would resolve it is not published.
The judgment: the toll is durable and the pricing lever is real, but both are bounded — at one end by mature phone volumes, at the other by licensees' willingness to fund a company that increasingly competes with them. A large licensee disclosing that it moved new design starts to RISC-V, or license revenue falling for several quarters while royalties keep rising, would undercut the first half of that view. So far the evidence shows the subscription formats spreading and the silicon push barely started. It does not yet show that licensees will accept both.
Questions worth following
- Will Arm's own silicon products add revenue without eroding licensees' willingness to pay?
The Arm AGI CPU makes Arm a potential competitor to companies that generate most of its royalties. Whether licensees treat that as additive or threatening decides if the silicon bet expands Arm's profit pool or taxes the licensing relationships that currently fund it.
- Can Arm keep raising the effective royalty rate per chip as mobile unit growth matures?
Royalties are the recurring part of Arm's model, and handset volumes can no longer drive them. Whether Arm can charge more per chip through newer architectures and subsystems sets the growth rate of its largest, most durable revenue stream, and how much cash it has to fund expansion.
Financial quality
Financial quality: cash tracks earnings, licenses set the timing
Financial quality: Strong. Arm licenses processor designs and collects a royalty on each chip shipped. That model produced a 92.5% gross margin in fiscal 2026, liabilities of 23% of assets, and $2.75 billion of cash against $0.06 billion of short-term borrowings. Free cash flow of $3.14 billion over fiscal 2022–2026 slightly exceeded net income of $3.07 billion, so the earnings are backed by cash. Two things keep the grade from being higher: returns are good rather than exceptional given the margin level, and reported profit leans more on the cash pile than on operations.
Five years of high margins, uneven profit
| USD bn unless noted | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 | FY2027 Q1¹ | Assessment |
|---|---|---|---|---|---|---|---|
| Revenue | 2.70 | 2.68 | 3.23 | 4.01 | 4.92 | 1.29 | Strong: 16% four-year CAGR |
| Gross margin % | 91.9 | 92.7 | 92.8 | 94.9 | 92.5 | 97.2 | Exceptional |
| Operating margin % | 25.2 | 25.4 | 3.1 | 20.7 | 18.5 | 7.8 | Adequate; FY2024 hit by IPO-year costs |
| Net margin % | 20.3 | 19.6 | 9.5 | 19.8 | 18.4 | 20.9 | Strong |
| ROE % | 15.5 | 13.8 | 6.5 | 13.1 | 12.0 | 3.2¹ | Adequate |
| ROIC % (est.) | 22 | 21 | 3 | 14 | 13 | n/a | Adequate |
| Operating cash flow / net income (x) | 0.84 | 1.42 | 3.52 | 0.51 | 1.69 | 3.33 | Strong over the cycle, volatile |
| Free cash flow | 0.38 | 0.65 | 0.95 | 0.18 | 0.98 | 0.69 | Strong |
| Capex / revenue % | 3.0 | 3.4 | 4.3 | 5.5 | 11.0 | 16.3 | Rising: expansion |
| Liabilities / assets % | 45.5 | 41.0 | 33.2 | 23.4 | 22.6 | 22.9 | Strong |
| Current ratio (x) | 2.22 | 2.60 | 2.79 | 5.20 | 6.00 | 5.25 | High; some idle liquidity |
| Receivable days | 174 | 159 | 128 | 173 | 183 | 141 | Weak: long, but stable |
¹ Quarter ended June 30, 2026. Flow ratios there are single-quarter figures, not comparable with full years.
Operating margin, not gross margin, is where the cost structure shows. At 18.5% in fiscal 2026 and 20.7% in fiscal 2025, research and staff costs absorb roughly four-fifths of gross profit. Fiscal 2024 shows how fast that bites: revenue rose 21% to $3.23 billion while operating income fell to $0.10 billion in the year containing the September 2023 IPO, cutting net margin to 9.5%. Cost drove that, and the next two years restored a margin near 20%.
Returns land in the adequate-to-strong band. ROE was 12.0% in fiscal 2026 and has ranged from 6.5% to 15.5% over five years, earned with equity at 77% of assets, so borrowing is not flattering the number. An estimate of ROIC, taxing operating income at about 22% and measuring against equity plus debt less cash, gives roughly 13% in fiscal 2026 and 14% in fiscal 2025. That is modest for a 92%-gross-margin business because $1.62 billion of goodwill and $2.75 billion of cash sit in the capital base earning little operationally.
Interest income has grown large enough to shape the reported result. Fiscal 2026 pretax income of about $1.15 billion compared with operating income of $0.91 billion, and in the June 2026 quarter net income of $0.27 billion was nearly triple operating income of $0.10 billion. The gap is consistent with interest earned on a cash balance above $3 billion, plus a small tax benefit in the June quarter. Returns measured on equity therefore include a treasury contribution that says nothing about licensing.
Working capital, not demand, drives the cash line
| USD bn unless noted | FY26 Q1 (Jun-25) | FY26 Q2 (Sep-25) | FY26 Q3 (Dec-25) | FY26 Q4 (Mar-26) | FY27 Q1 (Jun-26) | Assessment | Trend |
|---|---|---|---|---|---|---|---|
| Revenue | 1.05 | 1.14 | 1.24 | 1.49 | 1.29 | Lumpy, license-driven | Rising |
| Net income | 0.13 | 0.24 | 0.22 | 0.31 | 0.27 | Broadly tracks revenue | Rising slowly |
| Total assets | 9.39 | 9.71 | 10.18 | 10.70 | 11.20 | Cash and receivables | Rising |
| Total liabilities | 2.39 | 2.30 | 2.38 | 2.42 | 2.57 | Low; no debt | Leverage falling |
| Fixed assets (net PP&E) | Not disclosed | Not disclosed | Not disclosed | Not disclosed | Not disclosed | Not reported quarterly | n/a |
| Operating cash flow | 0.33 | 0.57 | 0.36 | 0.26 | 0.90 | Working-capital swings | Volatile |
| Operating cash flow / net income (x) | 2.54 | 2.38 | 1.64 | 0.84 | 3.33 | Collection-driven | Volatile |
| Receivables | 2.15 | 1.98 | 2.01 | 2.47 | 2.00 | Built at year-end, released in June | Cyclical within the year |
| Gross margin % | 94.3 | 97.4 | 94.2 | 93.1 | 97.2 | No compression | Stable |
Cash conversion is strong over the cycle and nearly useless quarter to quarter. Operating cash flow ran at 0.51 times net income in fiscal 2025 and 1.69 times in fiscal 2026; the June 2026 quarter's $0.90 billion arrived as receivables fell $0.47 billion, which is collection of earlier billings. Receivable days sit near 150 on a rolling-quarter basis, 183 at the March 2026 year-end, consistent with milestone billing on multi-year licenses. Receivables rose 68% in fiscal 2025 against 24% revenue growth before settling the following year. Deferred revenue, not debt, is the largest liability, so cash moves with signature and billing dates.
The September 2026 quarter repeated the pattern that matters for quality. Arm reported revenue up 27% to $1.44 billion and licensing up 50% to $773 million, while annualized contract value grew 12% and royalties 8%. Licensing revenue is recognized early in an agreement; royalties arrive only after customers finish designs and ship chips. When booked licensing outruns the contract base that far, reported growth is being pulled forward from a stream not yet earned.
Capex is rising for the first time
Spending on equipment, masks and software went from 3.0% of revenue in fiscal 2022 to 11.0% in fiscal 2026 and 16.3% in the June 2026 quarter. Against roughly $0.25 billion of depreciation that is more than 2x, an expansion ratio, and free cash flow fell to 64% of operating cash flow in fiscal 2026 from 88% in fiscal 2023. The cause is Arm's move into selling its own silicon. Reported expectations put initial gross margins on the AGI CPU at 30%–40%, and management has said it expects up to $2 billion of cumulative revenue from it in fiscal 2027 and 2028.
If that volume arrives, the blended gross margin falls toward the 80s, inventory appears on a balance sheet that carries none, and tape-out spending becomes recurring. The first honest test is whether ROIC holds near 13% once that capex and any inventory enter the capital base; a 92%-margin licensing business funding a 35%-margin hardware line would lower group returns even as revenue grows.
What to check
- License bookings against contract value. If annualized contract value keeps growing near 12% while licensing revenue runs at 50%, the mix is shifting toward one-time recognition, and reported growth will slow when signings normalize.
- Receivables and collections. A further rise in days alongside a licensing-heavy quarter would suggest terms are stretching to support signings.
- Goodwill of $1.62 billion. It has not moved in five years and equals about 20% of equity. A silicon program that misses its volume plan is the most plausible trigger for a write-down.
- The non-operating share of profit. If interest income keeps supplying a fifth of pretax profit after short rates fall, the operating margin is thinner than the net margin implies.
Questions worth following
- Does Arm's license revenue convert into cash on schedule?
License revenue is recognized early in an agreement, so cash arrives through milestone billings that Arm must collect. If collections keep pace, the premium rating rests on repeatable cash; if receivables outgrow revenue, reported growth is being financed with longer terms.
- Will Arm's move into selling its own silicon materially raise capital intensity and weaken free cash flow conversion?
The gap between operating and free cash flow widened to $0.54bn in fiscal 2026 from $0.22bn a year earlier, and Arm introduced the Arm AGI CPU in March 2026. If that transition consumes cash the way licensing never did, the returns on a much larger capital base fall.
- Does interest income make Arm's P/E look cheaper than it is?
Net income of $0.27 billion exceeded operating income of $0.10 billion in the June 2026 quarter as interest on $3.06 billion of cash and a tax credit did part of the work. On operating earnings the multiple exceeds 290x, so the headline number understates what buyers pay for the core business.
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Key risksWhat could break the investment case
- Valuation and margin of safetyWhat the current price assumes, and how much room for error it leaves
- SummaryOur overall judgement and what to watch next
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on Arm Holdings plc American Depositary Shares
Arm Holdings plc American Depositary Shares
Business, moat and key questions
Open company page