How ARM Holdings Makes its Money: Revenue Breakdown

How Does ARM Make its Money?

ARM Holdings plc (NASDAQ: ARM) designs the processor architectures and CPU core designs that power virtually every smartphone, most tablets, and a rapidly growing share of laptops, data centre servers, automotive systems, and IoT devices. The company generated $3.94 billion in total revenue for fiscal year 2025 (ending March 2025), up 39.2% from $2.83 billion in FY2024, with net income of $0.63 billion and an 85.3% gross margin — one of the highest gross margins of any company in the semiconductor ecosystem.

ARM makes money in a fundamentally different way from semiconductor companies like Qualcomm, Intel, or NVIDIA: ARM does not manufacture chips, does not sell chips, and does not design complete chips. Instead, ARM designs the fundamental blueprint — the instruction set architecture (ISA) and physical CPU core implementations — and licenses those designs to chip companies who then manufacture, sell, and support the resulting products. ARM earns revenue through two streams: Licensing fees (upfront payments for the right to use ARM’s IP in new chip designs) and Royalties (a per-chip fee on every ARM-based chip shipped to end customers, typically for the lifetime of the product).

Over 280 billion ARM-based chips have been shipped cumulatively, and ARM’s architecture is estimated to power approximately 99% of the world’s smartphones and a growing portion of cloud server infrastructure. The business model is structurally similar to a tollbooth on the semiconductor highway: every time a chip based on ARM’s architecture is designed or shipped, ARM collects a fee. The chipmakers bear all the design customisation, manufacturing, validation, and sales costs; ARM collects a royalty on their success.

Key Takeaways

• ARM generated $3.94B in FY2025 revenue (+39.2%) split between Royalty revenue ($2.33B, 59%) and License & Other ($1.33B, 41%); gross margin of 85.3% reflects near-zero marginal cost of distributing intellectual property — once a CPU core design is created, it can be licensed to hundreds of companies at minimal incremental cost
• Royalties are the high-quality recurring revenue stream: ARM earns a per-chip royalty (typically 0.5–5% of chip selling price, or a fixed cent-per-chip rate) on every ARM-based chip shipped; royalty revenue grows with both chip volume (more devices shipped globally) and royalty rate per chip (the Armv9 uplift)
• The Armv9 transition is ARM’s primary earnings growth lever: Armv9-based chips carry approximately 2x the royalty rate of the older Armv8 architecture; as the installed base of shipping chips shifts from Armv8 to Armv9 (a 3–5 year transition), ARM’s royalty revenue per unit shipped doubles without requiring any additional volume growth — this royalty rate uplift drove much of the +23.3% royalty revenue growth in FY2025
• Licensing revenue ($1.33B, +41.5%) comes from upfront IP access fees — most importantly ARM Total Access (ATA) and Flexible Access subscription programmes where customers pay annual fees for broad IP portfolio access rather than per-design licensing; ATA has 30+ customers including AMD, Intel, Samsung, and most of the world’s major chip companies; this subscription model converts lumpy per-design license deals into more predictable recurring revenue
• Data centre penetration is the biggest total addressable market expansion: ARM-based server CPUs — AWS Graviton4, Microsoft Cobalt 100, Google Axion, NVIDIA Grace Blackwell — are taking market share from Intel x86 in cloud computing; every percentage point of global server CPU market share gained from Intel directly adds royalty revenue; the server CPU market alone is approximately $20B annually at the silicon level, representing an enormous royalty expansion opportunity
• Custom silicon is a structural tailwind and a strategic risk: Apple (A-series/M-series), Amazon (Graviton), Google (Axion), Microsoft (Cobalt), and Qualcomm (Oryon) all design their own custom ARM-based chips rather than using ARM’s standard “off-the-shelf” Cortex cores; custom silicon customers typically negotiate lower per-chip royalty rates in exchange for greater design investment — this grows ARM’s total ecosystem but compresses per-unit royalty rates at the high-volume end
• RISC-V is the long-term architectural alternative: RISC-V is an open-source processor instruction set that any company can use without paying licensing fees to ARM; adoption in embedded/IoT applications is growing; the question is whether RISC-V can mature into a viable competitor at the high-performance computing tier (smartphones, servers, laptops) where ARM currently faces no architectural competition
• SoftBank owns ~90% of ARM following the September 2023 IPO (which floated approximately 10% of shares); SoftBank’s majority ownership means ARM’s strategic direction — licensing terms, M&A, capital allocation — is ultimately controlled by SoftBank’s Masayoshi Son; investors in the public float are minority shareholders in a SoftBank-controlled entity

ARM Holdings (ARM) Business Model

ARM’s business model is a pure intellectual property licensing company — the closest analogy in the semiconductor industry is not a chipmaker but rather an EDA software company like Cadence Design Systems or Synopsys, which also earn recurring fees for providing design tools that semiconductor companies cannot build chips without. The critical difference: ARM’s IP is embedded in every chip shipped, generating a royalty that continues for the life of the product.

The Licensing Revenue Stream: Paying for Access

Licensing revenue represents the upfront fees companies pay for the right to use ARM’s IP in new chip designs. This revenue is earned before any chips are manufactured or shipped — it is the “design phase” revenue that triggers when a chip company decides to build a new processor based on ARM’s architecture.

Types of licensing arrangements:

Per-design license (traditional model): A chip company pays a negotiated upfront fee (typically $1–30M+ depending on the IP scope) to license a specific ARM CPU core (e.g., Cortex-A720, Cortex-X925, Neoverse V3) for a specific chip design. This fee covers the right to implement the ARM core in that chip. The fee is negotiated and paid once; the royalty on shipped chips flows separately. Per-design licensing is lumpy — it depends on when chip companies start new design projects — and can vary significantly quarter to quarter.

ARM Total Access (ATA) — the subscription model: ARM Total Access is a subscription programme that gives licensees access to ARM’s complete current portfolio of CPU cores, GPU cores, and system IP for a single annual fee. Rather than paying per design, ATA customers pay a recurring annual subscription and can use any ARM IP for any number of chip designs during the subscription period. ATA customers include the world’s largest semiconductor companies (AMD, Intel, Samsung, MediaTek, and others) — companies that design many chips across many product lines and benefit from broad, flexible IP access rather than per-design negotiations.

ATA is strategically important for ARM because it converts lumpy project-by-project license revenue into predictable annual subscription revenue. It is the primary driver of the +41.5% license revenue growth in FY2025 as more large customers adopted the subscription model. The trade-off: ATA customers may pay lower total fees per chip design than a traditional per-design licensee, but ARM gains revenue predictability and often expands its share of each customer’s total chip portfolio.

Flexible Access: A lighter-weight programme for companies exploring ARM IP for prototyping and evaluation before committing to full ATA or per-design licenses. Used by emerging chip design companies and startups in adjacent markets (automotive, IoT).

Architecture licenses (custom silicon): A small number of highly sophisticated chip designers — Apple, Qualcomm, Amazon, Google, Microsoft — license ARM’s instruction set architecture (ISA) rather than ARM’s specific CPU core implementations. These customers design their own CPU cores from scratch (Apple’s Firestorm/Everest, Qualcomm’s Oryon) that are compatible with the ARM ISA (meaning they can run all ARM software) but built from custom transistor-level designs optimised for each company’s specific power, performance, and area targets. Architecture license fees are typically higher than per-design licenses (due to the breadth of IP access granted) but royalty rates for custom silicon are generally lower per chip shipped, as these customers contribute more of their own design investment.

The Royalty Revenue Stream: Earning on Every Chip Shipped

Royalty revenue is ARM’s most structurally valuable revenue stream: recurring, growing with global semiconductor volumes, and improving in per-unit rate as chips shift to newer architectures. ARM earns royalties under two main rate structures:

Percentage-of-price royalty: ARM earns a percentage of the chip selling price — typically ranging from 0.5% to 5% depending on the value of ARM’s IP contribution, the architecture version, and the negotiated terms. A $100 application processor (like the Qualcomm Snapdragon flagship) at a 1.5% royalty rate generates $1.50 per chip for ARM. A $5 microcontroller at 1% generates $0.05. The royalty rate is higher for more valuable, complex chips using more advanced ARM IP.

Fixed per-unit royalty: Some older or lower-cost chip agreements use a fixed rate per chip shipped (e.g., $0.05 or $0.10 per unit) rather than a percentage of price. As chip prices change, fixed-rate royalties do not capture the upside — percentage-of-price royalties are strategically preferable for ARM.

The Armv9 royalty rate uplift: This is the most important mechanical driver of ARM’s near-term royalty revenue growth. ARM’s Armv9 architecture — used in high-end smartphone chips (Apple A18, Qualcomm Snapdragon 8 Gen 3/4, MediaTek Dimensity 9300+) and increasingly in Neoverse server chips — carries approximately 2x the royalty rate of the preceding Armv8 architecture, which dominated smartphone chips for over a decade. As new chip designs shift to Armv9 and as Armv8 chips age out of production:

• Today: approximately 25–35% of shipped ARM chips are Armv9-based (high-end and mid-range smartphones, some servers)
• In 3–5 years: Armv9 becomes the default architecture for all performance-tier chips; Armv8 retreats to legacy embedded/IoT use

Each percentage point shift in the shipped chip mix from Armv8 to Armv9 increases ARM’s royalty revenue per chip shipped by approximately 2x the magnitude of the shift. This is a multi-year structural royalty tailwind that does not depend on semiconductor volume growth — it works even in a flat or declining unit environment.

The Data Centre Opportunity: ARM vs. Intel x86

The most significant total addressable market expansion for ARM is in cloud server CPUs — a market historically dominated by Intel’s x86 architecture with over 90% share as recently as 2020.

ARM-based server CPUs gaining traction:

• AWS Graviton (Amazon): Custom ARM Neoverse-based chips now power approximately 30%+ of AWS EC2 instances; Graviton4 (released 2024) delivers competitive performance at 40–60% better price/performance vs. x86 instances on many workloads
• Microsoft Cobalt 100: ARM Neoverse V2-based; deployed across Azure for first-party workloads
• Google Axion: ARM Neoverse V2-based; deployed in Google Cloud data centres
• NVIDIA Grace Blackwell: NVIDIA’s Grace CPU (ARM Neoverse-based) pairs with Blackwell GPUs in the Grace-Blackwell superchip; as NVIDIA’s AI infrastructure dominates hyperscaler capital expenditure, every Grace Blackwell unit shipped generates an ARM royalty on the CPU component
• Ampere Altra/AmpereOne: Arm-based server CPUs for enterprise cloud deployment

The server CPU market is approximately $20–25 billion annually at the silicon level. If ARM-based server chips reach 20–25% market share by 2026–2028 (from ~5–10% today), this represents billions of additional royalty revenue — and at higher ASPs (server CPUs sell for $200–1,000+ each vs. smartphone chips at $20–100), the per-chip royalty dollar amount is significantly larger than mobile.

Custom Silicon: The Double-Edged Sword

The trend toward custom silicon — where hyperscalers and consumer electronics companies design their own proprietary chips rather than buying off-the-shelf chips from Qualcomm, MediaTek, or Intel — is simultaneously ARM’s largest growth catalyst and its pricing power constraint.

Why custom silicon grows ARM’s market: Every hyperscaler that designs its own ARM-based chip (Amazon Graviton, Google Axion, Microsoft Cobalt, Apple A-series) creates a new ARM royalty relationship that didn’t exist when those companies bought Intel or Qualcomm chips. Apple, for example, pays ARM a royalty on every iPhone and Mac sold — a relationship that generates ~$300–400M+ in annual royalty revenue for ARM from a single customer.

Why custom silicon constrains ARM’s per-unit royalty rate: Architecture licensees (Apple, Qualcomm, Amazon) who design their own CPU cores from scratch negotiate lower per-chip royalty rates than companies using ARM’s standard Cortex cores. Their argument: they contribute the majority of the chip design investment and use only ARM’s ISA compatibility rather than ARM’s actual core implementations. ARM accepts lower per-chip rates to maintain these customers in the ARM ecosystem — losing Apple or Amazon to a competing architecture would be an existential threat to the smartphone and cloud royalty bases.

The net effect: custom silicon grows the volume of ARM-royalty-bearing chips but at lower per-unit rates than standard Cortex-based designs. The Armv9 uplift partially counteracts this rate dilution.

ARM Holdings Competitors

Qualcomm — largest ARM licensee and emerging competitive threat

Qualcomm is the largest single payer of ARM royalties — Qualcomm’s Snapdragon chips (used in Android smartphones, laptops, and automotive systems) are all ARM-based, generating hundreds of millions in annual royalties for ARM. The relationship is simultaneously ARM’s largest revenue source and its most contentious: following Qualcomm’s acquisition of Nuvia in 2021 (a startup founded by former Apple CPU architects), Qualcomm developed the Oryon CPU — a custom ARM architecture-licensed core that Qualcomm claims can be used across its entire product line without per-chip royalty obligations under the Nuvia-era architectural license. ARM disagrees and has sued Qualcomm, claiming Nuvia’s architectural license terminated upon acquisition. This litigation, if resolved in ARM’s favour, would require Qualcomm to pay significantly higher royalties on Snapdragon X Elite (laptop) and future chips using Oryon; if resolved in Qualcomm’s favour, ARM loses royalty revenue from one of its highest-volume customers. This is the most material pending legal risk in ARM’s business.

NVIDIA — the AI infrastructure partner

NVIDIA is both a significant ARM licensee (NVIDIA’s Grace CPU in the Grace-Blackwell superchip is ARM Neoverse-based) and a former would-be acquirer — NVIDIA’s attempted $40B acquisition of ARM from SoftBank in 2020–2021 was blocked by regulators. Today NVIDIA generates ARM royalties through Grace CPU shipments in AI data centres; as NVIDIA’s AI GPU infrastructure spending grows, ARM’s royalty revenue from Grace scales proportionally. NVIDIA is not a competitive threat to ARM’s IP licensing business — it is an ecosystem partner and royalty customer.

Cadence Design Systems — the closest business model comparison

Cadence Design Systems is not a semiconductor company but the closest structural analogy to ARM’s business: a software and IP company that earns recurring fees (software subscriptions + IP royalties) from chip designers who cannot design chips without Cadence’s EDA tools and design IP. Like ARM, Cadence operates a high-gross-margin IP and software business with strong switching costs — once a chip design is built using Cadence’s tools and verified against Cadence’s IP libraries, switching to a competitor’s tools requires revalidating the entire design. Both ARM and Cadence are “picks and shovels” businesses in the semiconductor gold rush — they earn regardless of which specific chip company wins in the end market.

RISC-V — the open-source architectural alternative

RISC-V is an open, royalty-free processor instruction set architecture that any company or individual can implement without paying licensing fees. The RISC-V International organisation governs the specification; SiFive, Andes Technology, Western Digital (in storage controllers), and a growing number of startups and Chinese chip companies are building RISC-V-based chips. RISC-V adoption has been fastest in: embedded/IoT microcontrollers (where ARM Cortex-M cores face price competition), Chinese chip design (where geopolitical restrictions limit access to ARM architecture licenses), and research/academic processors. The question for ARM investors: can RISC-V mature into a competitive alternative for high-performance applications — smartphones, laptops, servers — where ARM currently faces no viable architectural competition? The consensus view is that RISC-V’s software ecosystem (applications, operating system support, toolchain maturity) remains years behind ARM’s ecosystem, making near-term substitution at the high-performance tier unlikely but a long-term structural risk.

Intel x86 — the server CPU incumbent ARM is displacing

Intel is not a competitor for ARM’s licensing business — Intel designs and manufactures its own chips using its proprietary x86 architecture. But Intel is the primary incumbent that ARM-based server CPUs are displacing in the data centre, making Intel’s competitive response (continued x86 performance improvements, Intel Foundry manufacturing services) the primary factor determining how fast ARM gains data centre royalty revenue. Intel’s competitive struggle — facing cost, yield, and architecture challenges — has accelerated the window for ARM-based server CPU adoption.

Revenue Breakdown

Financial data sourced from ARM Holdings SEC Filings.

Royalty growing 23.3% reflects the Armv9 rate uplift (higher royalty per chip shipped even on similar unit volumes) plus volume growth from data centre chip shipments. Licence growing 41.8% reflects ATA subscription adoption — more customers on annual subscription programs pay license fees more consistently than under traditional per-design models, increasing license revenue predictability. The combined 39.4% total revenue growth is exceptional for any company, let alone one at ~$4B in revenue — it reflects both the cyclical upturn in semiconductor demand post-2023 inventory correction and ARM-specific structural growth from architecture transition and data centre penetration.

Revenue Trend (3-Year)

FY2024 was the trough year: a semiconductor industry inventory correction reduced chip shipments broadly, suppressing ARM royalty volumes; an impairment charge on a prior-year acquisition contributed to the net loss. FY2025 represents the full recovery plus structural growth: Armv9 transition ramping, data centre ARM CPUs beginning to scale, and licence subscription adoption increasing revenue predictability. The gross margin expansion (80% → 85.3%) reflects IP licensing revenue growing faster than the modest cost of revenue (primarily IP amortisation and engineering support for licensees).

ARM Holdings (ARM) Income Statement

Financial data sourced from ARM Holdings SEC Filings.

The most notable income statement observation: ARM’s gross margin (85.3%) far exceeds its operating margin (16.2%) — the gap represents $2.7 billion in annual operating expenses, primarily R&D ($1.87B, ~47% of revenue). ARM must continuously invest in next-generation CPU core designs (Cortex-A, Cortex-X, Neoverse), GPU and NPU architectures, physical IP libraries, and software tooling to maintain its architectural leadership. If ARM stopped investing in R&D, its existing IP would generate excellent profits for several years — but competitors would eventually erode its architectural advantage. The R&D investment is the economic moat maintenance cost.

ARM Holdings (ARM) Key Financial Metrics

• Gross Margin: 85.3% — Among the highest gross margins in the semiconductor industry, reflecting the near-zero marginal cost of IP distribution. Once an ARM CPU core design is created (requiring thousands of engineer-years of investment), the cost of licensing that design to the 100th customer is essentially the cost of contractual administration and engineering support — not a proportional variable cost. This is the structural advantage of IP licensing vs. chip manufacturing: TSMC’s gross margin is ~53%; Qualcomm’s is ~56%; ARM’s is 85%

• Operating Margin: 16.2% — The gap between 85.3% gross margin and 16.2% operating margin is R&D ($1.87B, 47% of revenue). ARM’s operating margin is structurally lower than its gross margin would imply because the company must run a continuous multi-year R&D programme to stay architecturally relevant. As revenue scales faster than R&D costs, operating margins should expand toward 30–40%+ over a multi-year horizon — the same operating leverage dynamic seen in software and EDA companies as they scale

• Royalty Revenue per Chip: Not directly disclosed but calculable from volume data. ARM has stated approximately 7.5–9 billion chips were shipped in recent fiscal years; $2.33B in royalty revenue on ~8.5B chips implies approximately $0.27 per chip blended royalty rate. The Armv9 transition is lifting this blended rate — Armv9 chips at the high end (smartphone flagships, server CPUs) carry royalty rates of $0.50–2.00+ per chip; legacy Armv8 embedded chips carry $0.01–0.10 per chip; the blended rate improvement is the core thesis

• Free Cash Flow: ~$1.0B — ARM generated approximately $1B in FCF in FY2025 despite significant R&D investment, reflecting the cash-generative nature of the IP licensing model (no capex, no inventory, no manufacturing costs). FCF should grow proportionally with operating income expansion as revenue scales

• Return on Invested Capital: ARM’s IP creates compounding returns — the R&D investment in Armv9 cores, once deployed to 280+ licensees, generates royalties on billions of chips annually without ongoing variable costs. The ROIC on ARM’s cumulative R&D investment is extremely high, justifying continued R&D spending as a capital allocation priority

• Stock-Based Compensation: ARM’s G&A expense spike in FY2025 reflects IPO-related SBC and public company costs; SBC is a meaningful portion of operating expenses and should be monitored as it dilutes existing shareholders. ARM’s approximately 7,500 engineers are expensive talent competing for compensation against hyperscaler and chip company engineering roles

Is ARM Holdings Profitable?

Yes — ARM reported net income of $634 million in FY2025 on $3.94 billion in revenue (16.1% net margin). This represents a significant reversal from the -$37 million net loss in FY2024, which was impacted by a prior-period acquisition impairment charge and the semiconductor cycle trough. On a free cash flow basis, ARM generated approximately $1 billion in FY2025 — the most relevant measure of the IP licensing business’s underlying cash generation.

ARM’s profitability path is structurally compelling: gross margins of 85.3% mean that each additional dollar of revenue contributes approximately $0.85 to gross profit; as revenue grows faster than R&D (which scales more slowly than revenue once the core architecture investments are made), operating margins expand automatically. At $5B+ in revenue — likely within 2–3 years on current growth trajectory — ARM’s operating margin should approach 25–30%, generating $1.5–2B+ in annual net income on the current cost structure.

ARM Holdings (ARM): What to Watch

1. Armv9 royalty rate mix shift — This is the single most important quarterly metric for ARM investors: the percentage of chips shipped that use Armv9 (vs. older Armv8/Armv7). Management discloses Armv9 royalty revenue as a percentage of total royalty revenue — watch this figure each quarter. As Armv9 grows from ~25–35% of royalty revenue toward 50–60%+, the blended royalty rate per chip shipped doubles for the transitioned portion. Any acceleration in Armv9 adoption (driven by smartphone mid-range and server chip proliferation) would be the most positive signal for near-term royalty revenue growth

2. Data centre royalty revenue trajectory — ARM’s data centre penetration is the largest TAM expansion in the company’s history. Watch for management disclosure of infrastructure royalty revenue growth rates or data centre chip volume commentary. AWS Graviton4, Microsoft Cobalt, Google Axion, and NVIDIA Grace Blackwell shipments are all growing; as AI data centre capital expenditure expands globally, the Grace Blackwell chip specifically (NVIDIA’s flagship AI chip with an ARM-based CPU component) becomes a meaningful ARM royalty contributor. Any data centre royalty disclosure showing growth above 40–50% annually would validate the server thesis ahead of schedule

3. Qualcomm litigation outcome — The ARM vs. Qualcomm lawsuit over the Nuvia architecture license is the most significant legal risk in ARM’s business. A ruling in ARM’s favour could require Qualcomm to renegotiate royalty terms on all Oryon-based chips (Snapdragon X Elite laptops, future Snapdragon smartphone chips using custom cores) — potentially adding hundreds of millions in annual royalty revenue. A ruling in Qualcomm’s favour would establish that architecture license rights transfer through acquisitions, potentially creating a precedent that other large licensees could use to negotiate reduced royalty rates. Watch for any court ruling or settlement announcement

4. RISC-V adoption in addressable markets — RISC-V is not currently a competitive threat in smartphones, laptops, or servers — the software ecosystem maturity gap is too large for near-term substitution. But watch RISC-V adoption in: (a) the Chinese chip market, where US export restrictions are accelerating RISC-V investment as Chinese companies seek ARM-independent architectures; (b) embedded/automotive microcontrollers, where RISC-V’s simplicity and royalty-free nature are competitive; (c) any major Silicon Valley hyperscaler announcing a RISC-V-based custom CPU programme would be a significant signal that the long-term architecture competition is intensifying

5. ARM Total Access subscription adoption — ATA customers on annual subscriptions represent the highest-quality licence revenue: predictable, recurring, and evidence of deep ecosystem commitment. Watch ATA customer count (currently 30+) and total ATA annual contract value disclosures. If ATA adoption stalls or customers revert to per-design licensing (which would reduce licence revenue predictability), it would signal ARM’s subscription transition is hitting limits. Continued ATA growth toward 50+ customers would represent a structural improvement in revenue quality

6. Operating margin expansion trajectory — ARM’s 16.2% operating margin in FY2025 is significantly below its long-term potential given 85.3% gross margins. R&D at 47% of revenue is the primary compression factor. Watch whether R&D grows in line with revenue (holding margin flat) or at a slower rate (driving margin expansion). Management has guided toward expanding operating margins over time; any FY2026 guidance implying operating margins above 20% would validate the operating leverage thesis. The investment question: at what scale does ARM’s R&D growth decelerate relative to revenue, and is that deceleration sustainable or does it require reducing innovation pace?

7. SoftBank’s stake and related-party dynamics — SoftBank owns approximately 90% of ARM and has significant influence over licensing terms, strategic decisions, and capital allocation. Watch for: (a) any SoftBank secondary share sales that increase ARM’s public float (beneficial for liquidity and index inclusion); (b) any strategic decisions that appear to benefit SoftBank’s broader portfolio (e.g., preferential licensing terms for SoftBank portfolio companies); (c) the potential for a larger ARM float or index inclusion if SoftBank reduces its stake, which could shift ARM’s investor base and valuation multiple

ARM Holdings (ARM) Financial Summary

ARM Holdings (ARM) generated $3.94 billion in total revenue in fiscal year 2025 (+39.4% YoY) with $634 million in net income and an industry-leading 85.3% gross margin — the result of a pure intellectual property licensing model where ARM collects royalties on the over 280 billion ARM-based chips shipped cumulatively across smartphones, servers, laptops, automotive systems, and IoT devices. The Armv9 royalty rate uplift (2x vs. Armv8), data centre ARM CPU penetration (AWS Graviton, NVIDIA Grace, Microsoft Cobalt, Google Axion), and ARM Total Access subscription adoption are the three structural growth drivers converging simultaneously. ARM’s competitive position — the only viable high-performance processor architecture for smartphones and a rapidly growing force in data centre CPUs — is among the most durable moats in the semiconductor industry. For ARM’s largest royalty customer and pending litigation context, see How Qualcomm Makes its Money. For the AI data centre chip that generates ARM Grace royalties, see How NVIDIA Makes its Money.