Arm Holdings (ARM): AI Royalty Growth vs. Rich Valuation

Thesis

Arm Holdings plc American Depositary Shares(ARM) is one of the strongest strategic assets in semiconductors and one of the most demanding stocks in the market at the same time. The bullish case is straightforward: ARM sits at the center of modern compute through a licensing and royalty model that scales across smartphones, cloud, automotive, IoT, and now AI infrastructure. Revenue grew 26.3% YoY to $4.67B, gross margin reached 97.5%, and management is showing that higher-value products like Armv9 and Compute Subsystems, or CSS, are lifting royalty per chip even when end-unit growth is uneven.

The medium-term opportunity is real. ARM is moving from being the quiet plumbing of mobile chips to becoming a richer monetization layer across AI servers, edge devices, robotics, and software-defined vehicles. Management said data center royalty revenue grew more than 100% YoY, CSS licenses reached 21 across 12 companies, and hyperscaler share is expected to approach 50%. That is not cosmetic growth. That is a change in where ARM captures value.

The problem is price. With a trailing P/E of 222.3x, forward P/E of 82.0x, EV/revenue of 37.2x, and free cash flow yield of just 0.35%, the stock already discounts a large part of the AI upside. The business looks elite. The stock looks expensive. For a balanced, moderate-risk investor with a medium-term horizon, that usually leads to one conclusion: respect the company, but demand discipline on entry.

The base-case view is that ARM deserves a premium multiple because of its asset-light model, ecosystem moat, and rising exposure to AI compute. But it does not deserve a blank check. The right stance is Buy on pullbacks, not chase at any price. If execution continues, the business can grow into a higher valuation over time. If growth merely stays good instead of spectacular, multiple compression becomes the main risk. In markets, a great engine can still be attached to an overpriced car.

Company Overview

Arm Holdings plc American Depositary Shares(ARM) architects and licenses CPU designs, related IP, system IP, graphics technologies, software, and development tools. Unlike Nvidia(NVDA), Advanced Micro Devices(AMD), or Intel(INTC), ARM does not primarily manufacture and sell chips. It licenses the architecture and collects royalties when customers ship products built on that IP. That distinction matters because it produces unusually high gross margins and broad exposure to semiconductor growth without the same capital intensity as a fab or foundry model.

The company was founded in 1990, is headquartered in Cambridge, U.K., and trades on NASDAQ under ARM. It employs 8,330 people and remains controlled by SoftBank. ARM completed its IPO in September 2023, but SoftBank still owns the overwhelming majority of shares, leaving a relatively small public float. That helps explain part of the stock’s volatility. When float is tight and narrative is hot, price action can behave less like a weighing machine and more like a voting machine with too much coffee.

ARM generated $4.01B of revenue in fiscal 2025, up from $3.23B in fiscal 2024. Revenue is split between royalty and license streams. In fiscal 2025, royalty revenue was $2.17B, or 54.1% of total revenue, while license and other revenue was $1.84B, or 45.9%. That mix gives ARM both recurring monetization from shipped chips and upfront monetization from new design wins and broader access agreements.

The company has reorganized its strategic framing around three AI-centric business units: Edge AI, Cloud AI, and Physical AI. This is more than branding polish. It reflects where demand is moving. ARM’s architecture already dominates mobile and embedded compute. The next leg is monetizing that installed base more deeply while gaining share in data center and automotive, where the royalty pool per chip is materially larger.

Business Segment Deep Dive

ARM does not report classic operating segments in the same way many semiconductor peers do. The clearest financial split is by revenue type: royalty and license. That is the right place to start because it tells investors how the machine earns money.

Royalty revenue is the core annuity stream. It reached $2.17B in fiscal 2025, up from $1.80B in fiscal 2024. In the most recent reported quarter, royalty revenue was $737M, up 27% YoY. This line benefits from three forces: more end devices shipping with ARM-based chips, higher royalty rates from newer architectures like Armv9, and richer content per chip through CSS and more complex compute platforms.

License and other revenue was $1.84B in fiscal 2025 versus $1.43B in fiscal 2024. In Q3 fiscal 2026, license revenue was $505M, up 25% YoY. This line is lumpier because it depends on the timing and size of large agreements. Management pointed to strong demand for next-generation architectures and deeper strategic engagements. It also disclosed that $200M of quarterly license revenue came from SoftBank-related technology licensing and design services, which is meaningful and worth watching.

That SoftBank contribution cuts both ways. On one hand, it supports near-term revenue and may reflect durable roadmap work tied to AI compute ambitions. On the other hand, it introduces a concentration question. When a parent company becomes a large customer, investors should separate durable demand from financial choreography. Management insists the revenue is durable, but the market will likely keep applying a small discount until that durability is proven over several periods.

By end market, mobile remains the largest business today, but the growth story is shifting. ARM’s investor materials show mobile consumer electronics share above 99%, cloud applications share rising from 9% to 20%, automotive from 36% to 44%, and other infrastructure from 44% to 50%. That is the key transition. Mobile is the base. Cloud and automotive are the torque.

Management also noted that CSS was approaching double-digit royalty mix last year, is now in the teens, and could move toward 50% over the next couple of years. If that happens, ARM’s monetization profile changes materially because CSS carries higher value, shortens customer design cycles, and embeds ARM more deeply into system-level decisions rather than just core IP selection.

Flagship Product Analysis

ARM’s flagship product is not a single chip sitting on a shelf. It is the architecture platform itself, increasingly led by Armv9 and CSS. Armv9 matters because it raises royalty rates versus prior generations. CSS matters because it packages more of the subsystem, reduces integration work for customers, and gives ARM a larger economic claim on each chip shipped.

In smartphones, ARM said the top four Android vendors are already shipping CSS-powered devices. That matters because mobile is still the largest royalty base, and richer content in a mature market is often more valuable than unit growth alone. Management was clear that each new smartphone CSS cycle generally comes with higher royalty rates year over year. In plain English, ARM is finding ways to get paid more even if the handset market does not suddenly become exciting again.

In cloud and AI infrastructure, the flagship product family is Neoverse and related server CPU platforms. ARM highlighted AWS Graviton5, Google Axion, Microsoft Cobalt 200, and Nvidia’s Vera CPU as examples of rising adoption. These are not side projects. They are proof that ARM is becoming a serious architecture layer in hyperscale compute, where power efficiency and core density matter more each year.

The strategic significance of CSS is especially high because it changes ARM’s role from ingredient supplier to partial system architect. Customers adopt CSS to cut cycle time roughly in half, according to management commentary. That is a strong value proposition in a world where advanced nodes are expensive, design complexity is rising, and missing a product window can destroy returns. When a product saves time, reduces risk, and improves economics, customers tend to come back. Management said every CSS customer eligible to renew for the next generation has done so.

ARM also launched its AGI CPU in March 2026 with Meta Platforms(META) collaboration cited in external context. That move suggests ARM is willing to push further into platform-level influence in the data center. Investors should watch carefully. If ARM can expand from architecture licensing into broader silicon and subsystem relevance without alienating partners, the upside is meaningful. If it oversteps, channel conflict becomes a real risk.

Innovation & Competitive Advantage

ARM’s moat rests on four pillars: ecosystem scale, power efficiency, design stickiness, and monetization upgrades. The ecosystem is enormous, with more than 22M developers and over 325B chips shipped since inception. That gives ARM a software and tooling advantage that is hard to replicate. In semiconductors, architecture adoption is not just about raw performance. It is about whether the whole stack works, from tools to libraries to developer familiarity.

Power efficiency remains ARM’s most durable technical edge. As AI workloads spread from cloud to edge and physical systems, efficient compute becomes more valuable, not less. Data centers care because power is now a design constraint. Smartphones care because battery life still matters. Cars and robots care because thermal budgets, reliability, and real-time behavior are unforgiving. ARM’s architecture fits all three.

The third advantage is stickiness. Once a customer designs around ARM IP, switching is expensive. Software stacks, validation, toolchains, and downstream product roadmaps all get built around the architecture. That does not make ARM invincible, but it does make displacement slow and costly. This is why RISC-V is a real threat over time but not an overnight wrecking ball.

The fourth advantage is monetization. Armv9 and CSS allow ARM to increase revenue per chip and per design win. That is the subtle but crucial shift. The company is not relying only on more chips shipping. It is improving the toll booth. Management said CSS could become up to 50% of royalty revenue over the next couple of years. If that plays out, ARM’s royalty engine becomes structurally more profitable and less dependent on mature mobile unit growth.

The main innovation risk is execution breadth. ARM is investing heavily in next-generation architectures, chiplets, and complete SoCs. That can expand opportunity, but it also raises complexity and partner sensitivity. Some customers like ARM best when it sells the map, not when it starts driving the car.

Operations & Supply Chain

ARM’s operating model is unusually attractive for a semiconductor company because it is asset-light. It does not need to build fabs, carry large inventories, or manage the same manufacturing swings as integrated device makers. That supports gross margins near 98% on a non-GAAP basis and 97.5% on the provided profitability data. Few businesses in semiconductors produce software-like margins with hardware-like strategic importance.

That said, ARM is not immune to supply chain issues. Its royalty stream depends on customers shipping chips and devices. If memory shortages or foundry bottlenecks reduce unit volumes, ARM feels it indirectly. Management addressed this directly on the earnings call, noting that a 20% reduction in smartphone units would likely translate to only a 2% to 4% impact on smartphone royalties and roughly 1% to 2% on total royalties. That is a useful reminder that ARM’s exposure is cushioned by mix and pricing, not eliminated.

The company is also increasing R&D investment aggressively. Non-GAAP operating expenses in Q3 fiscal 2026 rose 37% YoY to $716M. This reflects engineering headcount growth and spending on architectures, CSS, chiplets, and SoC exploration. For an IP company, talent is the factory. Rising R&D is not automatically a red flag. It is the cost of staying essential.

Operationally, the biggest external dependency is customer execution. ARM can sign licenses, but royalties arrive only when customer chips ramp successfully. That creates a lag between design wins and financial payoff. It also means ARM’s quarterly results can be influenced by customer product timing, not just ARM’s own performance. Investors need patience with that rhythm.

Market Analysis

ARM sits in several semiconductor markets at once, but the most important medium-term growth pools are cloud AI, automotive, and edge AI. The broader semiconductor market remains healthy, with industry estimates placing 2025 revenue around $700B to $800B and continued growth into 2026. AI is the dominant demand driver, with data center chips and advanced packaging pulling the cycle forward.

ARM’s own TAM framing is useful because it focuses on chips that can contain a processor. It estimated TAM at $202.5B in 2022 and $246.6B by 2025, with the fastest growth in cloud compute and automotive. Cloud compute was projected to grow from $17.9B to $28.4B, while automotive was projected from $18.8B to $29.1B. Those are exactly the markets where ARM’s share gains are now becoming visible.

In mobile, ARM is already dominant, so the opportunity is monetization rather than share gain. In cloud, the reverse is true. Share gains can still be large, and each win carries higher royalty potential. Management said data center revenue is now likely in the teens as a share of business and could become as large as or larger than smartphones in two to three years. If that happens, ARM’s growth mix improves materially because server and infrastructure silicon tends to carry richer economics.

Automotive and physical AI are the quieter but important growth vectors. ARM highlighted Rivian(RIVN), Tesla(TSLA) Optimus-related platforms, Nvidia(NVDA) Jetson and Thor, and Qualcomm(QCOM) DragonWing as examples of ARM-based momentum in robotics and autonomous systems. These markets are slower to ramp than smartphones, but once designed in, they can last for years.

The market backdrop is favorable, but not frictionless. Consumer electronics remains large but mature. Smartphone refresh cycles are longer. Memory supply constraints can distort unit shipments. That is why ARM’s ability to raise royalty per chip matters so much. It turns a mature market from a ceiling into a funding source for newer growth.

Customer Profile

ARM’s customer base spans semiconductor companies, hyperscalers, OEMs, and embedded system developers. The profile is broad, but the economic weight is concentrated among large, sophisticated buyers. These are not casual customers. They are companies building flagship smartphones, cloud CPUs, networking silicon, automotive platforms, and AI systems.

In mobile, ARM serves major Android ecosystem players and indirectly benefits from broad smartphone adoption. In cloud, key customers and partners include Amazon(AMZN) through Graviton, Alphabet(GOOGL) through Axion, Microsoft(MSFT) through Cobalt, and Nvidia(NVDA) through Grace and Vera-related CPU platforms. In automotive and robotics, ARM is increasingly embedded in next-generation compute platforms where power efficiency and software portability matter.

This customer profile is a strength because it aligns ARM with the companies spending most aggressively on compute. It is also a risk because large customers have bargaining power and some are building more custom silicon in-house. ARM’s job is to remain indispensable even as customers become more capable. So far, the evidence suggests it is doing that by moving up the value stack with CSS and broader platform offerings.

One specific customer issue to monitor is SoftBank. Management said the parent is not interested in selling shares and framed SoftBank-related licensing revenue as durable. Still, any time a controlling shareholder is also a major commercial counterparty, governance-minded investors should pay attention. It may be fine. It just should not be ignored.

Competitive Landscape

ARM competes across several layers. In server and PC compute, the main architectural rival is x86, represented by Intel(INTC) and AMD(AMD). In mobile, the threat comes less from x86 and more from custom CPU efforts and alternative architectures. Over the longer term, the most important structural threat is RISC-V, which offers an open instruction set and is attracting ecosystem investment.

Against x86, ARM’s case is power efficiency and design flexibility. Hyperscalers are increasingly willing to adopt ARM-based CPUs because performance per watt matters more in AI-heavy data centers. Google, AWS, and Microsoft are not making symbolic moves here. They are redesigning important parts of their infrastructure stack. That is a strong validation of ARM’s competitive position.

Against RISC-V, ARM’s defense is ecosystem maturity, software compatibility, and time-to-market. RISC-V is real and growing, but ARM still has the deeper toolchain, broader deployment base, and more proven path for high-volume commercial products. In other words, RISC-V is a credible challenger, but ARM still owns the highway while the challenger is building on-ramps.

The more subtle competitive issue is customer insourcing. Some large customers may want more custom cores, more internal IP, or more negotiating leverage. ARM’s answer has been to make its own offerings more valuable through CSS, subscriptions, and broader system support. That is smart. If customers are going to build more themselves, ARM wants to be the fastest way for them to do it.

Peer valuation data in the provided dataset is incomplete, but the broad comparison is still clear. ARM trades at a much richer multiple than most semiconductor peers because investors are valuing it as a scarce AI-enabler platform rather than a conventional chip company. That premium can persist if execution remains excellent. It can also compress quickly if growth cools even modestly.

Macro & Geopolitical Landscape

Macro conditions matter for ARM, but less through direct cyclical inventory swings and more through customer capex, device demand, and export policy. The current semiconductor backdrop is constructive because AI infrastructure spending remains strong, hyperscaler capex is elevated, and automotive compute content continues to rise.

The main macro risk is that AI enthusiasm outruns actual monetization for ARM’s customers. If cloud providers slow deployment or consumer demand weakens sharply, ARM would still grow, but the market might stop paying such a heroic multiple. High-beta stocks with premium valuations tend to get punished first and asked questions later. ARM’s beta of 3.34 underscores that point.

Geopolitically, China remains a meaningful risk. ARM disclosed that U.S. trade restrictions had an impact on direct and indirect licensing in fiscal 2025, and Arm China remains economically important. The company reported $670M of share-of-revenue from Arm China in fiscal 2025 and a $285M net receivable at March 31, 2025. That is not a footnote. It is a material exposure to a politically sensitive market.

There is also ongoing litigation risk involving Qualcomm(QCOM) and Nuvia. IP companies live by the sword and occasionally spend a lot of time in court polishing it. Litigation can be costly, distracting, and strategically important, especially when it touches architecture rights and customer relationships.

Finally, SoftBank ownership creates a market-structure issue. Public float is limited relative to shares outstanding, with float around 136.6M versus 1.06B shares outstanding. That can amplify volatility, squeeze short-term supply, and distort price discovery. It is one reason the stock can trade with more drama than the underlying business deserves.

Balance Sheet Health

Income Statement Strength

Estimates Outlook

Valuation Assessment

Target Prices & Recommendation

Closing

ARM is one of the rare semiconductor companies that can plausibly claim both strategic centrality and financial elegance. It has a dominant installed base, an asset-light model, exceptional margins, and real momentum in the most important compute transitions of this cycle: AI inference, hyperscale CPU adoption, edge intelligence, and automotive compute. The company is not selling shovels in a gold rush. It is licensing the shape of the shovel.

That said, investing is not just about finding great businesses. It is about finding great businesses at prices that still leave room for returns. ARM’s fundamentals support long-term optimism, but current valuation demands caution. For a balanced investor with a medium-term horizon, the right posture is Hold, with a willingness to upgrade on pullbacks or after further earnings growth closes the gap between business quality and stock price.

The key signals to watch over the next 12 to 18 months are clear: data center royalty mix, CSS penetration, Armv9 royalty uplift, durability of SoftBank-related license revenue, and any evidence that R&D spending is translating into broader platform monetization. If those indicators keep moving in the right direction, ARM can justify more of its premium over time. If they stall, the stock may learn the old market lesson that even brilliant companies are not immune to gravity.