Arm Holdings Plc

Business Model and Scope

Arm designs computing architecture and semiconductor intellectual property rather than manufacturing most chips. Customers license instruction-set architecture, processor cores, graphics and system IP, software tools, or more integrated compute subsystems, then design chips or products around that technology. Arm generally receives license and support revenue during development and per-unit royalties when customers ship qualifying chips. One license can therefore precede a long, uncertain stream of customer investment and royalty-bearing volume.

License models differ in customer freedom and economics. Architecture licenses allow extensive implementation flexibility; technology licenses provide Arm-designed components; Arm Total Access and Flexible Access broaden access across portfolios; compute subsystems combine more validated elements to reduce integration work. The fiscal 2026 Form 20-F also describes investment toward production-silicon products, including an Arm AGI CPU, chiplets, and complete chip solutions. Moving closer to finished silicon can increase customer value while changing margin, capital, competitive relationships, and execution risk.

Arm sits upstream of semiconductor designers, foundries, device manufacturers, cloud providers, operating systems, and application developers. Its design has economic value only when partners tape out, manufacture, and sell successful chips and when software runs reliably on the resulting installed base. Arm Technology (China) Co. Limited operates independently and provides access to the Chinese market under a commercial relationship; neither Arm nor SoftBank Group controls its operations.

The central question is whether architectural compatibility, energy-efficient designs, broad software support, and partner economics can sustain license and royalty cash while open architectures, customer insourcing, export controls, more integrated products, related-party dependence, and controlling-shareholder claims evolve.

Customers and Purchasing Decisions

Semiconductor companies license Arm to reduce the time, engineering risk, and software burden of building processors. They compare performance per watt, area, security, verification, roadmap, development tools, royalty and license terms, foundry readiness, customization, and available software with x86 suppliers, RISC-V, other licensed IP, and proprietary design. Switching architecture can require hardware redesign and software migration, but sophisticated customers may use several architectures or develop their own cores.

Original-equipment manufacturers and cloud providers care about the complete system: application compatibility, developer adoption, supply, performance, energy and cooling cost, security maintenance, and control over the roadmap. They may sponsor custom Arm chips while retaining substantial design economics themselves. A customer can license Arm technology without shipping a product, so signed contracts are not sufficient evidence of demand. Tape-outs, production ramps, end-product adoption, and sustained royalty units are the relevant behaviors.

Software developers are not usually direct royalty payers but are essential participants. Toolchains, operating systems, libraries, applications, and engineering skills lower the cost of using Arm. Developers can target multiple architectures, and compatibility layers can reduce switching friction, yet a large installed base makes abandonment costly. Foundries and electronic-design-automation vendors enable implementation; their process roadmaps and tools constrain Arm's practical performance.

Arm China is simultaneously a route to customers and a concentrated counterparty. The Form 20-F identifies dependence on a limited number of customers and significant related-party activity. SoftBank-controlled companies can also be customers or strategic partners. Growth from related parties must therefore be separated from independent market adoption and tested for arm's-length price, payment, and durable external demand.

Profit Creation and Value Capture

License and other revenue comes from IP access, software tools, design services, training, support, and maintenance. It is lumpy because negotiation, delivery, milestones, customer decisions, and revenue recognition do not align evenly. Royalty revenue depends on customer chip shipments, product mix, selling price where contractually relevant, royalty tier, and architecture generation. The fiscal 2026 Form 20-F states that royalty growth benefited from higher-rate products such as Armv9, showing why unit growth and royalty mix must be distinguished.

The attractive economic mechanism is reuse. Research and verification for an architecture or core can serve multiple customers and product generations, while a successful licensed design can generate royalties over years with limited per-unit Arm cost. However, engineering is a continuing expense, not a one-time asset. New nodes, security, software, AI workloads, and competitor performance require substantial research before revenue. Customer support and physical-design enablement also rise as products and use cases broaden.

Royalty contracts can include volume tiers, so royalty per chip may decline as a device scales, subject to agreed minima. Chip average selling prices also tend to fall over a product generation. Arm captures more when newer, higher-value designs justify higher rates or enter servers and other valuable systems; it captures less when partners retain integration value, units remain in mature low-price devices, or customers negotiate broad access. Foundries, chip designers, device brands, cloud operators, and software platforms can retain more industry profit than the IP supplier.

Working capital includes license receivables, contract assets, deferred revenue, support obligations, Arm China balances, employee compensation, and supplier commitments. Remaining performance obligations exclude uncertain future royalties, so they are not a complete backlog. Growth creates value when risk-adjusted license and royalty cash exceeds research, share compensation, customer concentration, and the capital required for more integrated products on a diluted per-share basis.

Industry Structure and Capital Cycle

Semiconductors are cyclical because demand forecasts, long design periods, wafer capacity, inventories, and end-market spending adjust at different speeds. Arm is less physically capital intensive than a manufacturer, but royalty units follow the same inventory and product cycle. License decisions can lead shipments by years, making current license growth a poor measure of near-term royalties and current royalties a lagging measure of prior design wins.

Architecture competition has a slower capital cycle. Software, engineers, tools, validated cores, and customer roadmaps accumulate over many years. x86 retains strong positions in personal computers and servers; RISC-V offers an open instruction set and strategic control; customers can build proprietary accelerators or cores. High Arm returns invite these alternatives, particularly among large customers able to fund software and silicon. The cost is engineering and ecosystem duplication rather than a new factory.

Arm's expansion from component IP toward CSS, chiplets, an AGI CPU, or complete solutions changes competitive boundaries. More integration can shorten customer development and raise Arm's share of value. It can also compete with licensees, require manufacturing or inventory commitments, increase verification liability, and reduce customer differentiation. If Arm becomes a chip supplier, its capital cycle can include tape-out cost, packaging, foundry allocation, inventory, and product obsolescence that the historical licensing model largely avoided.

Industry value shifts with workload. Mobile scale favored energy efficiency and software compatibility; cloud and AI increase the value of compute density, interconnect, accelerators, and complete systems. Export controls can split roadmaps by geography. A favorable market narrative about artificial intelligence does not prove Arm captures the economics: customers may use Arm control processors while accelerators, memory, foundries, and cloud services capture most profit.

Sources and Durability of Competitive Advantage

Arm's principal advantage is a coordinated architecture ecosystem: widely licensed instruction sets, proven processor designs, software compatibility, tools, engineering knowledge, foundry enablement, and a large installed base. More chips attract developer support; more software lowers adoption risk for device and cloud designers; more licensees finance broader research; reusable IP can spread cost across customers. Energy-efficiency heritage and neutral licensing strengthen that system where customers prefer to design rather than buy a finished processor.

Observable evidence should include independent design wins, tape-outs, royalty-bearing shipments, Armv9 and data-center adoption, stable or improving royalty value per relevant unit, customer renewal, software readiness, and returns after all research and share compensation. License revenue from a related party or a broad access contract is weaker evidence until external products ship. Reported market share is useful only when tied to economic royalty and customer profitability.

Switching costs are substantial but not absolute. A customer can maintain Arm for legacy products while starting new RISC-V or proprietary designs. Large cloud and device companies possess capital, workloads, software influence, and incentives to reduce supplier dependence. Open-source tools can narrow the ecosystem gap. Compatibility emulation and portable software can make architecture less visible to developers over time.

The advantage could also be weakened by Arm itself. Raising price or imposing terms faster than customer value increases encourages insourcing and alternative architectures. More complete silicon may make Arm less neutral to licensees. Security defects, roadmap delays, disputed license scope, or inadequate developer support can interrupt the feedback loop. The moat is durable when partners earn attractive returns using Arm, not when contractual dependence alone raises Arm's take.

Operating System and Strategic Trade-offs

Arm's system begins with architecture and workload research, then core and system design, verification, physical implementation guidance, software tools, security maintenance, foundry and EDA enablement, licensing, customer engineering, and royalty reporting. A design decision affects performance, power, area, manufacturing yield, software compatibility, customer schedule, and eventual royalty. Verification quality is crucial because a defect replicated across many designs can impose widespread remediation.

The company vertically integrates architecture, key processor IP, software enablement, and customer support while leaving most chip integration, fabrication, packaging, device manufacturing, and distribution to partners. This division keeps physical capital low and lets customers differentiate. CSS moves the boundary toward a more complete, validated subsystem. Production silicon would move it further, requiring capabilities and risk controls that three retained public filings cannot demonstrate across a full product cycle.

Research allocation is the central operating trade-off. Arm must support mature architectures that still generate royalties while funding Armv9, data-center Neoverse, automotive, edge AI, CSS, and prospective silicon. Too many roadmaps can dilute scarce engineering; too much focus on high-value compute can neglect the broad embedded base. Hiring and share compensation can build capability while transferring material economics away from outside shareholders.

Arm China requires contract enforcement, product delivery, royalty reporting, collections, and export-control review across an independently operated organization. Related-party governance requires comparable discipline. The system is difficult to reproduce because hardware, software, tools, and partners must work at release, but this interdependence also means Arm cannot unilaterally ensure customer tape-out, foundry yield, or product demand.

Financial Resilience

Arm's historical licensing model requires substantial research and people but limited factories and inventory. Cash and short-term investments provide liquidity, while receivables, contract assets, and related-party balances depend on collection and performance. Capitalized or acquired intangibles are less useful in stress than the continued relevance of architecture and employees. Operating leases, tax, litigation, customer support, and long-term research continue even if license timing weakens.

Revenue visibility must be interpreted carefully. Remaining performance obligations at March 31, 2026 excluded potential future royalties and included obligations whose timing can depend on customer action. Conversely, royalty cash from installed designs is not fully shown in that measure. Resilience analysis should construct separate schedules for committed license consideration, historical royalty cohorts, Arm China collections, related-party contracts, and essential research spending.

A severe but plausible scenario combines a semiconductor downturn, delayed customer tape-outs, lower device units, loss or restriction of Chinese revenue, a major license dispute, and overspending on production silicon. License milestones and royalties weaken while engineers, software support, litigation, tape-outs, and supplier commitments continue. Arm can pace some new programs and discretionary hiring, but deep cuts to architecture, security, or ecosystem support would damage the royalty base.

The balance sheet may withstand ordinary cyclicality without outside financing if liquidity remains accessible and large commitments are avoided. The strategic stress is more important: a multi-year shift toward open or customer-owned architectures could erode future royalties before accounting assets signal impairment. Any production-silicon inventory or foundry commitments should therefore be matched to customer commitments and tested without assuming capital markets or SoftBank support.

Capital Allocation and Shareholder Outcomes

Core reinvestment in architecture, verified cores, software, security, tools, and customer enablement has first claim because current royalties originate in earlier research. Projects should be evaluated by independent adoption, expected royalty pool, engineering opportunity cost, and the probability of customer production. License announcements and research expense are inputs, not returns.

CSS and production silicon require separate capital gates. A subsystem can create value when it shortens customer time-to-market enough to support higher license and royalty economics without displacing partners. A complete chip must earn after design, masks, foundry, packaging, inventory, support, and channel risk. Management should state stop conditions if expected customer volume or margin fails, rather than using the broad Arm ecosystem to subsidize an unproven product indefinitely.

Acquisitions should add scarce technical capability or software adoption at a return above internal development after retention. Share-based compensation is a direct dilution and cash-equivalent claim; repurchases, if used, should be assessed net of issuance and only at a price that discounts customer concentration, control, and the risk of a changing business model. Cash retained for resilience is valuable when it protects roadmap continuity, but excess cash controlled by a dominant shareholder requires allocation discipline.

SoftBank Group owned approximately 86.4% of Arm's issued share capital at May 21, 2026, according to the Form 20-F. Outside ADS holders have limited influence over board composition and transactions. Common shareholder outcome therefore depends not only on operating returns but on arm's-length related-party contracts, fair capital allocation, and value per diluted share after compensation and any expansion capital.

Legal and Regulatory Exposure

Export controls are a direct product and market-access constraint. The fiscal 2026 filing states that certain Neoverse V processors meet U.S. and U.K. thresholds requiring licenses before delivery to Chinese customers. Rules can change product eligibility, delay delivery, split roadmaps, or prevent sales. Probability of continuing change is high; severity is concentrated but material; duration may be multi-year; design of compliant alternatives offers only partial reversibility.

Arm China's independent operation creates contractual, collection, reporting, and governance exposure. A deterioration can reduce Chinese access even if local end demand remains. U.S.-China and U.K.-China policy can also affect non-Chinese customers selling into China. Compliance errors can damage relationships and trigger sanctions beyond a single transaction.

License scope and intellectual property are central to the model. The 2026 Form 20-F describes continuing litigation with Qualcomm after the earlier Qualcomm and Nuvia dispute. The economic consequence can include contract interpretation, architecture use, royalties, customer confidence, and future negotiating behavior, not only damages. Patent, open-source, and security obligations likewise require continuing defense and remediation.

Competition authorities may examine licensing terms or a move toward integrated silicon, particularly if customers depend on Arm while competing with its products. Cybersecurity and privacy rules cover design data, customer roadmaps, and employee information. As a U.K. foreign private issuer and Nasdaq controlled company, Arm uses governance provisions different from a widely held U.S. issuer. Legal form does not remove the need for transparent related-party and minority-holder protection.

Conclusion, Uncertainties and Disconfirming Evidence

Arm creates value by supplying reusable, energy-efficient computing architecture and verified IP that reduces customer development and software risk. It retains value through license terms, per-unit royalties, ecosystem scale, software compatibility, research knowledge, and switching costs. Those economics are potentially durable, but customers, open architectures, export controls, and Arm's own move toward integrated products can redistribute value.

The financial structure can withstand ordinary semiconductor volatility if liquid resources protect core research and production-silicon commitments remain bounded. Common shareholders receive the benefit only when royalties and license cash exceed research, share dilution, related-party claims, and new product capital. SoftBank control and material related-party activity require a governance discount in the analytical sense even if every disclosed transaction is compliant.

The thesis would be invalidated by sustained independent-customer movement to RISC-V or proprietary architecture, stalled tape-outs despite license growth, lower royalty value after product mix, loss or prolonged nonpayment from Arm China, integrated products alienating licensees, production silicon consuming capital without committed demand, or license litigation materially weakening contract enforceability. A rising installed base with declining per-share cash after compensation would also disconfirm value retention.

On evidence through May 26, 2026, Arm owns an unusually broad architecture ecosystem, but only three retained annual filings do not establish a complete cycle or the return on its more integrated strategy. Business quality must remain distinct from valuation. Investment attractiveness requires conservative assumptions for royalty mix, related-party revenue, China access, research intensity, dilution, and terminal competition.