Chapter 094　What Did Arm Redefine?

Prescription for Incumbents What did Arm redefine? To answer “a semiconductor company without factories” is to state part of a fact. To answer “a company that sells designs” is still not enough. What we establish in this chapter is how this company came to stand where it stands, as the common foundation of an industry. And what that position brought it, and what it did not. Here the central structure of the age of AI appears in the open: creating value and being paid for it are not the same thing. This chapter works from public information only. What can be confirmed and what cannot are kept apart. Praise is not analysis.

1 What makes this enterprise worth a question

Start by setting two numbers side by side. The moment they sit together, the strangeness of this enterprise is visible. The first. Revenue for fiscal 2026, ended March 31, 2026, was $4.92 billion, up 23 percent year on year. Within that, royalty revenue was $2.613 billion and license and other revenue was $2.307 billion. It was the third consecutive year of revenue growth above 20 percent since listing (results materials announced May 6, 2026). The following first quarter of fiscal 2027 ended June 30, 2026. Revenue was $1.29 billion, up 22 percent year on year (announced July 29, 2026). The second. Chips based on this company’s designs have shipped in cumulative volumes of more than 350 billion units. More than 22 million software developers use its development environment, said to be over 80 percent of the world total. The company describes itself by saying that 100 percent of the connected global population uses Arm-based products. The sources are the company’s official website and the shareholder letter released February 4, 2026 (verified August 1, 2026). Set those two beside each other. An enterprise present in almost every computer in the world has annual revenue of about $5 billion. The orders of magnitude do not line up. There sits the structure at the center of this chapter. Creating value and receiving payment for it are separate problems. This company is extraordinary at the first and ordinary at the second. The gap is not an accident. It is a consequence of design. There are three reasons this enterprise is worth a question. First, the divergence between value creation and value capture appears here in as pure a form as it ever does. We treated that issue in Vol. VII, Ch. 069. Arm is the textbook instance of it. Second, the ownership structure changed three times and the core of the business did not move. From listed company to private company, to acquisition target, and back to a listing. Through all of it, what the company sells has not changed. Third, that premise began to shift in 2026. In that year the company put into the world a data center CPU of its own design. A company that supplied only designs is becoming a company that supplies products.

2 Conventional answers and their limits

We take up three explanations about this company that are in circulation. Each is partly right. Each drops the part that matters. The first conventional answer: “Margins are high because it has no factories” This is an explanation about the fabless business form. No capital expenditure, no inventory, therefore high margins. But the explanation does not match the accounting facts. NonGAAP operating income in the first quarter of fiscal 2027 was $531 million, an operating margin of 41.2 percent. GAAP operating income for the same quarter was $91 million, an operating margin of

7.1 percent (announced July 29, 2026). That difference is not small.

Having no factories changes the form of the cost. It does not remove the cost. Design, verification, and the upkeep of an ecosystem all require continuous spending. “Fabless, therefore profitable” is a coarse explanation. The second conventional answer: “It is strong because it holds the de facto standard” This is the explanation that whoever owns the standard owns the value. By the company’s own account, it has reached a share of about 50 percent of the CPU compute base of the major hyperscalers (announced May 6, 2026). But holding a standard and being paid for it are different things. More than 350 billion chips have shipped in cumulative terms, and annual royalty revenue is $2.613 billion. The owner of a standard does not necessarily get to set its share of the value created on top of that standard. If anything the reverse holds. To be adopted as the common foundation, the share had to be kept low. The conditions for becoming a standard and the conditions for making money often point in opposite directions. The third conventional answer: “It won in mobile, so it will win in the data center” This is the explanation that the same strength, power efficiency, works in data centers now that they face power constraints. The direction is right. The company’s Neoverse CPUs reached a cumulative 1.5 billion cores shipped by the end of June 2026. Of those, 500 million accumulated in the previous nine months (announced July 29, 2026). Data center royalties are said to have more than doubled year on year. But the shape of the competition is not the same. In mobile, this company’s customers were chip companies. In the data center, the customers at the center are cloud operators that design their own. In a market where the buyer is the designer, the supplier’s bargaining power changes. The same strength may work in a market without the same way of earning working there too. What all three conventional answers miss All three ask what kind of company this is by asking what it sells. But the second dimension of Enterprise Redefinition does not ask what a firm sells. It asks what value the firm delivers. Change the question and what you can see changes. What this company provides is not a blueprint. It is the state in which an entire industry can share a foundation for computation. To create that state, not becoming a manufacturer was a condition.

3 What was redefined — an analysis across the five

dimensions We read along the five dimensions of Enterprise Redefinition. The order follows the canon: Purpose, Business, Organization, Capital, and Leadership. What follows is what can be read from public information. It is not an assertion about internal decisions. Purpose — computation per unit of power as the core The company places its own starting point in a goal: to design computers on the premise that they run on batteries (company official website). That one sentence governed the four decades that followed. The core is how much computation you can get per unit of power. That core has not moved. What moved is where it is applied. From embedded devices to mobile phones, to smartphones, to vehicles, and to the data center. As the canon states, Core Purpose can remain stable while its expression and its means of realization evolve. This company is an instance. Power constraints grew tighter as the era advanced. By not moving the core, the company ended up meeting the era on its way in. Business — the decision not to manufacture decided what could be sold The redefinition of the business dimension can be reduced to one point. The decision not to manufacture. If you do not manufacture, what you can sell is not a product. It is the right to use a design. From there the revenue structure split into two layers: license revenue for the right to use a design, and royalty revenue arising on each chip shipped. In fiscal 2026 the ratio was roughly half and half. This two-layer structure has an important property. License revenue is booked at contract; royalty revenue is booked years later at volume production. The company’s income statement therefore reflects the results of contracts signed several years earlier. The quarter’s numbers do not reflect the quarter’s management. In recent years the company has raised the unit of what it supplies. From individual CPU designs to Compute Subsystems (CSS), which bundle several building blocks together. As of the end of December 2025, there were 21 CSS licenses across 12 companies. Five of those companies were shipping CSS-based chips (announced February 4, 2026). In its shareholder letter, the company states that CSS expands both the value Arm delivers and the economics Arm captures. That sentence matters. The company itself recognizes the delivery of value and the capture of value as separate variables. Organization — an organization whose boundary lies outside the firm This company’s organization is not complete inside its own walls. The design sits inside; manufacturing, implementation, and software all sit outside. In the May 2026 announcement, more than 50 companies were listed as supporting the expansion of Arm’s compute platform. AWS, Broadcom, Google Cloud, and Marvell appear there. Microsoft, Micron, NVIDIA, Oracle, Samsung, SK hynix, and TSMC are named as well. The figure of 22 million developers is also an asset outside the company. The canon defines an organization not as a collection of people but as a value-creation system made of people, AI, partners, universities, and customers. This company’s organization is close to that definition. But close is not the same as strong. Assets outside the boundary cannot be controlled. Capital — accumulation that never appears on the income statement The capital dimension matters most at this company. What it built over four decades was not equipment. It was a convention called an instruction set, the total volume of software sitting on top of it, and the skills of the developers who can work with it. None of these are items the company capitalizes. All of them support the whole of its bargaining power. The third equation of the canon defines capital as follows. Future Capital = Financial × Human × Learning × Trust × AI × Knowledge × Ecosystem × Purpose This is multiplication. If any single term is zero, the whole product is zero. In this company’s case, the Knowledge and Ecosystem terms compounded over a long period. At the same time, the Trust term played a peculiar role. We take that up in the next dimension. Leadership — a management that keeps choosing to be neutral The mark of the leadership dimension is the maintenance of neutrality. As a principle, the company has offered the same designs to all customers. Team up with one customer and shut out the rest, and the position of common foundation does not hold. Competing enterprises able to fight on the same foundation: that was the company’s product. First Principle 8 states, Trust Compounds Faster Than Capital. Trust compounds faster than capital. In this company’s case, trust did not merely compound. It also set the ceiling on revenue. As long as it stays neutral, it cannot step into its customers’ business territory. And in 2026 the company put a hand on that boundary. We treat that issue in section 5.

4 Structure — the divergence between value creation

and value capture

4.1 Where the divergence comes from

Fix the terms. Value Creation is the total volume of value brought into the world by an enterprise’s activity. Value Capture is the part of that which the enterprise receives as payment (→ Vol. VII, Ch. 069). In this company’s case, the two magnitudes plainly do not match. More than 350 billion chips run on its designs. Royalty revenue in fiscal 2026 was $2.613 billion. Against the economic scale of the computation running in the world’s mobile devices, vehicles, and data centers, that is an order of magnitude apart. Why do they diverge? Three reasons can be read from public information. First, the unit of charging sits at the bottom layer of the value chain. What the company charges for is the right to use a design and the number of units shipped. The value added stacked above it — implementation, manufacturing, system integration, and service — falls outside what the company charges for. Second, diffusion itself was the objective, so prices could not be raised. The condition for being adopted as a common foundation was to keep the barrier to adoption low. Price is part of that barrier. Third, as long as neutrality holds, the company cannot stand on the same ground as its customers. The shortest route to a larger share is to move downstream. That erodes the position of a foundation. None of the three is a weakness of this company. All three are consequences of the position it chose.

4.2 Reading it through the equations

The first equation of the canon defines value as follows. Value = Purpose × Trust × Capability × Time This is multiplication as well. If any single term is zero, the whole product is zero. This company held the Trust term extremely high. At the same time, the constraints that maintained that Trust narrowed the range over which Capability could be converted into payment. A principle can be read out of this. The terms of an equation are not independent. An action that maximizes one term can lower the convertibility of another. Multiplication does not license optimizing each term separately. The time term is worth looking at as well. The fifth equation reads as follows. Future Value = Future Time × Future Capability This is multiplication too; a zero in either term empties the whole. Several years pass between a contract and volume production of the company’s designs. That lag has two effects on management. Downturns in bad periods are softened, and the effects of good turns also arrive late. Time is at once a constraint and a cushion for this company.

4.3 Where in the Future Value Chain the value was created

The causal order fixed by the canon is as follows. Purpose → Learning → Redefinition → Creation → Enterprise Value Lay this company’s history over that order and the point at which value arose is Redefinition. The point at which it decided not to manufacture. Before that decision, the company was part of an organization that built computers. After it, the company became an organization that distributed designs to an industry. Creation came next. An enormous number of chips were made by companies other than this one. Enterprise Value came last. Its enterprise value as a listed company only came into question at the final stage of that order. The order must not be read in reverse. This company’s enterprise value is the result of the redefinition, not its cause.

4.4 Where it sits on the Enterprise Redefinition Maturity Model

Within what public information allows, maturity varies by dimension. Nothing can be asserted. In the business dimension, the business model evolves continuously. From the right to use a design, to the higher unit of CSS, and in 2026 to silicon of its own design. That is behavior close to a Continuous Redefinition Enterprise. It changes the unit it supplies before external disruption requires it. In the purpose dimension, the Core Purpose has stayed stable while its expression adapted. Widening the field of application without weakening organizational identity reads as a high level against the paper’s assessment question. The capital dimension, on the other hand, is hard to assess from outside. Three changes of ownership structure mean the party allocating capital has itself changed. The organization dimension cannot be observed sufficiently from outside either. Here we recall the canon’s notes. The Enterprise Redefinition Maturity Model (ERMM) evaluates organizational coherence rather than isolated excellence. An organization can be at Level 4 in AI capability while remaining at Level 2 in leadership. Placing this company at a single level is itself a misuse of the model. Maturity is also assessed across all five dimensions in balance: exceptional technological capability with weak leadership redesign does not produce higher maturity, and strong purpose without adaptive organizational systems remains insufficient. And Level 5 must not be treated as a target to be reached as fast as possible. The appropriate level differs by industry and environment.

5 What it looks like in practice — the pivots, and the

risks Redefinition is not the act of adding. It is the act of deciding what to protect and what to release. We set out this company’s pivots together with what was given up. Pivot 1 — deciding not to manufacture At the first pivot, the company gave up most of the price of the product. Sell only the design and what you receive is a small fraction of the whole. This is the kind of judgment that reliably reduces near-term revenue. It also multiplies, by orders of magnitude, the number of parties willing to adopt you. It was a trade of a lower share for a larger base. Most enterprises cannot choose that trade, because existing revenue falls. Pivot 2 — choosing a market with power constraints The market where the company first won wide adoption was not one where performance came first. It was battery-powered devices. At the time, that sat on the periphery of high-performance computing. What it gave up was the main battlefield of the day. What it chose was the place where the constraint was tightest. Capability sharpened where constraints are tight paid off when the constraints later spread. When data centers hit the wall of power, this company’s design philosophy moved to the center. Pivot 3 — three changes of ownership structure The ownership structure moved three times in a decade. Here we set out facts only. Whether the investment judgments were sound is not the subject of this chapter. On July 18, 2016, SoftBank Group announced the acquisition of ARM Holdings plc. The total was about £24 billion, at 1,700 pence per share, described as a 43 percent premium to the previous closing price. At the same time it was stated that the headquarters would be maintained in Cambridge and that the company would be run as an independent business within the group. A policy of doubling UK employee numbers within five years was also set out. The acquisition by NVIDIA announced in 2020 was terminated on February 7, 2022. Both companies cited significant regulatory challenges as the reason. The $1.25 billion NVIDIA had already paid stayed with SoftBank Group, and NVIDIA retained a 20-year Arm license. SoftBank Group stated that it would begin preparing for a public offering of the company’s shares. On September 13, 2023, the company priced its initial public offering on Nasdaq. It was $51.00 per ADS, for 95,500,000 ADSs, under the ticker ARM. The selling shareholder was a wholly owned subsidiary of SoftBank Group. From the standpoint of redefinition, three things can be read from public information about this decade. First, the private period granted distance from quarterly accountability. Second, the termination of the acquisition was an event in which outside authorities effectively acknowledged that neutrality is an asset of this company. Third, the relisting brought back the discipline of the capital markets. All three, however, are readings. They are not proof of internal decisions. Pivot 4 — 2026, silicon of its own design In May 2026 the company announced a data center CPU of its own design. It is the first data center chip the company designs. It is manufactured on TSMC’s 3nm process. It carries up to 136 Neoverse V3 cores per chip, with power consumption in the 300W class. Against x86-based configurations, it claims more than twice the performance per rack. Meta is named as the principal co-development partner. Systems are supplied by Supermicro, Lenovo, Quanta, and ASRock Rack. Broad availability was set for the second half of 2026. Customer demand is said to exceed $2 billion across fiscal 2027 and 2028, described as double the assumption held at the time of the announcement (announced May 6, 2026). The company explains that customers can use the same architecture in the form of IP, CSS, and now silicon. What is happening here is an attempt to raise the unit of value capture by one step. Note that even at this stage the company holds no factories. Manufacturing remains outside. What changed is not whether it manufactures but whose name the product is sold under. The conditions under which this structure breaks We organize this company’s fragility into four points. All are about the future, and none can be asserted. First, competition with customers. This company’s largest capital was its neutrality. Put out products of its own and some customers become competitors. If trust in it as a foundation is eroded, recovery takes time. Second, the side effects of raising the unit of capture. Action that increases your share raises the barrier to adoption. The position of common foundation rested on that barrier being low. Third, concentration of demand. Data center royalties more than doubled year on year. But the buyers in that field can be counted on one’s fingers. If those buyers raise their own capability to design, the shape of the negotiation changes. Fourth, the thinness of accounting profit. The GAAP operating margin in the first quarter of fiscal 2027 was 7.1 percent. As long as a large gap from the non-GAAP 41.2 percent persists, any discussion of enterprise value depends heavily on which premises are chosen.

6 What transfers, and questions for the executive

What transfers from this case, and under what conditions? Not an imitation of the business form. There are four things to take. First, your share is not decided by technical capability. It is decided by the unit of charging. This company’s designs sit in nearly every computer in the world, yet what it receives stays a small fraction of the product price. Many incumbent component and materials companies sit inside the same structure. They win on technology and lose on share. The problem is not the technology. It is which layer of the value chain you charge at. Second, neutrality is capital and a constraint at the same time. Being able to work with anyone is strong. But competing with no one also means giving up the value added downstream. Chanting “be open” without a conscious view of that trade is not a decision. The reason for staying neutral, and the conditions for releasing it, have to be put into words in advance. Third, ownership structure is a condition of management. Listed or private tends to be discussed as a question of financing. What actually changes is the length of time management can use and the party it must explain itself to. This company’s decade is a rare case in which that fact can be observed from outside. What time horizon does your own ownership structure make possible? That is not a question of capital policy. It is a question of management design. Fourth, becoming a foundation and making money are separate decisions. If you want both, you have to design at the outset which layer you charge at and how. Raising your share after the fact erodes trust in you as a foundation. What this company stepped into in 2026 is exactly that difficult ground. The result is not yet in. And one caution. This company must not be imported as a model. The appropriate level on the Enterprise Redefinition Maturity Model differs by industry and environment. Making the fastest possible arrival at Level 5 the objective is something the source paper explicitly warns against. Finally, three questions. Each can be taken up at your next executive meeting. Question 1 — Does the unit you charge at match the layer where value is created? If it does not, who receives the difference? If you cannot answer, your company has handed the right to decide its share to someone else. Question 2 — Is your neutrality an asset, or inertia? Are you choosing the state of competing with no one, deliberately? If you are not choosing it, that is not a strategy. It is an absence of decision. Question 3 — How many years ahead does your ownership structure allow you to manage for? This is the question of Recognize, the first stage of Enterprise Redefinition. What assumptions about our enterprise are becoming obsolete? AI cannot answer that. AI can test assumptions. Which assumptions should be doubted is decided by people. What Arm redefined was neither the semiconductor nor the method of design. It was the way an industry shares a foundation for computation. The company chose not to build, and in exchange it created a state in which anyone could build for themselves. Computers across the world running on the same convention is the result of that choice. But the enterprise that created that state does not automatically receive more for it. Value creation and value capture are variables that have to be designed separately. First Principle 2 states, Future Value Precedes Enterprise Value. Before enterprise value, there is Future Value. Arm’s four decades are evidence of that order. And in 2026 the company is trying, on the far side of that order, to design its capture again.

In brief

What Arm redefined is not a method of design. It is the way an industry shares a foundation for computation.
What can be read from public information is that the business and purpose dimensions show the behavior of a Continuous Redefinition Enterprise.
Value was not created at Creation. It was created at the Redefinition where the company decided not to manufacture.
Release neutrality to raise your share and trust in you as a foundation is eroded; the advantage can reverse.

Key concepts

Enterprise Redefinition / the Enterprise Redefinition Maturity Model / Future Value Chain / Future Capital / the Layer Shift Pattern (→ Vol. VI, Ch. 059)

The chain of ideas

Core Purpose → Redefinition (not manufacturing) → Ecosystem → Future Value → Enterprise Value

Related first principles

Principle 2 — Future Value Precedes Enterprise Value. Principle 3 — Capital Exists to Create Possibility. Principle 6 — Enterprise Exists to Redefine Itself. Principle 8 — Trust Compounds Faster Than Capital.

Related chapters

Vol. VII, Ch. 069 “Does Innovation Become Enterprise Value?” — why value creation and value capture do not coincide
Vol. VI, Ch. 051 “What Does It Mean to Redefine Competitive Advantage?” — the separation of becoming a foundation from making money
Vol. VIII, Ch. 077 “What Is Capital Strategy in the Age of AI?” — how ownership structure governs the time horizon of management
Vol. V, Ch. 049 “What Does It Mean to Redefine a Brand?” — how credibility as a common foundation becomes capital

Papers and companion volumes

Kadowaki, N. (2026). Future Value Theory: A Management Framework for Enterprise, Capital, and Society in the Age of AI. VURA Working Paper. SSRN: https://ssrn.com/abstract=7120980 / Zenodo: https://doi.org/10.5281/zenodo. 21255662
Kadowaki, N. (2026). Enterprise Redefinition: Toward an Enterprise Evolution Theory for the Age of AI. VURA Working Paper. (Published on Zenodo; under review at SSRN)
100 Questions on Management in the Age of AI, #071 “What Is a Strength That Cannot Be Copied in the Age of AI?”

Chapter 094 What Did Arm Redefine?