Chapter 091　What Did IBM Redefine?

Prescription for Incumbents What did IBM redefine? The question is different in kind from the ones this series has put to other companies. This enterprise has rewritten itself repeatedly over more than a century. And it holds both the periods when the rewriting worked and the periods when it came late. What we treat here is not a winner’s story. It is the fact that, inside one company, a stretch in which redefinition operated sits beside a stretch in which it did not. When the object of comparison is the same company, the cause of the difference narrows to management rather than environment. Neither praise nor prosecution. We read only that difference.

1 What makes this enterprise worth a question

Start with the published numbers. The numbers are not the conclusion. They are the starting point of the question. Full-year revenue for the period ended December 2025 was $67.5 billion, up 8 percent on the prior year and up 6 percent at constant currency. By segment, Software was $29.96 billion, up 11 percent. Consulting was $21.06 billion, up 2 percent, and Infrastructure was $15.72 billion, up 12 percent. Free cash flow was $14.7 billion, an increase of $2 billion on the prior year. The generative AI book of business was shown at more than $12.5 billion on a cumulative basis (results release of January 28, 2026). The following second quarter of 2026, announced July 22, 2026, produced revenue of $17.2 billion, up 1 percent. Software was $7.8 billion, up 5 percent; Consulting was $5.3 billion, flat; and Infrastructure was $3.8 billion, down 7 percent. Gross margin was 57.7 percent. For the full year, revenue growth of 4 to 5 percent at constant currency was indicated. Set those two side by side and no single story holds. Inside one company, three businesses are running on separate clocks. That is exactly why this enterprise is worth a question. There are three reasons. First, the length of time available for observation. Across a business history of more than a hundred years, the principal business has changed hands several times. A twenty-year case teaches you twenty years’ worth. Second, both the successes and the lag sit inside the same company. With other cases, we can only compare separate firms. That introduces the confounds of different industries, different eras, and different executives. Comparing periods within one company reduces them. Third, the turn is still under way. The target year for quantum computing is publicly stated as 2029. We are reading a case whose ending we do not know. One personal note here. I worked at this company for a long time. For that reason, this chapter uses nothing I saw or heard inside it. What is used is public information only, and we do not write figures, rankings, or shares we could not verify. The best thing someone who knows the inside can do is not to talk about the inside.

2 Conventional answers and their limits

Three explanations of this company circulate, and they point in different directions. Each is partly right. Each runs aground in the same place. The first conventional answer: “IBM is a company of the past” A mainframe company, beaten in the cloud race, and off the growth markets. This account is consistent with the record of the 2010s. It is not consistent with the numbers for 2025. Up 8 percent for the year, with free cash flow of $14.7 billion. In the fourth quarter, IBM Z grew strongly and was disclosed as having lifted the growth of hybrid infrastructure. Companies of the past do not produce numbers like that. Nor can we rush to the opposite conclusion. Infrastructure fell 7 percent in the second quarter of 2026. Pulling out only the quarters that point up and calling it a revival is the same error as pulling out only the quarters that point down and calling it decline. The second conventional answer: “The Red Hat acquisition won the cloud race” On July 9, 2019, the company completed its acquisition of Red Hat. The equity value was about $34 billion. The announcement at the time set out a next-generation hybrid multicloud platform built on Linux and container technology. The limit of this account is that it mistakes the ground on which the contest is being fought. The company has not chosen to compete on the scale of cloud infrastructure itself. What can be read from public information is a position that holds the layer running across multiple clouds and the customer’s own facilities. The announcement also stated a policy of widening partnerships with other major cloud providers. So this was not an acquisition made to win a competition. It was an acquisition made to move the definition of the competition. And moving the definition guarantees nothing. The third conventional answer: “It was late to AI” Memories of the Watson era keep this assessment alive. Yet as of the end of 2025, the generative AI book of business was disclosed at more than $12.5 billion cumulatively. The breakdown given in the earnings materials put more than $10.5 billion of that in Consulting and more than $2 billion in Software. What to look at here is not the amount but the composition. Most of the company’s AI revenue comes not from selling models but from implementing them inside enterprises. This is a strength, and at the same time a fragility. Implementation is tied to human hours, and it does not replicate the way software does. What all three answers miss All three ask what kind of company this is — 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 can be seen changes. The value this enterprise has delivered for more than a century is neither machines nor services nor software. It is that society’s core operations do not stop. Tabulating machines, mainframes, outsourced operations, and hybrid cloud were each one expression of that.

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. Everything below stays within what can be read from public information. It is not an assertion about internal decisions. Purpose — the core did not move, the expression moved many times The canon places an important note. The five dimensions do not change at the same frequency. Core Purpose can remain stable while its expression and its means of realization evolve. Few cases fit that note as well as this one. The object was consistently the core of the enterprise. What changed was the means by which that core was held up. There was an era of holding it up with machines, an era of holding it up with outsourced operations, and now an attempt to hold it up with cloud and AI. At Think 2026 on May 5, 2026, the company presented a framework it called the AI operating model. The concept integrates four things: agents, data, automation, and hybrid. CEO Arvind Krishna said that running AI inside an enterprise requires a new operating model. The claim is that AI-driven systems should be managed with the same rigor and governance as the most critical infrastructure. The words are new. The position has not moved. Because the core has not moved, everything around it can be swapped out as often as needed. Business — separating into three layers was the redefinition In the business dimension, the largest rewriting was the separation of the business into three layers. Software, Consulting, and Infrastructure. That separation was completed with the Kyndryl spin-off of November 3, 2021. The decision carved out the managed-operations business as an independent company. Krishna explained it at the time as one of a series of actions sharpening the focus on hybrid cloud and AI. The meaning of the carve-out is not a reduction in scale. It moved the definition of the business from “we take custody of the customer’s facilities and run them” to “we let the customer run anywhere.” The first is tied to a place. The second is the value of letting the customer choose the place. And the three-layer structure separated the clocks on which growth runs. Organization — the boundary opened outward In the organizational dimension, two movements can be read from public information. One is opening. On May 21, 2024, the company released its Granite family of language and code models under the Apache 2.0 license. Krishna said that open means choice. The sense of it is that more eyes see the code, more minds turn to the problem, and more hands reach toward a solution. A development method called InstructLab, built jointly with Red Hat, was also presented. The other is the preservation of independence. The announcement at the close of the acquisition stated explicitly that Red Hat’s independence and neutrality, and its commitment to open source, would be maintained. It is a design that does not dissolve the acquired asset into the acquirer. As the canon defines it, an organization is not a collection of people. It is a value-creation system made of people, AI, partners, universities, and customers. Publishing models and preserving the independence of an acquired company are designs close to that definition. But the reality of the organization cannot be observed sufficiently from outside. Capital — the dimension on which judgment splits most The capital dimension is decisive for reading this enterprise. The same company has made opposite capital allocations in different periods. “Roadmap 2015,” presented by the then CEO in 2010, put a single target at its center: non-GAAP earnings per share of at least $20 by 2015. The plan came with a policy of returning a substantial share of free cash flow to shareholders through dividends and buybacks. On October 20, 2014, the company announced that it was abandoning the target. What was being allocated in that period was not possibility. It was earnings per share. 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. However much the financial capital term is manipulated, Future Capital does not grow unless the Learning and Knowledge terms grow. The Red Hat acquisition of 2019 has a different character. It committed about $34 billion to an area that was not then the principal business. First Principle 3 states, Capital Exists to Create Possibility. Capital exists to create possibility. Ten years apart, the same company stood on both the wrong side and the right side of that principle. Leadership — not a forecast, a published timetable In the leadership dimension, the most observable item is the handling of quantum computing. On June 10, 2025, the company published its path to large-scale fault-tolerant quantum computing. The target is to realize “Starling” in 2029 and to run circuits on the scale of 100 million quantum gates over 200 logical qubits. Roles have been assigned to the intervening years as well, under the names Loon, Kookaburra, and Cockatoo. On November 12, 2025, the company announced “Nighthawk,” a 120-qubit processor. A target of demonstrating verifiable quantum advantage by the end of 2026 was presented alongside it. This is not a forecast. It is the publication of a timetable. The fourth equation of the canon defines leadership as follows. Leadership = Purpose × Question Design × Capital Allocation × System Architecture × Trust In a management that hands out a timetable, the terms doing the work are System Architecture and Trust. As First Principle 9 states, Leadership Means Designing the Future. Leadership means designing the future. But a published timetable is also a verifiable promise. Delay reduces the Trust term directly. Here again, the strength and the fragility sit in the same place.

4 Structure — maturity and the value chain

4.1 Where it sits on the Enterprise Redefinition Maturity Model

We apply the Enterprise Redefinition Maturity Model (ERMM). We assert no level, and confine ourselves to the behavior that can be observed. In the business dimension, behavior beyond a Transformation Enterprise is visible from 2019 on. An acquisition and a separation were carried out within the same few years, and the composition of the business itself was rebuilt. Whether this is a permanent organizational capability or was a periodic project cannot yet be judged from outside. In the capital dimension, behavior differs sharply by period. Capital allocation in the first half of the 2010s was worked backward from a single indicator, earnings per share. The state the source paper describes for an Improvement Enterprise overlaps here. This is not criticism. It is the application of a definition. In the leadership dimension, the quantum roadmap shows the behavior of designing a future ecosystem. But a Future Value Enterprise is the stage at which a firm competes through superior enterprise evolution rather than superior execution. Whether that level has been reached cannot be tested until the results of 2029. The organizational dimension cannot be observed sufficiently from outside. Here we recall the canon’s notes. The Enterprise Redefinition Maturity Model 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 this case shows one more thing. Maturity rises and falls by period inside the same company. It is not a static rating. A company that once reached a high level can return to the behavior of a lower one through a single choice of indicator. Level 5 must also not be treated as a target to be reached as fast as possible. The appropriate level differs by industry and environment.

4.2 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 that order over this company’s three periods. In the first period, the turn at the end of the twentieth century, the order held. The company recognized that the market for machines had changed structurally, learned, and redefined itself from a machine company into a services and software company. The recovery in enterprise value came after. In the second period, the first half of the 2010s, the order was inverted. A target on the Enterprise Value side, earnings per share, was set first, and capital allocation was worked backward from it. Learning and Redefinition are not things you can work backward to. As a result, the company’s revenue fell below the prior year for twenty-two consecutive quarters, a run that finally ended in the fourth quarter of 2017. In the third period, from 2019 on, the order is returning. The acquisition and the separation, the publication of the models, and the long-range quantum plan. None of them can be explained by working backward from Enterprise Value. Enterprise value comes last. And two periods in this company’s history are a field record of what happens when the order is reversed. The second equation confirms the same thing. FVCC = Purpose × Learning × Redefinition × AI Integration × Ecosystem × Capital Allocation × Trust Seven terms, multiplied. What weakened in the first half of the 2010s were the Redefinition and Capital Allocation terms. However large the other terms, if those two are small the whole is small.

4.3 Time as a term

The fifth equation places time as an independent term. Future Value = Future Time × Future Capability The target year for quantum is 2029. None of that value appears in the 2026 results. Whether a company can keep allocating capital to something that does not appear is the substance of Future Time. The Value Equation says the same thing. Value = Purpose × Trust × Capability × Time The asset of a hundred-year company is not only accumulated capability. It is the structure that can handle long time at all. Yet in the first half of the 2010s, the same company was run on a five-year horizon. Holding time and using it are different things.

5 What it looks like in practice — four turns, and what

separated them Redefinition is not the act of adding. It is the act of deciding what to protect and what to release. We set out four phases together with what was released. Phase 1 — the turn at the end of the twentieth century The center of gravity of the information industry moved from large machines to distributed systems, and the company’s business structure shook at its foundation. Here the company chose not to split itself, and instead moved its center of gravity toward services and software. What it released was its self-image as a hardwarecentered firm. We could not confirm primary sources for the loss figures or headcount numbers of that period in the reporting for this chapter. So we do not state them. Write what cannot be confirmed, and the credibility of every other statement falls with it. Phase 2 — the stall of the 2010s This phase is the center of the chapter. The company was not doing nothing during it. Divestitures and acquisitions were made, efficiency programs were run, and new technologies were funded. Improvement continued. And still revenue fell below the prior year for twenty-two straight quarters. What was missing? The business model itself was never questioned. “Roadmap 2015” was not a plan that asked what the enterprise should be. It was a plan that asked where earnings per share should land. Under that plan, buybacks and efficiency programs are rational actions. Each individual judgment is correct, and only the direction of the whole is undetermined. The source paper’s description of an Improvement Enterprise names this state precisely. Improvement is systematic but incremental, and existing business models are rarely questioned. Organizations become increasingly efficient while remaining fundamentally unchanged. Here is the most important observation in this chapter. The company that fell behind was not standing still. It was moving. It was simply running improvement rather than redefinition. From outside, those two are hard to tell apart. They become distinguishable in the revenue curve a few years later. Phase 3 — 2019 to 2021, buying and cutting The Red Hat acquisition closed on July 9, 2019, and the Kyndryl separation completed on November 3, 2021. In two and a half years, the company bought big and cut big. The two look like separate judgments. They are two faces of the same one. Capital went in to take a new layer called hybrid cloud, and a managed-operations business tied to place went out. What was released was scale itself. The side to learn from is not the acquisition but the cut. Acquisitions clear internal approval. Separations do not. Phase 4 — AI from 2023, and quantum, which has not yet produced a result In the generative AI phase, the company took two positions: the implementation side and the open side. The first shows up in the composition of the $12.5 billion noted above, and the second in the release of the Granite models under Apache 2.0. It is a position that does not aim at monopoly over the models themselves. The core side is moving too. The z17, announced April 8, 2025, carries the Telum II processor and was presented with up to 450 billion AI inference operations per day (generally available from June 18 of the same year). It is a design that puts AI where the transaction happens rather than carrying the transaction out to where AI lives. What separated the turns that worked from the turn that lagged Set the four phases side by side, and the difference narrows to three points. First, what was released. Every phase that worked released something. A self-image, scale, the possibility of monopoly. What was released in the phase that lagged was investment and headcount. When the object released is an expense rather than an asset, that is not redefinition. It is contraction. Second, the order. In the phases that worked, Purpose and Learning came first. Third, the handling of time. The phases that worked had horizons measured in decades. The phase that lagged had a horizon of five years and a destination that was a single financial indicator. The conditions under which this structure breaks Praise is the enemy of analysis. We set out four fragilities readable from public information. First, the composition of AI revenue. Revenue that comes from implementation is tied to human hours. Growth in scale comes with growth in headcount. Second, the cyclicality of infrastructure. Infrastructure fell 7 percent in the second quarter of 2026. A business subordinate to product cycles swings the quarterly picture hard. Third, the uncertainty of quantum. The target year of 2029 is published, but it is not a promise of achievement. Fourth, change in the underlying assumption itself. The value of hybrid holds only as long as enterprises use several environments for different purposes. All of these are about the future, and none can be asserted. What we can say is only that the present configuration is effective and conditional at the same time.

6 What transfers, and questions for the executive

What transfers from this case, and under what conditions? Four things. First, being an old firm is not a disadvantage. Companies with long histories are commonly said to be incapable of change. This case shows the reverse. A structure that can handle long time is itself an asset. The question is whether that length has been converted into a horizon for the next decade. Second, the indicator governs the management. A plan built around earnings per share was operated rationally within its own terms. The ROE and PBR targets incumbents set can have the same structure. The indicator is not the problem. Management worked backward from an indicator inverts the order of the Future Value Chain. First Principle 2 states, Future Value Precedes Enterprise Value. Before enterprise value, there is future value. Third, buying and cutting are one judgment. Most incumbents have made acquisitions. Fewer have made separations. A company that only buys accumulates assets while losing focus. As First Principle 6 states, Enterprise Exists to Redefine Itself. An enterprise exists to redefine itself. Fourth, do not erase the period of lag from your own corporate history. This company’s 2010s remain in its financial filings and in the press coverage. Because they remain, we can learn from them. First Principle 5 states, Learning Is the Ultimate Competitive Advantage. The teaching material with the highest learning efficiency is your own failure. One strong caution here. This company must not be imported as a model. The appropriate level on the Enterprise Redefinition Maturity Model differs by industry and environment. The source paper is explicit that making the fastest possible arrival at Level 5 the objective is a mistake. Finally, three questions. Each can be answered at your next executive meeting. Question 1 — Is your company improving right now, or redefining? From inside, the two are hard to tell apart. There is one way to tell. Check whether, in the past three years, the assumptions underlying your own business model reached the agenda. If they did not, what is happening is improvement. Improvement is not bad. Calling improvement redefinition is what is dangerous. Question 2 — From which end is your medium-term plan worked backward? If it is worked backward from a financial target, the plan starts at Enterprise Value. If it starts from what future you want to bring into existence, it starts at Purpose. The first page of the plan will tell you. Question 3 — What will your company release next? Redefinition is not the act of adding. It is the act of deciding what to protect and what to release. A company that cannot name a single candidate for release has not begun redefining. And while AI can evaluate the options, what to discard is decided by people. As First Principle 4 states, AI Optimizes. Humans Define. AI optimizes. Humans define. What IBM redefined was neither a product nor a business area. It was the correspondence between one role — holding up the core of the enterprise — and the technology that fills that role in each era. And what can really be learned here is not the etiquette of success. It is the fact that one company, holding the same assets and the same people, redefined itself in one decade and did nothing but improve in another. What produced the difference was not the environment. It was the design of the order — what gets placed first. There is Purpose, then Learning, then Redefinition, then Creation, and Enterprise Value comes last. A decade that kept that order and a decade that inverted it sit side by side in one corporate history. What we should read is that arrangement.

In brief

What IBM redefined is the correspondence between one role — holding up the core of the enterprise — and the technology that fills it in each era.
What can be read from public information is that the business dimension sits close to a Continuous Redefinition Enterprise, while capital allocation in the first half of the 2010s overlaps the description of an Improvement Enterprise.
Value was not created at Creation. It was created on the step that runs from Purpose to Redefinition.
If the quantum timetable slips, and if the assumption that enterprises use several environments is lost, this configuration breaks.

Key concepts

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

The chain of ideas

Core Purpose → Recognize → Redefinition → Capital Allocation → Enterprise Value

Related first principles

Principle 2 — Future Value Precedes Enterprise Value. Principle 3 — Capital Exists to Create Possibility. Principle 5 — Learning Is the Ultimate Competitive Advantage. Principle 6 — Enterprise Exists to Redefine Itself.

Related chapters

Vol. V, Ch. 044 “What Is the Enterprise Redefinition Maturity Model (ERMM)?” — why maturity rises and falls by period
Vol. VI, Ch. 059 “Cases of Enterprise Redefinition” — the turns in this chapter, reread as six patterns
Vol. VI, Ch. 056 “What Does It Mean to Redefine Investment?” — the danger of capital allocation worked backward from a financial indicator
Vol. II, Ch. 017 “Will Companies Live Longer in the Age of AI?” — the conditions for turning a structure that can handle long time into an asset

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 stract=7120980 / Paper. Zenodo: SSRN: https://ssrn.com/abhttps://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, #017 “Will Companies Live Longer in the Age of AI?” (Vol. II, Ch. 017 of this series) / #059 “Can Large Enterprises Survive in the Age of AI?”

Chapter 091 What Did IBM Redefine?