The Number That Explains Everything
On January 29, 2025, Arvind Krishna stood before analysts and delivered a figure that would have been incomprehensible to anyone who watched IBM's stock flatline for most of the 2010s: $5 billion. That was the cumulative generative AI book of business IBM had amassed since inception — up nearly $2 billion in a single quarter. The number mattered not because it was large by the standards of hyperscalers burning through GPU clusters like kindling, but because it represented something far more improbable. IBM, the company that had fumbled Watson so spectacularly that the name became a punchline in Silicon Valley, was now selling enterprise AI at scale. The same company that had lost half its revenue from its 2011 peak. The same company whose stock price, adjusted for inflation, spent the better part of two decades going nowhere while Microsoft, Apple, Amazon, and Google became the most valuable entities in human history.
And yet here it was, still alive, still relevant, still generating $12.7 billion in annual free cash flow — a figure that would make most SaaS unicorns weep. $62.8 billion in revenue, up 3% at constant currency. Operating gross profit margin expanding by 130 basis points. Software revenue growing 9% at constant currency. The z16 mainframe declared the most successful program in the company's history.
This is the paradox at the center of the IBM story. The company has existed for 114 years. It has been declared dead at least three times — by antitrust regulators in the 1970s, by the market in 1993, by the cloud-native generation in the 2010s. It invented or co-invented the punch card, the mainframe, the hard disk drive, the relational database, the UPC barcode, the magnetic stripe, the personal computer, DRAM, and the scanning tunneling microscope. Six IBM researchers have won Nobel Prizes. The company holds more patents than any other in America, a streak maintained for decades. And yet it has also been the author of some of the most catastrophic strategic errors in business history — licensing PC-DOS to Microsoft, selling the PC division to Lenovo, the Watson debacle — each one a case study in how incumbents fumble the future they helped create.
How does a company survive 114 years in the fastest-moving industry on earth? And what does survival cost?
By the Numbers
IBM in 2024
$62.8BTotal revenue (FY2024)
$12.7BFree cash flow
$5B+Generative AI book of business (inception to date)
300,000+Employees across 170+ countries
$7B+R&D investment in 2024
9%Software revenue growth (constant currency)
19Research facilities across 6 continents
113+Years of continuous operation
The Religion of the Salesman
To understand IBM, you have to understand Thomas J. Watson Sr. — not as corporate hagiography demands, but as the peculiar, contradictory, occasionally monstrous force that imprinted the company's DNA so deeply that it persists a century after his arrival.
Watson was a farm boy from upstate New York who talked his way into a sales job at National Cash Register under John Henry Patterson, the original American sales evangelist — a man who built a corporate culture so totalizing it anticipated the twentieth-century corporation by decades. Watson absorbed everything: the evangelical sales meetings, the rigid dress codes, the conviction that selling was not merely a commercial act but a moral one. He also participated in Patterson's illegal scheme to destroy competitors, which landed both men antitrust convictions in 1912. Watson's conviction was overturned on appeal. The experience didn't humble him. It sharpened him.
When financier Charles Ranlett Flint — the "Father of Trusts," a man who had bundled rubber and wool companies with the practiced indifference of a Wall Street butcher — needed someone to run the unwieldy Computing-Tabulating-Recording Company he'd stitched together in 1911 from three firms making time clocks, scales, and tabulating machines, Watson was the pick. He arrived in 1914 and would not leave for 42 years.
Watson's genius was not technological. It was cultural. He understood, before the concept had a name, that a company's operating system is its culture — and that culture, once correctly installed, scales. His first act at C-T-R was not an engineering initiative or a cost reduction but a meeting in Endicott, New York, where he stood before employees and uttered a single word that would become the most famous corporate motto in American history: "THINK."
The word appeared on signs in every IBM office, on desk plaques, in company songs (yes, there were company songs), and eventually on the notebooks that
Steve Jobs would name a product after decades later. But the word itself mattered less than the ideology it encoded. Watson believed that the combination of information and technology could create an entire industry. More radically for the early twentieth century, he believed that investing in the education and dignity of workers would compound into commercial dominance. IBM offered training programs when most manufacturers offered pink slips. It hired women into professional roles in 1935, employed Black workers in managerial positions long before the civil rights movement, and established a company-wide culture of no drinking, conservative dress, and absolute loyalty to the institution.
I believe in getting behind the individual and backing him up, helping him to strengthen himself … and bringing out the best there is in him.
This was paternalism, to be sure — Watson's IBM was less a corporation than a secular church, complete with hymns and tithing. But it worked. By the mid-1930s, when the Great Depression had gutted American industry, Watson made what might be the most audacious bet in the company's first half-century: he kept his factories running. While other manufacturers slashed production and fired workers, Watson ordered IBM to continue building tabulating machines even though demand had evaporated. He stockpiled inventory, invested in R&D, launched new products into a market that couldn't afford them. When the Social Security Act of 1935 created the need for the largest data-processing operation in human history — tracking the wages of 27 million Americans — IBM had the machines ready. The IBM 077 Collator became the backbone of the New Deal's bureaucratic apparatus.
Watson had bet on the future arriving. It did, and IBM was the only company standing in the right spot.
The revenue doubled during the Depression, a fact so remarkable that Watson reportedly promised to send his entire sales force to Europe as a reward. War intervened. The trip was postponed for twenty years — then finally honored in 1962, when Thomas Watson Jr. sent 187 original salespeople and their spouses across the Atlantic. A promise kept across two decades and two generations of leadership. This was not business. This was religion.
The Shadow on the Ledger
But religion can curdle. The IBM that built the Social Security system also built the infrastructure that administered genocide.
Edwin Black's IBM and the Holocaust documented what the company has never fully addressed: that IBM's punch card technology was used by the Nazi regime to identify, track, and ultimately deport Jews and other targeted populations. The German subsidiary, Deutsche Hollerith Maschinen Gesellschaft (Dehomag), operated under Watson's direct oversight. Major German correspondence was translated for review by the New York office. Machines were leased, never sold — IBM was the sole source of all punch cards — and Watson personally traveled to Berlin at least twice annually from 1933 to 1939 to supervise operations. In 1937, he accepted the Merit Cross of the German Eagle from the Nazi government.
When Germany invaded Poland, Black's research revealed, IBM New York established a special subsidiary called Watson Business Machines — named, chillingly, after the CEO himself — whose sole purpose was to service the Nazi occupation. Punch card machines with English-language labels operated in offices across the street from the Warsaw Ghetto, calculating "exactly how many Jews should be emptied out of the ghettos each day" and routing them to trains bound for Auschwitz and Treblinka.
Watson returned the Nazi medal in 1940. IBM has preferred to suggest that the Third Reich seized control of its European subsidiaries, curtailing the company's influence. The historical record is more ambiguous and more damning than that framing allows. As Leon Krzemieniecki, the last surviving person involved in the Polish administration of rail transportation to the death camps, testified: "I knew they were not German machines… The labels were in English."
This chapter does not appear on IBM's otherwise exhaustive Heritage website. It is, in every sense, the shadow on the ledger — the price of Watson's conviction that IBM could do business anywhere, with anyone, in service of any system that required data processing. The same culture of relentless customer service that built the Social Security Administration's infrastructure served, without apparent moral hesitation, the machinery of the Holocaust. Not a single sentence written by IBM personnel has been discovered questioning the morality of automating the Third Reich.
If IBM's story is one of reinvention, it is also one of complicity. The company's capacity for adaptation — its greatest corporate virtue — has always contained within it the capacity for moral blindness.
The Five Billion Dollar Gamble
Thomas Watson Jr. was not supposed to be the man who transformed IBM into the most dominant technology company on earth. He was, by his own admission in
Father, Son & Co., a mediocre student, an anxious young man who flew transport planes in World War II partly to escape his father's gravitational field. Where the elder Watson was a master salesman who built a culture, the younger Watson was an engineer's temperament trapped in a salesman's body — a man who understood that the future would be built with circuits, not punch cards, even as his father's company remained the world's dominant purveyor of punched card equipment.
The succession was bitter. Watson Sr. clung to power; Watson Jr. pushed for computers. Father and son fought openly, viciously, in a dynamic that mirrored the company's own internal war between its profitable past and its uncertain future. When Watson Jr. finally took full control as CEO in 1956 — his father died six weeks later — he inherited a company generating enormous cash from tabulating machines and facing an existential question: Would IBM lead the computer age or be consumed by it?
His answer came on April 7, 1964, and it was the largest corporate bet in history to that point.
System/360 was not merely a new computer. It was an entirely new architecture — a family of six mutually compatible machines spanning the full range of commercial computing needs, from small businesses to government agencies to scientific research institutions. The name referenced 360 degrees: the system would encompass every application. The bet was $5 billion in development costs (roughly $50 billion in today's dollars), at a time when IBM's annual revenue was approximately $3.2 billion. Watson Jr. was wagering the entire company.
The gamble's strategic logic was elegant and terrifying. IBM at the time sold seven different computer product lines, each incompatible with the others. A customer who bought one system could not run its software on another IBM machine, let alone a competitor's. This fragmentation was profitable in the short term — it created captive customer bases for each product line — but it meant IBM was competing against itself, cannibalizing its own sales force, and leaving openings for focused competitors.
System/360 would destroy all of that. One architecture. One instruction set. One ecosystem. Software written for the smallest System/360 machine would run on the largest. This meant that customers could start small and grow without rewriting their entire software stack — a concept that seems obvious now but was revolutionary in 1964. It also meant that IBM was deliberately obsoleting its entire existing product line, writing off billions in sunk development costs, and betting that the new architecture would win enough market share to justify the investment.
This is the bet. If it doesn't pay off, I'm going to be the son of a bitch who bet the company and lost.
It paid off. System/360 became the most commercially successful computer product in history. It established the mainframe as the backbone of modern enterprise computing, created an ecosystem of third-party software and services that would become IBM's true moat, and generated a market dominance so total that the U.S. Department of Justice filed an antitrust suit in 1969 that would drag on for thirteen years before being dismissed as "without merit" in 1982.
The System/360 bet reveals the paradox that would define IBM for the next sixty years: the company's greatest strategic triumphs always required the willingness to destroy its own most profitable businesses. And the company's greatest failures always stemmed from the inability — or refusal — to do so again.
A corporate bet that created the modern computing industry
1961Watson Jr. greenlights internal planning for a unified computer architecture.
1964System/360 announced on April 7. Development cost: $5 billion (~$50B in 2024 dollars). IBM's annual revenue: ~$3.2 billion.
1966System/360 family begins shipping in volume. Orders exceed capacity.
1969U.S. Department of Justice files antitrust suit against IBM, citing mainframe market dominance.
1970System/370 successor announced, extending the architecture for another decade.
1982DOJ drops the antitrust case after 13 years as "without merit."
Good Design Is Good Business
Watson Jr. did something else that receives less attention than System/360 but may have been equally consequential: he made IBM beautiful.
In 1956, he hired Paul Rand — already one of the most celebrated graphic designers in America — to redesign the IBM logo. Rand replaced the stolid Beton Bold typeface with City Medium, giving the letters a more grounded, balanced appearance. Then, in 1972, he introduced the "8-bar" — the horizontal-striped version of the three letters that became one of the most recognizable corporate symbols on earth. The 8-bar has remained unchanged for over fifty years, a fact almost without precedent in corporate branding.
But the logo was merely the visible tip of a much deeper commitment. Watson Jr. coined the phrase "good design is good business" and established IBM's first formal Design Program, hiring Eliot Noyes as consulting design director. Noyes brought in Charles and Ray Eames to create IBM's famous exhibitions and films. He commissioned Eero Saarinen to design the Watson Research Center in Yorktown Heights, New York — the sweeping, glass-walled masterpiece that would become the birthplace of countless innovations and, eventually, a symbol of both IBM's ambitions and its ossification.
The design philosophy wasn't decorative. It was strategic. Watson Jr. understood that IBM sold to C-suite executives and government administrators — people who equated visual sophistication with institutional trustworthiness. The sleek mainframes, the impeccable documentation, the Saarinen buildings, the Rand logos: all of it communicated that IBM was not merely a vendor but a partner worthy of managing the informational infrastructure of civilization. The design program was, in essence, a trust machine.
This insight — that design is a competitive moat in enterprise sales — would lie dormant at IBM for decades, then be revived in 2012 when Phil Gilbert, an entrepreneur IBM had acquired along with his company Lombardi Software, was tasked with bringing design thinking to the entire organization. Gilbert's challenge was staggering: get 400,000 people to adopt a new methodology when none of them reported to him. His solution was characteristically non-IBM. Instead of mandating adoption from the top, he treated the change program as a product, IBM as a marketplace, and internal teams as customers. Employees opted in rather than being compelled. IBM went on to hire over 1,000 designers to embed in cross-functional teams — the largest design transformation in enterprise history.
The Invention Factory
The relational database. DRAM. The scanning tunneling microscope. The floppy disk. FORTRAN. The hard disk drive. The magnetic stripe card. SQL. RISC architecture. Fractals. The UPC barcode. Deep Blue. Watson (the Jeopardy-winning one, before the name was tarnished). Six Nobel Prizes. More U.S. patents than any other company for nearly three decades straight.
IBM Research, founded in 1945 as the Watson Scientific Computing Laboratory at Columbia University, is arguably the most prolific corporate research institution in history. Across 19 facilities on 6 continents, IBM researchers have contributed foundational work to virtually every branch of computer science, materials science, and applied mathematics. Edgar F. "Ted" Codd's 1970 paper "A Relational Model of Data for Large Shared Data Banks," written at IBM's San Jose Research Lab, didn't just propose a theory — it created a multibillion-dollar industry. The relational database became the standard for processing financial records, personnel data, and logistical information worldwide.
Larry Ellison's Oracle built the first commercial implementation; IBM's own DB2 followed. But the idea — that data relationships should be based on values, not on separately specified linking or nesting — was pure IBM Research.
Ted's basic idea was that relationships between data items should be based on the items' values, and not on separately specified linking or nesting. This greatly simplified the specification of queries and allowed unprecedented flexibility to exploit existing data sets in new ways.
The paradox of IBM Research is that the institution has consistently produced breakthroughs that other companies have commercialized more successfully than IBM itself. The relational database made Oracle. The personal computer made Microsoft and Intel. The hard disk drive became an industry dominated by Seagate and Western Digital. Even IBM's 1970s research into reduced instruction set computing (RISC) would eventually fuel the ARM architecture that powers virtually every smartphone on earth — none of them made by IBM.
This pattern — invent the future, then watch someone else profit from it — is not merely a failure of execution. It is a structural consequence of IBM's business model. A company that generates the majority of its revenue from serving large enterprises has fundamentally different incentive structures than a company built to serve consumers or developers. IBM Research optimized for prestige and patents; the rest of IBM optimized for the next mainframe sale. The gap between the lab and the sales floor was, for much of the company's history, a chasm.
The Day the Mainframe Almost Died
By the late 1980s, IBM was the most valuable company in America. In 1987, its market capitalization peaked at approximately $105 billion. It employed over 400,000 people worldwide. The mainframe business — descendants of System/360, now running on System/370 and System/390 architectures — generated margins that would make a drug cartel envious. IBM's sales force, still operating under Watson Sr.'s quasi-religious culture of customer service, was the most feared in technology. The standard joke was that nobody ever got fired for buying IBM.
Then, very quickly, everything broke.
The personal computer revolution — which IBM had helped launch with the IBM PC in 1981 — turned feral. IBM's army of lawyers had signed a contract allowing Microsoft to sell a version of PC-DOS to any PC maker, paving the way for a generation of PC clones that made Microsoft, not IBM, the dominant platform. Intel supplied the processors. Compaq, Dell, and a dozen others manufactured the boxes. IBM's PC division became a commodity player in a market it had created.
Simultaneously, the minicomputer and workstation makers — Digital Equipment Corporation, Sun Microsystems, Hewlett-Packard — attacked from below, offering cheaper, more flexible alternatives to mainframes. The client-server revolution dispersed computing power away from centralized data centers and toward desktops and departmental servers. IBM's mainframe margins, which had funded everything from research labs to country clubs, began to erode.
Between 1991 and 1993, IBM lost $16 billion — cumulative losses that remain among the largest in corporate history. Revenue, which had peaked near $69 billion in 1990, began a decline that would not truly reverse for years. The stock price cratered from over $43 (split-adjusted) to under $10. Forty thousand employees were laid off, then another thirty-five thousand. The culture of lifetime employment, Watson Sr.'s original covenant with his people, was shattered.
The board, desperate, did the unthinkable: they hired an outsider.
The Elephant Learns to Dance
Louis V. Gerstner Jr. was not a technologist. He was a McKinsey man, an American Express executive, and most recently the CEO of RJR Nabisco — a cigarette and cookie company. When he was appointed CEO of IBM on April 1, 1993 (April Fool's Day, a coincidence the press savored), the conventional wisdom was that Big Blue should be broken up into a dozen or more semi-autonomous units — the "Baby Blues" — each focused on a specific product line. Hardware here. Software there. Services somewhere else. The logic was that IBM was too big, too slow, too burdened by its own complexity to compete against focused attackers.
Gerstner, a Mineola, New York native whose father drove a milk truck, had spent enough time at McKinsey to understand that the conventional wisdom was almost always wrong in exactly the way that mattered most. His first and most consequential decision was to reject the breakup. IBM's greatest asset, he argued, was not any individual product line but the ability to integrate across all of them — to serve as the single vendor that could design, build, implement, and manage a customer's entire technology infrastructure. In a fragmenting industry, integration was the one thing that nobody else could provide at scale.
This was counterintuitive to the point of heresy. Every analyst, every consultant, every business journalist was calling for disintegration. Gerstner bet the company on the opposite thesis: that as technology became more complex, customers would pay an enormous premium for someone to make it all work together.
He was right. But getting there required a degree of corporate violence that would have horrified both Watsons.
Gerstner fired 35,000 employees beyond the 40,000 already gone. He sold off unproductive assets, including real estate and IBM's collection of fine art. He killed OS/2, the operating system IBM had built to challenge Microsoft Windows, accepting that the war was lost and the resources were better deployed elsewhere. He pegged compensation to corporate performance rather than individual results, obliterating the culture of divisional loyalty that had calcified IBM into warring fiefdoms. "People do what you inspect, not what you expect," he said — a sentence that could serve as the epitaph for Watson Sr.'s paternalism.
The strategic pivot was toward services and middleware. IBM would become, in Gerstner's formulation, the impartial integrator — happy to help customers whether or not the hardware bore the IBM name. This required abandoning the bundled-product religion that had defined IBM since the 1960s. It required admitting that the hardware margins were never coming back. It required, in short, the kind of self-destruction that Watson Jr. had practiced with System/360 — except this time, the thing being destroyed was the culture itself.
Services revenue rose from $7.4 billion in 1992 to $30 billion in 2001. IBM's share price went from $13 to $80 during Gerstner's tenure, adjusted for splits. Market capitalization climbed from $29 billion to approximately $168 billion. Gerstner made an early bet on the internet and "e-business" — a term IBM coined and backed with a $500 million marketing campaign — correctly predicting that the web would shift computing power from PCs back toward servers and networks, playing to IBM's traditional strengths.
If I had a vote, the most significant legacy of my tenure at IBM would be the truly integrated entity that has been created. It certainly was the most difficult and risky change I made.
Gerstner's memoir,
Who Says Elephants Can't Dance?, became a canonical text in corporate turnaround literature — the rare CEO autobiography that reads as strategic analysis rather than self-congratulation. He died on December 28, 2025, at 83. Arvind Krishna, in an email to IBM employees, wrote: "His leadership during that period reshaped the company. Not by looking backward, but by focusing relentlessly on what our clients would need next."
The Long Plateau
The story that followed Gerstner is less dramatic and more instructive — a lesson in what happens when a company wins the turnaround but not the next war.
Sam Palmisano, who succeeded Gerstner in 2002, continued the services strategy and pushed IBM toward what he called the "globally integrated enterprise." He oversaw the sale of IBM's PC division to Lenovo for $1.75 billion in 2004 — the final, irrevocable admission that IBM had lost the personal computer market it created. He drove the $3.5 billion acquisition of PricewaterhouseCoopers Consulting in 2002, transforming IBM's Global Services division into a consulting powerhouse of over 100,000 professionals. He set an ambitious financial target — $20 in earnings per share by 2015, the "EPS Roadmap" — that would prove to be both a brilliant focusing mechanism and, eventually, a straightjacket.
The EPS Roadmap worked through financial engineering as much as organic growth. Share buybacks, margin optimization, divestitures of low-margin businesses — Palmisano and his team executed the playbook with mechanical precision. But the focus on near-term earnings came at a cost. While Amazon was building AWS, while Google and Microsoft were investing billions in cloud infrastructure, IBM was optimizing for quarterly performance. Revenue peaked at approximately $106.9 billion in 2011, then began a decline that would last 22 consecutive quarters under Ginni Rometty.
Rometty, who became CEO in January 2012 — the first woman to lead IBM — inherited a company that was financially optimized but strategically adrift. Cloud computing was reshaping the industry. Amazon Web Services, launched as a side project in 2006, was becoming the operating system of the internet. Microsoft under
Satya Nadella would soon pivot Azure into a multi-hundred-billion-dollar cloud platform. Google Cloud was spending lavishly. And IBM, despite its massive services business and vaunted research labs, was late.
Rometty's strategic bet was Watson — the AI system that had famously defeated human champions on Jeopardy! in 2011. The victory was a genuine technological achievement and a marketing masterstroke. But the gap between a game-show AI and enterprise-grade artificial intelligence turned out to be a chasm as wide as the one between IBM Research and the sales floor. Watson was over-promised and under-delivered. IBM marketed it as a solution for healthcare, legal research, financial services — but the technology was not ready for the messy, unstructured reality of enterprise data. Watson Health, the crown jewel of the initiative, consumed billions in investment and was eventually sold in 2022 for an estimated $1 billion — a write-down that symbolized the broader failure to translate research excellence into commercial dominance.
And yet Rometty's tenure was not merely a story of decline. She made a decision that, viewed from 2025, may have saved the company: the $34 billion acquisition of Red Hat in 2019. It was the largest software acquisition in history at the time, and it was a bet — once again, the pattern — that IBM could reinvent itself around a new platform paradigm.
The Red Hat Thesis
Arvind Krishna drove the Red Hat deal. Born in Andhra Pradesh, India, educated at IIT Kanpur and the University of Illinois at Urbana-Champaign, Krishna had spent three decades inside IBM in progressively senior roles — heading IBM Research, pioneering the hybrid cloud business, founding the company's security software unit, even helping create the world's first commercial wireless system. Wired named him one of "25 geniuses who are creating the future of business" in 2016, specifically for his foundational work on blockchain. He was, in every sense, the anti-Gerstner: not an outsider parachuting in to save a dying company, but a lifer who understood the internal mechanics well enough to know where to plant the dynamite.
The Red Hat acquisition was predicated on a simple but powerful thesis: that the future of enterprise computing would not be a single public cloud (AWS, Azure, Google Cloud) but a hybrid environment — workloads running across on-premises data centers, multiple public clouds, and edge computing environments. Customers needed a platform that could operate across all of these environments without vendor lock-in. Red Hat's OpenShift, built on Kubernetes and Linux, was that platform.
The $34 billion price tag was staggering. IBM financed it partly through a $20 billion bond offering — one of the largest corporate debt issuances in history. The 2019 prospectus listed tranches ranging from floating-rate notes to 30-year bonds at 4.25%, a monument to IBM's ability to borrow at scale. Critics called it a desperation move, a declining company mortgaging its future to buy relevance.
Krishna became CEO in April 2020 — in the first weeks of a global pandemic. One of his earliest public acts was a letter to Congress calling on legislators to enact reforms to advance racial justice and announcing that IBM was getting out of the facial recognition business entirely. "IBM firmly opposes and will not condone uses of any technology, including facial recognition technology offered by other vendors, for mass surveillance, racial profiling, violations of basic human rights and freedoms," he wrote. The facial recognition business generated negligible revenue for IBM. The decision was significant precisely because it was symbolic: a declaration that the company's values, articulated by Watson Sr. a century earlier, would be applied even — especially — to technology with morally ambiguous applications. Whether this was genuine ethical leadership or shrewd positioning as the industry's "safe" AI vendor is a question the market hasn't fully resolved.
What followed was a systematic portfolio transformation. In November 2021, IBM spun off its managed infrastructure services business into a new public company called Kyndryl — shedding approximately $19 billion in annual revenue and 90,000 employees. This was the opposite of a growth strategy. It was a deliberate contraction — cutting away the low-margin, labor-intensive services business to reveal a smaller, more focused, higher-margin company underneath.
🔄
IBM's Portfolio Transformation Under Krishna
The strategic narrowing that reshaped a century-old company
2019$34 billion acquisition of Red Hat closes, the largest software deal in history.
2020Arvind Krishna becomes CEO in April. IBM exits facial recognition business in June.
2021IBM spins off Kyndryl, shedding ~$19 billion in managed infrastructure services revenue and ~90,000 employees.
2022Watson Health sold for an estimated $1 billion. IBM declares itself "a software-led, fully integrated platform company."
2023IBM launches watsonx AI platform. Generative AI book of business begins accumulating.
2024Revenue reaches $62.8 billion.
Free cash flow of $12.7 billion. 11 acquisitions closed. GenAI book of business exceeds $5 billion.
The Anti-OpenAI
IBM's AI strategy under Krishna looks radically different from the consumer-facing arms race between OpenAI, Google, Meta, and Anthropic. Where those companies compete on model scale, benchmark performance, and consumer mindshare, IBM has positioned itself as the enterprise AI company — focused not on the biggest models but on the most efficient, most governable, most cost-effective ones.
The watsonx platform, launched in 2023, provides tools for developing AI applications, managing data, and governing the entire lifecycle of AI models. IBM's Granite family of models — open-source, fit-for-purpose, designed to be tuned with proprietary enterprise data — represents a philosophical bet that enterprise customers don't need trillion-parameter frontier models. They need models that are affordable to run, transparent enough to audit, and small enough to deploy on hybrid infrastructure without sending every query to a public cloud.
This is the anti-OpenAI thesis. And it may be exactly right.
The logic runs like this: most enterprise AI use cases are not creative writing or scientific reasoning. They are customer service automation, code generation, document summarization, compliance monitoring — tasks where a well-tuned, domain-specific model running on a customer's own infrastructure will outperform a general-purpose frontier model at a fraction of the cost. IBM claims its Granite models can deliver "up to 90% improved cost efficiency" when tuned with proprietary data. If that number is even directionally correct, it represents a formidable value proposition for CIOs who need to show AI ROI to their boards.
The risk, of course, is that the frontier models get cheap enough, fast enough, that the efficiency argument disappears. If OpenAI or Google can deliver GPT-level performance at Granite-level costs, IBM's positioning collapses. It is, in essence, a bet that the AI cost curve will not drop as precipitously as the hyperscalers hope — or that enterprise customers will pay a premium for sovereignty, governance, and the ability to keep their data off someone else's servers.
Three years ago, we laid out a vision for a faster-growing, more-profitable IBM. I'm proud of the work the IBM team has done to meet or exceed our commitments. With our focused strategy, enhanced portfolio, and culture of innovation, we're well-positioned for 2025 and beyond.
The Machine That Keeps Running
The most underrated fact about IBM is the mainframe.
In 2024, Krishna declared the z16 "the most successful mainframe program in our history." This is a product category that has been declared dead every decade since the 1980s. And yet the IBM Z series — direct descendants of System/360, running on an architecture that traces its lineage to 1964 — continues to process an estimated 68% of the world's production workloads, handles 87% of all credit card transactions, and runs critical systems for 44 of the top 50 banks in the world. These numbers are difficult to verify independently and IBM is incentivized to emphasize them, but even if they overstate the case by half, the mainframe remains one of the most quietly dominant platforms in all of technology.
The mainframe business operates on a cyclical refresh model — new hardware generations every few years, with customers locked into upgrade cycles by the accumulated weight of their mission-critical software running on the Z architecture. The z16, launched in 2022 with an integrated AI accelerator chip, generated sufficient demand that even as the broader infrastructure segment declined 3% at constant currency in 2024 (in line with product cycle expectations), the program's installed base continued to expand.
This is the hidden flywheel. The mainframe generates recurring revenue through hardware upgrades, software licenses, and maintenance contracts. Customers who run their core banking, insurance, or government systems on IBM Z cannot easily migrate without years of effort and billions in cost. The switching costs are measured not in months but in decades. And because the mainframe is the system of record for the world's most regulated industries, IBM's position is reinforced by compliance requirements, audit trails, and the sheer terror of migration risk.
The mainframe also creates the gravitational pull for IBM's software and consulting businesses. Customers who run Z need Red Hat OpenShift to modernize their mainframe applications. They need IBM Consulting to manage the hybrid environments that connect mainframes to cloud workloads. They need IBM security software to protect the crown jewels. Every dollar of mainframe infrastructure revenue generates multiple dollars of software and services revenue downstream.
Quantum's Long Bet
There is one final dimension to IBM's strategy that defies conventional analysis: quantum computing.
IBM has been the most visible corporate investor in quantum technology for over a decade, building superconducting quantum processors, publishing the most quantum-computing research papers of any company, and making its quantum systems available via the cloud through the IBM Quantum Network. In 2023, IBM unveiled the 1,121-qubit Condor processor, then pivoted its roadmap toward modular quantum systems that could be connected to scale beyond the limits of any single chip.
Quantum computing has no significant commercial applications today. The technology is pre-revenue in any meaningful sense. IBM's competitors in the space — Google, Honeywell (via Quantinuum), IonQ — are in the same position. The bull case requires believing that quantum computers will eventually solve problems in drug discovery, materials science, cryptography, and optimization that are intractable for classical computers, and that IBM's early investment will translate into platform dominance when that day arrives.
This is a multi-decade bet with no guaranteed payoff. It is also exactly the kind of bet that IBM has historically excelled at — the System/360 of the quantum age, a massive upfront investment in a platform architecture that, if successful, will create an entirely new computing paradigm. The question is whether IBM's track record of inventing the future and then watching someone else profit from it will repeat.
At the Watson Research Center in Yorktown Heights — Saarinen's glass-walled masterpiece, now home to some of the most advanced quantum hardware on earth — the next generation of IBM's quantum processors are being calibrated in dilution refrigerators cooled to temperatures colder than outer space. The building that was designed in the early 1960s to capture the spirit of technological advancement still does. Whether the company that built it can capture the commercial value is the question that has defined IBM for 114 years.
The z16 mainframe running the world's banking transactions, built on an architecture from 1964. The watsonx platform selling AI governance to enterprises terrified of getting it wrong. The quantum processors in Yorktown Heights, cooled to 15 millikelvins, solving problems that don't exist yet. Three bets, three time horizons, one company. The oldest survivor in the most volatile industry on earth, still running.
