IBM Didn't Expand Into New Markets. It Kept Escaping Old Ones.

In 2014, IBM did something companies almost never do: it paid someone to take a business away. To get out of chip manufacturing, IBM handed GlobalFoundries roughly $1.5 billion to absorb its fabrication plants.⁵ Not a sale — a payment. The world's most storied technology company wrote a check to make a part of itself disappear. That single transaction is the cleanest window into how IBM actually expands beyond its core. It doesn't march into bright new adjacencies from a position of strength. It backs out of burning rooms, and learns to call the exit a strategy.

The official story is a celebrated one: IBM, the great reinventor, repeatedly walked away from commoditizing hardware to claim higher ground in software, services, and the cloud — a masterclass in self-disruption. The truer story is that every one of those famous pivots arrived under duress, with a regulator, a near-death loss, or a shrinking, loss-making unit doing the deciding.

Here is the thesis a smart friend could repeat at dinner: IBM's adjacency moves are not advances. They are monetized retreats — the company has spent six decades getting paid to leave positions it was already losing, and dressing the departure as vision.

The first 'pivot' came with a lawsuit attached

On June 23, 1969, IBM announced it would unbundle — sell software and services separately from its hardware rather than giving them away with the machine.¹ This is now taught as the moment IBM invented the modern software industry, a far-sighted choice to create a new revenue stream. The timeline is awkward for that reading. The Department of Justice had filed an antitrust suit against IBM on January 17, 1969 — five months earlier.¹ The motives are genuinely disputed: IBM has always insisted it was responding to changing business conditions, and serious historians acknowledge both readings are defensible.² But 'visionary expansion' and 'partly coerced concession' are very different stories, and the second one fits the pattern better. The first time IBM broke its own product apart to sell the pieces, it did so with a federal lawsuit sitting on the desk.

The decision that actually mattered wasn't 'go to services'

Skip to the legend everyone knows: Lou Gerstner arrives at a dying IBM and turns the hardware dinosaur into a services powerhouse. The crisis was real and enormous. On January 19, 1993, IBM announced an $8.10 billion loss for fiscal 1992 — at the time the largest single-year corporate loss in U.S. history, part of nearly $16 billion of net losses across 1991 to 1993.³ By the spring of 1993, investment bankers were already placing breakup values on IBM's divisions, pricing the company as parts worth more dead than alive.⁴

Here is the part the 'shift to services' narrative buries. Gerstner's own most important decision, by his own account, was not to chase services — it was to keep IBM together and refuse the breakup the bankers were already underwriting.⁴ The services build-out followed from that structural refusal, not the other way around. IBM had always carried service revenue; what Gerstner did was scale it and reprice it, because a unified IBM needed something to sell across all those divisions he'd just declined to sell off. The pivot, again, was downstream of a near-death event — not the cause of a renaissance, but the consequence of refusing to die. By 2001, the rebuilt company earned $7.7 billion.⁴ The recovery was genuine. The framing — proactive vision — is the part that doesn't survive contact with the dates.

Read the moves as a ledger of exits

Line the famous reinventions up next to what actually drove each one, and a single shape appears. These aren't expansions into strength. They are departures from weakness, timed to the moment the weakness became impossible to carry.

The 2014 hardware exits make the logic naked. IBM sold its Intel-based System x server line to Lenovo for $2.3 billion — and in the same year paid GlobalFoundries about $1.5 billion to take its chip-manufacturing facilities off its hands.⁵ One business it could still sell. The other it had to buy its way out of. When the exit price goes negative, 'focusing on higher-value areas' stops being strategy and starts being damage control with a press release.

The most expensive proof: a spinoff dressed as a strategy

Then comes the cleanest case of all. On November 2, 2021, IBM spun off its managed infrastructure services business as Kyndryl, presented as a forward-looking move to sharpen IBM's focus on hybrid cloud and AI.⁷ The filings tell a blunter story. That same business had been shrinking for years — $21.8 billion in revenue in 2018, $20.28 billion in 2019, $19.35 billion in 2020 — and in 2020 it posted a net loss of about $2.011 billion after charges.⁷ IBM didn't release a thriving unit into the wild. It detached a declining, loss-making one and let the market score the remaining company on cleaner numbers. The spinoff was not the planting of a new business. It was the amputation of an old one.

And the asset IBM kept — the part it actually wanted — it had bought, not built. In July 2019 it closed its acquisition of Red Hat for roughly $34 billion, its largest-ever deal, paying $190.00 a share in cash to buy its way into hybrid cloud rather than win there organically.⁶ The pattern holds even in the supposed offense: when IBM needed a new core, it didn't grow one. It purchased one, then carved away the declining flesh around it.

Isn't 'knowing when to leave' the whole skill?

The fair objection is that this is too cynical. Kodak couldn't leave film; Blockbuster couldn't leave stores; most incumbents die clutching the business that's killing them. IBM, on this reading, is rare precisely because it can let go — and shedding a $2-billion-loss unit while doubling down on software is exactly the discipline that destroys lesser companies when they refuse it. That's true, and it matters. Reactive is not the same as wrong. Gerstner's refusal to break up IBM was a genuine act of judgment, and the company that emerged was real and profitable.⁴

But notice what the steelman concedes. It defends IBM as a brilliant exiter, not a brilliant expander — and those are opposite muscles. A company that compounds from strength keeps finding new ground its core makes it uniquely able to take. A company that survives by exiting keeps discovering, late and under pressure, which ground it should never have been standing on. Both can be valuable. Only one is the story IBM tells about itself. The evidence supports the survival skill and quietly contradicts the expansion myth — and the difference shows up where it always does, in the price. You don't pay $1.5 billion to leave a market you expanded into from strength. You pay it to escape one that was eating you alive.

Give IBM its due: it is still here, software ARR climbing past $23.6 billion in 2025 and a generative-AI book of business already worth billions.⁸ Most companies that faced its 1992 loss are footnotes now. But durability and vision are not the same virtue. IBM's genius was never knowing which markets to enter from strength. It was knowing — usually a beat after the lawsuit, the loss, or the bleeding unit forced the question — which markets to pay its way out of. The company didn't expand beyond its core so much as keep relocating it, one forced retreat at a time. Survival by exit is a real and rare skill. It is just not the heroic forward march the brand keeps selling — and the $1.5 billion check it once wrote to leave a business tells you exactly which story is true.