IBM

When IBM sold its PC division to Lenovo in 2004, critics called it the death of an American icon. The ThinkPad was synonymous with serious computing. But CEO Sam Palmisano saw what others couldn't: PCs had become a commodity business with razor-thin margins, while services and consulting offered 25%+ gross margins and recurring revenue. The sale wasn't a retreat — it was the final move in a decade-long strategic transformation that doubled IBM's earnings per share within five years.

On July 14, 2026, IBM lost about a quarter of its value in a single session — its worst day on record, worse than Black Monday 1987 — after warning that second-quarter revenue would miss. Roughly $67 billion gone in a day. Everyone reached for the obvious story: another old-tech giant left behind by AI. The warning says something stranger. IBM isn't being passed by the AI boom. It's being billed for it.

The story is that IBM brilliantly reinvented itself across software, services, and cloud. The pattern in the filings tells another one: a $8.1B loss, a chip plant it paid $1.5B to give away, a loss-making unit spun off in 2021 — each pivot a managed retreat, not a bold advance.

IBM tells a story of three coherent moonshots: Red Hat, Watson, and quantum. The truth is asymmetric - a $34B adjacency that compounds, a healthcare 'revolution' it quietly sold for around $1B, and a quantum 'utility' claim that independent labs poked holes in within weeks.

IBM spent $34 billion on Red Hat in 2019, a 63% premium over the share price. The story is that it was a cloud bet. The truth is more uncomfortable: after 22 straight quarters of shrinking revenue, IBM was buying its way back into a conversation it had been losing.

IBM's two great exits look unrelated — a 2005 sale to Lenovo, a 2021 spinoff. They're the same move twice: shed a big, shrinking, margin-poor unit to re-rate the rest. The PC division had lost ~$965M over four years before it left.

IBM called Kyndryl's spinoff a liberation. The numbers say escape: a $19.7B managed-infrastructure business in multi-year decline, cut loose so the parent could become the $34B Red Hat bet it had already made. Kyndryl left home posting a $2.3B loss.

The legend says IBM naively handed the PC to the world. The real story is colder: an open design was the only way to ship in a year, and a single non-exclusive license cost IBM the platform - and roughly 80% market share in 1982-83, fallen to about 20% a decade later.

The legend says IBM naively gave the PC's brain to a 24-year-old. But IBM's own negotiator said the quiet part on record: IBM wanted Microsoft to own the software, to dodge lawsuits IBM kept losing. The give-away was a strategy, not a blunder.

The legend says Watson Jr. risked $5 billion on a single heroic gamble in 1964. The real decision was made two years earlier by a committee of thirteen — and the famous 'you bet your company' line was never even his.

IBM spent more than $4 billion building Watson Health, then sold the assets for $1.065 billion in cash. The failure wasn't that AI couldn't do medicine. It was that IBM marketed a system trained on synthetic patients as if it ran on real ones.

In 1993 IBM was bleeding cash — an $8.1 billion loss in one year alone — and the board's plan was to chop it into pieces. Lou Gerstner did the opposite, and that single refusal was the comeback. Market value went from $29 billion to $168 billion.

IBM is praised for serial reinvention - mainframes to services to cloud. But each pivot followed a $5 billion loss or a years-long decline, not foresight. The pattern isn't vision. It's the same captive base, milked until it broke.