IBM Sold the PC and Spun Off Kyndryl. Both Were the Same Trade.

Sixteen years apart, IBM ran the same play and almost nobody noticed it was the same play. In May 2005 it handed the business that put the IBM logo on millions of desks — the ThinkPad, the Personal Computing Division — to a Chinese company called Lenovo.³ In November 2021 it cut loose a $19-billion services arm, christened it Kyndryl, and gave the shares away to its own stockholders.⁶ One was a sale to a buyer. One was a spinoff to nobody. They look like different decisions made by different IBMs. They were the same trade twice.

The official story is that IBM was pruning — trimming a tired commodity here, a low-growth service there, to keep the portfolio fresh. The real story is narrower and more deliberate: IBM was deliberately throwing revenue overboard. Both units were big. Both were shrinking or bleeding. And both dragged down the one number IBM had decided to be valued on — margin. The genius and the cost of these exits are the same fact: IBM chose to be smaller on purpose.

The trade hiding inside both deals

Here is the move, stripped to its mechanism. A diversified company is priced as a blend. The market takes its high-margin, fast-growing parts and its low-margin, slow-or-shrinking parts and gives the whole thing one multiple — a weighted average that punishes the good businesses for sitting next to the bad ones. Cut out the worst-rated unit and two things happen at once: revenue falls, which looks like failure, and the average margin and growth rate of everything that remains rises, which is the entire point. You accept a smaller top line today to buy a higher multiple on a cleaner business tomorrow. That is a cannibalization trade — eating your own revenue to feed your own valuation.

The PC division was the textbook case, because it wasn't merely low-margin — it was losing money. IBM's own SEC-filed financials for the unit showed operating losses of $397M in 2001, $171M in 2002, and $258M in 2003, plus another $139M in the first half of 2004.² That is roughly $965M of operating losses over four years, and it kept happening even after IBM outsourced most of the manufacturing in 2002.⁵ The popular telling that the PC unit was 'barely breaking even' is too kind by about a billion dollars. IBM wasn't trimming a flat business. It was stopping a hemorrhage — and getting paid to do it.

What the $1.25 billion headline left out

The deal announced in December 2004 is remembered as a $1.25 billion sale, and that number is real but partial — it covers only the cash and equity Lenovo paid.¹ On top of it, Lenovo assumed roughly $500M of IBM liabilities, which is why contemporaneous reporting put the total transaction at about $1.75 billion.⁴ And IBM didn't fully walk away: it took an 18.9% equity stake in Lenovo at closing.¹ So the structure wasn't 'sell the loser and leave.' It was 'hand off the operating burden, keep a slice of the upside, and book the rest as someone else's problem.' The combined PC business carried about $12B of annual revenue and 19,000 employees out the door — revenue IBM was, quite literally, happy to lose.¹

Kyndryl rhymes with all of it. IBM didn't sell it; it distributed the shares tax-free to existing stockholders on November 3, 2021, which is just a different financial wrapper around the same intent — get the unit off IBM's books without triggering a tax bill.⁶⁸ And the registration statement said the quiet part in plain corporate English: both companies would each have 'a more focused business better positioned to invest more in growth opportunities.'⁸ 'More focused' is the polite word for what was happening to IBM's reported numbers. Strip out a roughly $19B unit and the rest of IBM instantly looks more like a software-and-cloud company, because it suddenly is one — by subtraction rather than by growth.

Both businesses got worse the moment they left

Here is the detail that turns this from a tidy story into a real one. IBM did not spin off growing assets and then watch them flourish elsewhere. It cut loose units that were already in decline — and the decline continued on the other side of the door. Kyndryl is the cleanest example: its revenue had been falling inside IBM for years, from $21.8B in 2018 to $20.3B in 2019 to $19.4B in 2020, and after the spin it kept right on falling, to about $18.3B for the fiscal year ending March 2022 and roughly $17B the year after.⁷ IBM didn't free a trapped growth engine. It identified the slowest car in the train and uncoupled it.

Isn't this just dressing up failure as strategy?

The honest objection is sharp: if you keep cutting the parts that don't grow, you don't have a strategy, you have a company slowly disappearing — and re-rating the remainder is a magic trick, not a business. There's truth in that. A cleaner average is not the same as more profit, and a company that grows mainly by subtraction eventually runs out of things to subtract. But the objection misses what made these specific cuts defensible. The PC unit wasn't slow growth; it was a documented cash drain that even outsourced manufacturing couldn't fix.²⁵ Kyndryl wasn't a misjudged keeper; it was a unit already shrinking three years running, whose decline simply continued in independent hands.⁷ Cutting a part that is destroying value frees real capital and management attention for software and cloud — and choosing which businesses to no longer be in is one of the few genuine strategic decisions a giant company has left. The trade is only a trick when you cut something that would have thrived. IBM, by the numbers it filed itself, mostly didn't.

IBM is the company that taught the world to type on a keyboard with its name on it, and then decided that owning that keyboard made it worth less. That is the whole logic of both exits: in a portfolio priced as a blend, your biggest, most famous business can be the one quietly holding down your value, and the bravest financial move is to let it go to someone who wants it more — or to no one at all. IBM ran that calculation twice, in 2005 and in 2021, and twice it chose to be smaller, cleaner, and harder to copy. The PC and Kyndryl both kept shrinking after they left. That wasn't IBM's failure of foresight. It was the proof IBM had read the train correctly.