It was the most valuable company on earth. Now it’s the 85th.

The System/360 in 1964 was a bet on compatibility: one instruction set across an entire product line, so a customer could upgrade without rewriting anything. It worked so well that it became a moat. Once a bank's ledgers ran on 360 architecture, leaving was not a purchasing decision, it was a decade-long project.

By the mid-1970s the U.S. government had been suing IBM for monopolisation for years — a case filed in 1969 and abandoned in 1982, after thirteen years. The lawsuit is the clearest measure of the position: the state believed IBM's dominance was structural, not temporary.

In 1985 the company employed roughly 405,000 people and was the largest single component of the S&P 500 — as it had been for most of the decade. Owning the American stock market meant, to a first approximation, owning IBM.

Six dollars of every hundred invested in America's largest companies were IBM. Today it is about thirty cents. Nothing was seized; the index simply grew around it.

Speed was the goal. A small team in Boca Raton shipped the IBM PC in about a year by buying nearly everything off the shelf: an Intel processor, and an operating system licensed — non-exclusively — from a 32-person company called Microsoft. Both suppliers were free to sell the same parts to anyone.

They did. Compaq reverse-engineered the BIOS in 1982, and within a decade the industry standard was “IBM-compatible” rather than IBM. The hardware became a commodity with commodity margins; the profit pooled in the processor and the operating system. IBM sold the PC business to Lenovo in 2005 for $1.75 billion.

This is the single most expensive decision in the history of corporate strategy, and it is worth being precise about the price. Not the lost PC revenue — the platform.

The company that wrote the operating system for IBM's machine is now worth about {{ msRatio }} times the company that commissioned it.

In 1993 IBM reported the largest annual loss in American corporate history to that point. Breakup was the consensus plan. Louis Gerstner arrived from RJR Nabisco, kept the company whole, and argued that its scarce asset was not machines but the ability to make a customer's whole estate work.

He was right, and the turnaround is real: IBM became a services and software company, and for a decade it was a very profitable one. But the pivot also fixed the company's centre of gravity on selling labour and maintenance to the installed base — a business that grows with a client's IT budget, not with the industry.

Every subsequent chapter follows from that. When the next platform shift arrived, IBM's instinct was to service it for others rather than own it.

Revenue peaked in 2011 and fell for the better part of a decade, while earnings per share kept rising. The shaded years mark the Aptiva home PC: pulled from US retail shelves in October 1999, discontinued in May 2001. IBM has had no home-market product since; the remaining PCs went to Lenovo in 2005.¹⁰

The 2010 “Roadmap 2015” promised $20 of operating earnings per share by 2015. It was a commitment to a per-share number, not to a business, and there are three ways to hit it: grow, cut, or shrink the denominator. IBM used the last two hardest.

Roughly $176 billion went to buybacks over two decades — capital that Amazon, Google and Microsoft were spending on data centres in exactly the same years. The roadmap was abandoned in 2014. Revenue then declined for 22 consecutive quarters.^4,9

This is the mechanism worth understanding as an investor: nothing here was fraudulent or even unpopular. Shareholders asked for capital returns and got them. The compounding just went to the wrong place.

In February 2011, Watson won at Jeopardy! on national television. For a moment IBM owned the public idea of artificial intelligence outright — years before anyone outside research had heard of a transformer.

What followed was a brand rather than a platform. Watson was sold to hospitals as an oncology adviser; MD Anderson's project was shelved after roughly $62 million.⁹ Watson Health, assembled through billions in acquisitions, was sold to a private-equity firm in 2022 for a reported ~$1 billion.

The lesson repeats the PC lesson at a different layer. IBM had the demo, the research bench and the brand; it did not build the developer platform underneath. When the actual AI wave arrived, the ecosystem had already assembled somewhere else.

{{ wallNote }}

A classical computer simulating quantum mechanics pays an exponential tax. A quantum computer does not. That makes chemistry, materials and certain optimisation problems the first real targets: catalysts, batteries, drug binding — problems where a single answer can be worth more than the machine.

IBM's edge is the thing it failed to build twice before: the platform. More than 90 quantum systems deployed, a network of over 340 organisations running real workloads, and Qiskit — used by roughly 70% of quantum developers, with over four trillion circuits executed.^5,6

And it is now funded like a priority rather than a research line: more than $10 billion committed over five years in June 2026, plus $1 billion of cash into Anderon, a dedicated quantum wafer foundry backed by the U.S. Department of Commerce.⁶