In 1987, IBM staked its future on a machine designed to take back the personal computer standard it had unintentionally set free. The Personal System/2, launched on April 2nd of that year, was the most expensive proprietary gamble in the history of personal computing. The company wagered its market leadership and hundreds of millions of dollars on the belief that the industry would follow wherever it led. The industry refused.
The PS/2 was discontinued in 1995, and IBM eventually sold its entire PC division to China’s Lenovo for $1. 75 billion. Yet the small round six-pin connector IBM engineers designed in Boca Raton, Florida for that machine still ships on motherboards today. It outlived the computer, the division, and the very idea that any single company could own the personal computer.
To understand why IBM built the PS/2, you have to understand what IBM was before the personal computer existed. In 1980, International Business Machines was the most dominant technology corporation the world had ever seen. Its market capitalization would peak at roughly $90 billion by the middle of the decade, making it the most valuable company on Earth. It employed 405,000 people.
Its mainframe computers ran the banking systems, the airline reservations, the defense logistics, and the government payrolls of the Western world. Nobody ever got fired for buying IBM. That was not a joke. It was a description of market gravity.
But by the end of the 1970s, small personal computers from Apple, Commodore, and Radio Shack were showing up on desks, and IBM had nothing to put beside them. The company’s internal product development cycle ran four years long. Personal computing was moving in months. A lab director in Boca Raton named William Lowe walked into IBM’s corporate management committee in July 1980 with an idea the company had never tried.
Do not build it the IBM way, he told them. Buy the processor from Intel. License the operating system from a small software company in Seattle called Microsoft. Publish the architecture so outside developers could write software and build hardware for it.
And ship it in twelve months, not four. Lowe assembled a small team under a project codenamed Chess, installing them in a low-rise building on the Boca Raton campus, away from the layers of oversight that governed everything else IBM built. He handed day-to-day leadership to an engineer named Don Estridge. Estridge was informal in a company that measured hierarchy by the whiteness of your shirt, impatient with process in an organization that elevated procedure to religion, and personally convinced that a computer should be something an individual owned and enjoyed.
He took Lowe’s plan and built it in a year. On August 12th, 1981, IBM unveiled the PC at the Waldorf Astoria Hotel in New York. The company expected to sell perhaps 200,000 units over three years. Within the second year, customers were buying 200,000 a month.
First-year revenue reached $1 billion. By 1983, IBM held roughly 80 percent of the personal computer market. The three letters on the case were not just a brand. They were the standard itself.
But Estridge had built a door into the machine that could not be closed. The architecture was open. Intel sold the same processor to anyone with a purchase order. Microsoft licensed DOS to anyone who asked.
The only proprietary component was the BIOS, a small program burned onto a chip that handled the lowest level hardware communication. IBM held the copyright on its BIOS, but copyright protects expression, not function. If you could write a BIOS that performed every function IBM’s did without copying a single line of its source code, you could build a machine that ran every IBM program and owed IBM nothing. In February 1982, three engineers who had left Texas Instruments sketched a business plan on a placemat at a Houston pie shop.
Compaq hired engineers and divided them into two teams using what the industry came to call a clean room process. One team studied the IBM BIOS and documented every behavior. That team never wrote a line of code. A second team, completely isolated from the first, took the functional documentation and wrote a new BIOS from scratch.
The result was 95 percent compatible with IBM’s software library. The Compaq Portable shipped in November 1982, and by 1985 Compaq’s annual revenue had reached $53 million. In 1984, a small firm called Phoenix Technologies industrialized the process, building a compatible BIOS and selling it to any manufacturer who wanted to build a PC clone. The floodgates opened.
Dell, AST, Epson, Zenith, Tandy, and dozens of smaller assemblers began shipping machines that ran every IBM program at a significantly lower price. By 1986, IBM’s share of the PC market had collapsed from 76 percent to 26 percent in three years. The clones had not defeated IBM. They had become IBM at two-thirds of the price.
On August 2nd, 1985, Don Estridge boarded Delta Airlines Flight 191 in Fort Worth, Texas with his wife Maryanne. The Lockheed L-1011 hit a microburst on approach to Dallas–Fort Worth International Airport and crashed short of the runway. 137 people died. Estridge was 48 years old.
The engineer who had championed the open architecture, who had fought the corporate layers to keep the PC team fast and autonomous, was gone. The bureaucracy he had held at arm’s length for four years closed around his division like water filling a space. By 1986, the strategic consensus inside IBM’s Armonk headquarters was clear. The problem was not the clones.
The problem was the architecture. The openness that Don Estridge had built into the original PC was the wound. The only cure was a new machine with a new bus behind a new wall of patents and licensing fees that would pull the industry back under IBM’s roof. It was mainframe logic applied to a market that had already learned to live without the mainframe.
The machine that barrier would live inside was called the Personal System/2. On April 2nd, 1987, IBM unveiled the PS/2 in Miami. Four models launched, ranging from the entry-level Model 30 at $1,695 to the flagship Model 80 running Intel’s 386 with a starting price of $6,995 that climbed past $10,000 with options. The machines were redesigned from the ground up: smaller cases, 3.
5-inch floppy drives, integrated controllers, a new video standard called VGA that pushed resolution to 640×480 pixels in 16 colors, and a new 15-pin analog connector that would become the monitor standard for the next two decades. The memory sat in a new 72-pin module format. Keyboard and mouse plugged into those new six-pin mini-DIN connectors. But the strategic core of the PS/2 was not on the surface.
It was inside the expansion slots. Every model above the entry-level Model 30 ran Micro Channel architecture, a new proprietary bus designed by IBM engineer Chet Heath. Micro Channel was by every technical measure superior to the old ISA bus. It operated at 10 megahertz instead of 8.
It supported 32-bit data paths and bus mastering. It offered shared interrupt lines and a software-based card configuration system that replaced the old jumpers with something approaching what the industry would later call plug-and-play. It was also completely and deliberately incompatible with every expansion card the ISA world had ever produced. Every peripheral manufacturer who wanted to sell hardware for the PS/2 would need to design new products.
And to do that, they needed a license from IBM at a royalty of one to five percent of revenue. IBM had fenced the Micro Channel design with patents and made clear it was prepared to enforce them. The industry IBM had created looked at the proprietary bus, looked at the licensing fees, looked at the patent wall, and said no. In September 1988, seventeen months after the PS/2 launch, Compaq led a coalition of nine PC manufacturers in announcing a competing bus standard called EISA, the Extended Industry Standard Architecture.
EISA matched most of Micro Channel’s technical features, but it did something Micro Channel deliberately refused to do. It was backward compatible with every existing ISA expansion card. And it was licensed royalty-free. The message to IBM was blunt and public.
We do not need your bus. We do not need your patents. We do not need your permission. The PS/2 sold respectably in its first eighteen months, moving three million units by September 1988.
IBM was spending over $500 million a year on advertising for the line. But sales volume could not change the structural math. The Micro Channel bus had split the peripheral market into two incompatible worlds, and the smaller world was IBM. Third-party expansion card manufacturers followed ISA and EISA because that was where the volume was.
A buyer who chose a PS/2 walked into a store with a fraction of the expansion options available to someone who bought a Compaq or a Dell. The innovations that had nothing to do with the bus war told a different and more generous story. VGA was a genuine leap, adopted instantly and universally by clone makers and monitor manufacturers alike. The PS/2 keyboard layout set the template every PC keyboard has followed since.
The six-pin mini-DIN ports would carry the PS/2 name into the next century. The 72-pin memory module became the RAM standard for an entire generation. The 3. 5-inch floppy format replaced the old 5.
25-inch disc across the industry within two years. The clone makers took every one of those innovations. They took everything that was not locked behind a patent wall and left the Micro Channel bus on the other side of it. The market did not reject IBM’s engineering.
The market rejected IBM’s control. And then Microsoft delivered the blow the hardware rivals alone could not. IBM had commissioned Microsoft to build an operating system for the PS/2 era, a protected-mode multitasking system called OS/2 that was meant to be the software half of the lockdown. But OS/2 version 1.
0, released in late 1987, was text-only, unstable, demanded expensive memory upgrades, and cost $340 on top of a machine that already carried a steep premium. Meanwhile, Microsoft was building Windows. Windows 3. 0 launched in May 1990 and ran beautifully on cheap ISA clones that cost half what a PS/2 did.
Microsoft sold two million copies in the first six months. The IBM–Microsoft partnership came apart publicly in 1992. OS/2 peaked at roughly five to six percent of the operating system market. Windows reached ninety.
The PS/2 did not collapse in a single quarter. It eroded over years, and the erosion dragged the rest of IBM’s personal computer business down with it. The company that had employed 407,000 people at its peak began cutting. In 1992, IBM posted a net loss of $5 billion.
In 1993, it posted an $8 billion loss, at the time the largest annual loss in American corporate history. John Akers was forced to resign in January 1993, and the board, in a move without precedent in IBM’s history, hired an outsider. Lou Gerstner, a former RJR Nabisco executive with no background in technology, became the first external chief executive the company had ever had. The PS/2 line was quietly discontinued in July 1995.
IBM reverted to building machines on the industry-standard ISA architecture it had tried to replace. In December 2004, IBM announced the sale of its entire personal computer division to Lenovo for $1. 75 billion, less than two percent of its peak market capitalization. A company that once held 80 percent of the PC market sold the whole operation for roughly what a few good months of mainframe revenue were worth.
IBM did not die. It reinvented itself as a services, software, and consulting firm, and it remains profitable today. But the personal computer left IBM’s hands for good. The machine is gone, but the things it introduced are not.
The VGA connector shipped on monitors, laptops, and projectors for more than 25 years after the computer was discontinued. The 3. 5-inch floppy format outlasted the PS/2 by nearly a decade. The 72-pin memory module became the RAM standard for an entire generation of PCs.
The 101-key keyboard layout is the keyboard layout the world types on today. And the six-pin mini-DIN connector, the PS/2 port, is still manufactured and still soldered onto gaming motherboards, still used by competitive players and IT professionals who value the fact that it handles input on a hardware interrupt rather than through USB polling. Nearly four decades after the computer it was named for shipped, the PS/2 port remains one of the oldest active interface standards in personal computing. It carries the name of a machine that was built to close a door that should never have been opened, by a company that could not accept that the door was the whole point.
IBM invented the open personal computer. IBM watched the open standard escape and become the most successful hardware platform in history. IBM built the PS/2 to take that standard back. IBM did not lose the personal computer because it lacked engineering.
The PS/2 was full of innovations good enough to become industry standards for decades. IBM lost the personal computer because it tried to own what had already become everyone’s. The market kept exactly what it wanted from the PS/2 and discarded the rest, including IBM itself.
The small round port that still ships on motherboards forty years later is the proof of it.
