Inside the Disaster at IBM: The Day Bill Gates Overthrew Big Blue

Wall Street Journal, Eastern edition, New York. Dow Jones & Company Inc. 16 August 1993.

“You think you can sneak around behind my back,” Jim Cannavino, the head of IBM’s personal-computer business, told Bill Gates, the chairman of Microsoft Corp. “But let me tell you, anytime you try to do business with one of my big customers, even before you’re out the door they’re calling me to tell me what you’ve said.” Through clenched teeth, Mr. Cannavino muttered a phrase that would become his refrain: “I wouldn’t want to be in your shoes. I wouldn’t want to have to compete with the IBM company.”

It was the fall of 1989, and Mr. Cannavino had been assigned the job of saving IBM’s personal-computer business, which, company officials were beginning to realize, had compiled a nearly perfect record of strategic mistakes for the previous five years under Mr. Cannavino’s predecessors. IBM’s new line of PCs wasn’t selling. Crucial new software was turning out to be cumbersome and slow. Strategies for selling software applications and peripheral equipment were nonexistent.

By then, IBM and Mr. Gates’s software company were in the ninth year of an increasingly strange business alliance. It had been conceived in misunderstanding, when an IBM executive, on meeting the 24-year-old Mr. Gates, mistook him for Microsoft’s office boy. Microsoft had provided IBM with all-important software for its PCs, but Mr. Gates had retained ownership of the software and grown rich collecting royalties on it. And Mr. Cannavino wasn’t doing any better than his predecessors at IBM had at understanding Bill Gates.

The rivalry between these two men unfolded in an industrial drama that upended the entire computer business. It looked like an unfair contest: a street-smart executive backed by one of history’s great corporations pitted against a waiflike nerd. When it was over, the nerd had become the richest man in America, and IBM was damaged so badly that it had to take writeoffs equivalent to a third of the profits it had made in its entire 80-year history.

What follows is an inside look at the struggle between Mr. Gates and Mr. Cannavino at a critical time in the sudden, sad decline of International Business Machines Corp.; the moment when the balance of power in the computer industry changed, in an embarrassingly public showdown, while the nation’s top software executives looked on agape.

By 1989, the PC explosion had finally begun to undercut demand for mainframe computers, IBM’s bedrock business, which used to provide $4 billion or so a year in profits and routinely made IBM the most profitable company in the world. IBM had fumbled so many opportunities in the PC business that its partners and competitors were making all the money.

Though only a handful of people know it, Mr. Gates had actually offered to let IBM buy the crucial PC software, called DOS, for about $75,000 in 1980, but IBM passed. In 1986, Mr. Gates made IBM another offer that might have given IBM the upper hand: He asked IBM to buy 10% of Microsoft for around $70 million to try to smooth over some problems in the relationship. In another world-beating miscalculation, IBM declined—passing up what could have become a $2.5 billion profit—because it worried about seeming to dominate the PC business too thoroughly.

IBM and Microsoft began their relationship in 1980 at opposite ends of the business spectrum: Big Blue had 340,000 employees and $3.6 billion of annual earnings, and Mr. Gates’s company was a start-up with 32 employees and the slimmest of profits. But by 1989, IBM executives were scratching their heads wondering how to bring their increasingly powerful partner back under control. The job fell to Mr. Cannavino, who inherited the troubled PC division in January of that year.

Mr. Cannavino, then 44 years old, had grown up on the mainframe side of IBM, and once had sneered that if God had meant for man to have personal computers, He would have built them into people’s wrists. Now Mr. Cannavino was having to educate himself in a hurry about the PC business—having to drink out of a fire hose, as they said at IBM.

IBM executives knew they couldn’t just dump the 33-year-old Mr. Gates. They had been trying on and off for six years to come up with a new version of DOS on their own but had repeatedly failed. Knowing that IBM needed a successor to DOS to give PCs more capabilities and pull their business out of a hole, IBM people had learned to put up with Mr. Gates’s unique business style. He was generous to a fault with advice about IBM’s management, telling executives where he thought they should cut back on programmers, how they should focus differently on a particular software problem, and so forth. He didn’t let up. He’d sit in meetings, rocking back and forth with his arms folded across his stomach, firing off ideas in his machine-gun style.

Mr. Cannavino never got used to it. He hadn’t gone to college, and while that made most people view his success at IBM as all the more extraordinary, his lack of formal education made him sensitive to condescension. Mr. Gates didn’t have a college degree either, but he had, after all, dropped out of Harvard. Mr. Cannavino had merely spent time at a technical institute while working at a grocery store and at his family’s pizza parlor outside Chicago.

Mr. Gates, for his part, took Mr. Cannavino for a blowhard. Mr. Cannavino would go on and on about his horses, or his cars, or how IBM was showing American manufacturers how to stand up to the Japanese. Mr. Cannavino might chat for the first 2 1/2 hours of a three-hour meeting, while Mr. Gates strained at the bit to get down to business. Mr. Gates said later: “I kept wanting to say to Cannavino, “We need a shorthand because these meetings are taking too long. Every time you say ’13,’ I’ll know that what that means is that all you want to do is what the customer wants. And for every one of these other gibberish slogans we can also get little numbers. There are lots of small integers available. We’ll just tighten these meetings up. You know, Cannavino, if you want to talk about how you’re going to save the U.S. educational system, OK, we’ve heard that story. That’s a good 15-minute one. That can be No. 11. If, Cannavino, you want to give that speech about how you’ve cut manufacturing overhead and how you’ve done such a great job running things and how you’re such a tough guy, that one we can give a little shorthand, too, because you’re getting good at that little speech.'”

If the relationship between Mr. Gates and Mr. Cannavino wasn’t sufficiently complicated to begin with, they were trying to function simultaneously as collaborators and competitors, tussling over what kind of software would set the standard for the next generation of personal computers.

The problem was this: IBM needed an upgraded “operating system” for its new line of personal computers, called PS/2, the line that was supposed to restore the company’s fortunes. An operating system is the layer of software that translates a computer’s binary native tongue—unwieldy strings of ones and zeroes—into a form simple enough for a programmer to use handily. Once this translating software is available, programmers can decide how to write the “software applications”—spreadsheets, word-processing programs and so on—that make a personal computer worth buying. Without the translating software, the thousands of independent software companies that write all these applications simply can’t do their jobs.

Microsoft had developed the operating system for IBM’s original line of PCs; the system was called DOS, for Disk Operating System. Microsoft had kept the rights to this software, licensing it to IBM and to anyone else who wanted to buy it—an extraordinarily lucrative business.

Now IBM wanted to develop a broad rewrite of DOS to use in its new machines. It had assigned its programmers to write the new software, called OS/2, together with Microsoft. Recognizing the size of the mistake it made when it ceded royalties on the original DOS to Microsoft, IBM wanted to regain control of its PC destiny by bringing the operating system back under its aegis.

But at the same time, Mr. Gates had begun to develop an addendum to his original DOS operating system that would let the new IBM PCs imitate the user-friendly graphics made wildly popular by Apple Computer. He called the addendum Windows.

If a robust Windows took off, it would make OS/2 less attractive.

Like so much else at IBM, OS/2 became a casualty of bureaucratic sclerosis. Some 1,700 programmers were working on OS/2-related projects at four sites on two continents. With all those people working on the project, coordination of their efforts became so tough that the software made little progress.

It only reinforced Mr. Gates’s conviction that IBM was just too cumbersome to compete in the PC market. Later on, he recalled taking a break in an IBM office one day for a bowl of cereal, glancing idly around, and realizing that most of the PCs in use around him were more than seven years old. “This tells me more about IBM than I’ve ever seen,” he said.

As the fall of 1989 began, the problems with OS/2 became a Continental Divide for the entire PC software industry. Hundreds of companies had staked their futures on OS/2 by designing their spreadsheets, word processors and such to run on it. Those applications wouldn’t run if the computer user didn’t have OS/2, so if OS/2 didn’t get its act together fast, the companies were going to have to write off the fortunes they had spent on the development.

But things were even worse than that. It was becoming apparent that Microsoft was developing applications of its own to run on Windows. If Windows took off before OS/2 did, then Microsoft, as the only big company with Windows applications on the market, would clean up. Mr. Gates was already a tough competitor, far larger than any of his rivals. If Windows succeeded, he might become unstoppable. IBM, having essentially guaranteed that it would make OS/2 work, found itself with a lot of Mr. Gates’s unhappy software competitors beating on its door.

Mr. Cannavino held a series of small meetings with software makers, including Jim Manzi, Lotus Development Corp.’s chief executive; Fred Gibbons, chief executive of Software Publishing Corp.; and Dave Liddle, chief executive of Metaphor Computer Systems Inc. Most of the software executives resented Mr. Gates; the few who didn’t resent him hated him. Mr. Gates’s personal wealth had just passed $1 billion, yet he seemed rapacious to his less fortunate competitors. He would use his dominance of the software market to drive the hardest possible bargain.

The meetings were fiery. The software executives would jump all over Mr. Cannavino, saying he’d better cut Mr. Gates and Windows off at the knees. Mr. Cannavino would pace around his conference room, hinting that he had a plan that would take care of Mr. Gates. The crux of it seemed to be the threat that he might be able to use some limited rights IBM had acquired from Mr. Gates to begin selling DOS in direct competition with Microsoft. Rather than just selling DOS installed on its own PCs, Mr. Cannavino hinted, IBM might start selling packages through dealers that would compete with Microsoft’s sales of retail DOS packages. Mr. Cannavino might also try to do deals with clone makers to get them to buy DOS from IBM rather than from Microsoft. DOS was still by far the biggest single piece of Microsoft’s business, so cutthroat competition on pricing would have stung.

Mr. Gates recalls that Mr. Cannavino “kept saying to us, `You can’t do any more work on Windows. You’ve got to wind up your work on Windows.” Mr. Gates says he worried that federal antitrust authorities would regard such talk as showing that IBM and Microsoft were trying to carve up the market for PC operating systems—and the Federal Trade Commission did, in fact, investigate whether Microsoft had acted improperly. Mr. Gates adds: “My lawyers always tell me that if somebody says something like that I should knock over my drink (to get everyone’s attention) and say, `Just remember, I never agreed to that.'”

Mr. Gates was troubled by the deteriorating relations with IBM. He was also very interested in having IBM endorse Windows to help Microsoft get independent software companies to develop applications that would run on Windows—computer users, after all, were more interested in what tasks they could perform on their computers than in what the operating system did. While software companies suspected Microsoft of plotting to corner the market for applications, Mr. Gates insists that he was looking for all the outside help he could get in anticipation of the next Windows version, scheduled for May 1990.

With the PC industry abuzz over the obvious friction between IBM and Microsoft, Mr. Cannavino and Mr. Gates decided they needed to say something about the state of their relations in early November 1989 at Comdex, a trade show in Las Vegas where all the industry’s major players—and some 100,000 other people—gather every year. Both sides kept sending teams of staff people to hammer out issues, alternating between Redmond, Wash., where Microsoft is based, and offices in Westchester County, N.Y., satellites of the IBM headquarters in Armonk. Mr. Gates and Mr. Cannavino often flew off to neutral sites in the middle of the country for one-on-one meetings to break deadlocks.

With time running out, they finally reached a very tentative compromise. Mr. Cannavino would give Mr. Gates the endorsement of Windows that Mr. Gates craved. In return, Mr. Gates would state that Windows was aimed at low-end personal computers while OS/2 was the industrial-strength operating system that should be used on more powerful PCs. And he would concentrate Microsoft’s efforts on developing OS/2, de-emphasizing refinement of Windows.

The compromise came too late to allow for a full-fledged contract, which made both sides a little nervous. But both decided they would go ahead, assuming that things went well at a final meeting they planned for Sunday morning at the Hilton Hotel in Las Vegas. Mr. Gates and a few other Microsoft executives spent all day Saturday and much of the night pounding out details of their version of the compromise with IBM.

The meeting at the Hilton was a success; the compromise held. So, everyone then went off to a dinner that had been arranged with 25 or so top executives from the biggest PC software companies, a dinner that was to be a turning point for IBM’s fortunes and for the whole computer industry.

The dinner had been arranged hastily, once it appeared that a compromise was possible, but IBM had worked it hard. A senior PC executive had called and personally invited each of the software executives. For each guest, IBM arranged to have one of its senior executives attend the dinner as a sort of handler—glomming on to the software executive, sitting next to him at dinner, chatting him up. IBM planned the seating carefully, to make it clear which executives were considered friends and which were seen to be too close to Microsoft. The tables were arranged in a U shape, with Mr. Cannavino and Mr. Gates next to each other at the bottom of the U, with IBM’s friends along the Cannavino leg of the U and with Microsoft’s friends along the Gates leg.

The dinner struck everyone as odd right from the beginning. So many restaurants had already been booked by other companies arranging big Sunday night dinners at Comdex that the IBM organizers could only find room at Chateau Vegas, a place that was tucked in behind the convention center and that, with its garish red-and-black wallpaper, made some of the executives think they’d walked into an economy-class cocktail lounge. The room was too small and the ceiling was low; the 60 executives there felt as though they were sitting on top of each other.

Mr. Gates and Mr. Cannavino were sitting in a sort of booth at the head of the table, making it hard for them to stand up and speak. A noisy air conditioner at the far end of the table dripped on those sitting underneath it and drowned out most of what Mr. Gates and Mr. Cannavino said. A sound system had been installed, but it worked only intermittently. Dinner consisted of plates piled high with steak, chicken and lobster, far more food than anyone could eat and more than anyone even cared to look at. The IBM handler sitting next to Fred Gibbons of Software Publishing was so drunk that those around him kept watching to see whether he’d fall face down into his mound of food.

Mr. Cannavino and Mr. Gates talked briefly to lay out the compromise. They stuck pretty much to the script, but neither could quite bring himself to go all the way. Mr. Cannavino never really endorsed Windows, and Mr. Gates never completely said that Windows development would be curtailed from then on. Each was annoyed at what he saw as the other’s evasions.

Questions were then invited. There were only a few, but they were telling. One person asked Mr. Cannavino point-blank whether he was endorsing Windows. Mr. Cannavino fudged. Mr. Gibbons, an IBM ally, then stood up and said to Mr. Gates: “Bill, you walked into my office a while back and told me I should be developing applications for Windows. Are you now walking into my office and telling me I should be developing applications for OS/2 instead?” Mr. Cannavino assumed the answer was going to be, “Yes.” Instead, Mr. Gates gave a long, complicated answer that indicated he thought software developers should do applications both for Windows and for OS/2.

Dave Liddle, the chief executive of Metaphor Computer, sitting next to Mr. Cannavino, tried to clarify, to hold the compromise together. “OK, Bill,” he said. “But if you were advising the people in this room about how they should spend their development dollars, you’d suggest they spend them on OS/2, right?” Mr. Cannavino waited, assuming that he’d finally hear a simple affirmative. Instead, Mr. Gates said, “The people in this room are all the heads of their own software companies. They’re all bright people who are fully capable of making up their own minds. I wouldn’t presume to tell them how to run their businesses.”

Mr. Cannavino was livid. The software-industry executives who had met with him ahead of time were stunned. Microsoft wasn’t going to cut off Windows, they thought. Instead, it seemed determined to make Windows succeed, and it now seemed that Windows would become popular long before OS/2. Mr. Cannavino had seemed to think he could bring Mr. Gates to heel, but the software executives suddenly realized that Mr. Gates had become too powerful even for IBM to handle.

“That’s when I knew the game was over,” Mr. Gibbons says. “The numbers were on the scoreboard, and Microsoft had won.” Mr. Manzi of Lotus, annoyed at how things were going, got up and walked out.

When the dinner broke up and some people gathered around Mr. Gates, Mr. Cannavino tried to shove through the group to demand an explanation. But a Gates assistant headed Mr. Cannavino off, promising that they’d all get together to talk later. Mr. Gates, in the middle of a crowd, wandered off, oblivious to Mr. Cannavino’s heated concerns.

When the software executives went outside, they found that the IBM organizers had slipped up. There was a long line of limousines outside waiting to take the IBM executives back to the Las Vegas Hilton but no cars for anyone else, and the restaurant was far enough out of the way that taxis were hard to come by. The IBM executives trooped out, piled into their limos and drove off, leaving an important group of potential allies stranded on the curb. Several gave up and hiked back to the Hilton. They gathered in the bar to hash through what they had just witnessed and, finding a group of reporters camped out there, lamented in front of them about how the IBM-Microsoft compromise had fallen apart.

Throughout the next day, Monday, Mr. Gates and Mr. Cannavino struggled to patch up their differences. They settled into a trailer in the parking lot behind the convention center, where Mr. Cannavino had a mobile office—complete with big desk, plush carpeting, a bathroom, several phone lines and room for several assistants. Mr. Cannavino didn’t want his executives to miss out on the comforts of home while they worked at the trade show, so he had outfitted eight lavish trailers and had them driven cross country to Las Vegas from New York. Mr. Gates went all the way through the day and most of the way through the night, sometimes meeting just with Mr. Cannavino and sometimes getting caught alone with as many as nine IBMers. When he left the trailer shortly before dawn, it seemed that the compromise was back in place.

Mr. Cannavino, however, still wasn’t convinced. He didn’t quite see what all the fuss was about Windows. To a mainframe guy like Mr. Cannavino, Windows felt like a toy operating system. Mr. Cannavino assumed customers wanted something solid, something that wasn’t glitzy, something that felt like a mainframe—in other words, something like OS/2. So Mr. Cannavino didn’t want software developers or customers thinking he was endorsing Windows, despite the compromise he had worked out with Mr. Gates. In the late morning, some of Mr. Cannavino’s handlers began running him through a series of mock questions that they thought might come up at the news conference that afternoon. The first question was, of course: “Aren’t you really saying that you’re endorsing Windows?”

“Hell, no!” Mr. Cannavino exclaimed.

One of the questioners then said, “It looks like a duck, it walks like a duck and it quacks like a duck, so it must be a duck”—meaning that Mr. Cannavino certainly seemed to be backing Windows.

Mr. Cannavino was astonished. He had somehow convinced himself that he was offering the mildest of endorsements for Windows, and that everyone in the industry would still understand that OS/2 was the future. He glanced around the room at his fellow IBMers and said that if even one person might interpret his comments as full-fledged support for Windows, then perhaps he should cancel the news conference. Even when Mr. Cannavino left his trailer and started walking through the convention center, he was telling his assistants he wasn’t sure he should proceed with the event. When they arrived at the entrance to the news conference room, they discovered just how many people wanted to hear Mr. Cannavino’s position on Windows. The room was supposed to hold only 140 people, but 800 people showed up to watch the fireworks—all jockeying for position as they tried to cram into the room. Some of Mr. Cannavino’s staff were forced to call security guards to keep the restless crowd under control. Mr. Cannavino had to shove his way through the crowd as he approached the door.

Mr. Gates, coming along a few minutes later, saw the huge crowd in the hall and wondered aloud whether he and Mr. Cannavino should just hold the news conference in the corridor, where there was much more room. But the IBMers were set up inside, so they weren’t about to alter their plans.

As Mr. Gates stood at the podium, Mr. Cannavino, arms crossed and a scowl on his face, was standing 15 feet away. It seemed Mr. Cannavino wanted to put as great a distance as possible between himself and Mr. Gates. Even when Mr. Gates invited him to the podium, Mr. Cannavino never got closer than six or seven feet. The two adhered to their script on the compromise—with Mr. Cannavino offering halfhearted support for Windows and Mr. Gates holding out the prospect that OS/2 was the future—but things fell apart once the questions began. The strains between them became apparent, just as they had to those who attended the dinner two nights before. Of course, the first question was whether IBM was finally committing itself to installing Windows on its hardware and encouraging software developers to write applications that would run on Windows. Mr. Cannavino swallowed hard, shifted his weight and gave a rambling, 400-word nonanswer. When someone followed up with a question to try to pin Mr. Cannavino down, he responded with an even longer equivocation.

Mr. Cannavino left the news conference sure he had artfully avoided any public commitment to Windows, but he seemed to be the only one in the industry with that impression. Mr. Cannavino had done irreparable damage. He had lost face with the software industry and unintentionally pushed them to abandon OS/2 and embrace Windows—which, the following year, would take the world by storm.

Mr. Cannavino also left the room thinking he retained at least modest control over Microsoft, people on his staff said later. But he was alone in that opinion, too. Many of the reporters covering the conference had been at one of the bars in the Hilton two nights before, listening to executives from independent software companies grouse about how Mr. Gates had run roughshod over Mr. Cannavino at dinner. Now it was becoming clear to everyone in the industry—at least to everyone outside IBM. Mr. Gates was taking control of the direction of the software industry, and soon he would be unstoppable.