Before Control Data came along, IBM had been at the top of the supercomputing game. Our flagship project in the late ’50s was a machine called STRETCH that grew out of a contract with the weapons lab at Los Alamos. STRETCH was the brainchild of an engineer named Stephen Dunwell and was going to be IBM’s masterpiece: it was a daring design, with all sorts of exotic innovations. We promised customers that it would perform a hundred times as fast as our biggest commercial processor. Looking back, I think STRETCH was so ambitious that it is a wonder we got it built at all. But when it came out in 1961, behind schedule and only 60 percent as powerful as planned, I was disappointed and irate. I thought our engineers needed to be taught a lesson about not letting customers down. So, in the course of a press conference at an industry convention, I announced that the new computer had failed to meet specifications, and that I was hereby cutting the price on the machine from $13.5 million to $8 million to reflect its performance. We couldn’t make any money at that price, and before long the project got shelved.
Making an example out of STRETCH shook up the engineers all right, but it turned out to be a grievous mistake. The engineers understood me to be saying, “No more of those big machines around here,” and it was true that I would have snapped at anybody who brought up the subject. So for two years IBM did almost nothing in supercomputing, leaving the field wide open for Norris and his men.
In August 1963 came their bombshell: a machine called the 6600 that everyone recognized as a triumph of engineering. For seven million dollars, it delivered three times the power of STRETCH. Word of the new machine made me furious, because I thought the distinction of building the world’s fastest computer should belong to IBM. At that point the System/360 was the most advanced set of designs we had, and nothing in that whole product plan was even remotely comparable to the 6600. On August 28, 1963 I sent my top men a memo:
Last week Control Data had a press conference during which they officially announced their 6600 system. I understand that in the laboratory developing this system there are only 34 people, including the janitor. Of these, 14 are engineers and 4 are programmers, and only one person has a Ph.D., a relatively junior programmer.
Contrasting this modest effort with our own vast development activities, I fail to understand why we have lost our industry leadership position by letting someone else offer the world’s most powerful computer.
This note, eventually introduced as evidence in the antitrust suit, became famous as the “janitor memorandum.” It wasn’t illegal but it was where our troubles began, because the moves IBM made in response to my anger were too close to the limits of the law. Even though our engineering staff was overloaded, we tried to catch up with Control Data, and at the System/360 unveiling the following April, we said we would bring out a supercomputer at the top of the line that would leapfrog Control Data’s machine. Control Data hadn’t yet delivered its first 6600, and the effect of this announcement was to put a chill on its market: suddenly it became hard for their salesmen to close deals. Even though our supercomputer didn’t exist yet, a lot of customers decided to hold off until they could get a look. In those days virtually every computer company used this tactic of announcing “paper machines” to keep a competitor from getting too far ahead, but it was exactly the kind of thing I’d told our salesmen to be careful about because of IBM’s great size. The impact of our announcement on Control Data was so great that Norris cut his prices in a panic, and the company plunged into the red.
As it turned out, IBM was never able to beat the Cray design. Within two years Control Data was back on its feet; scores of 6600s had been installed and its salesmen were dropping hints about even faster machines being built in Seymour Cray’s lab; meanwhile we had embarrassed ourselves by announcing four different versions of our supercomputer and still not delivering a one. Finally my business sense overcame my pride and I belatedly figured out that we couldn’t compete with Control Data for the same reason that General Motors can’t compete with Ferrari in building two-hundred-mile-an-hour sports cars. Supercomputers had become so highly specialized that even if we came up with a design equal to theirs, it would never fit in with the rest of our product line, our style of selling, our volume and profit targets, and so on. I was afraid to twist IBM inside out, trying to capture what was actually a small segment of the market. Control Data outsold us more than fifteen to one, and in the end we canceled our supercomputer program after delivering only a limited number of machines.
Because my temper was partly to blame for IBM’s erratic behavior in the supercomputer market, I felt personally responsible for the Control Data lawsuit. Norris wanted compensation for the ordeal we had put his company through. His salesmen had kept detailed records of IBM’s tactics as we fought for supercomputer accounts, and the suit listed no fewer than thirty-seven different ways in which we had supposedly abused our market power. The suit was so detailed and specific that industry wags nicknamed it the “IBM Sales Manual.” The most famous accusation, which came directly from the supercomputer race, was that IBM had marketed “paper machines and phantom computers” in order to block Control Data from winning orders. I never met Norris, but he was a formidable adversary because of his understanding of the industry—and the depth of his antagonism for IBM. He shared everything he knew with federal investigators, and his case and the government’s became intimately entwined.
The wheels of justice slowly began to turn. First came the discovery process, in which each side requisitions documents from the other and takes depositions from witnesses. Control Data demanded the records of sixty IBM departments and the memos and files of over one hundred executives. We had an entire warehouse near Nyack, New York, where hundreds of clerks and paralegals did nothing but process paper for the case. The first batch of material they put together consisted of seventeen million documents—enough to fill a file drawer two miles long. Control Data’s lawyers reviewed forty million IBM memos and documents in all, decided that a million of them were relevant to the case, and had them microfilmed. Then they winnowed this collection down to eighty thousand key documents and used a computer to build a sophisticated electronic index. It was set up to support the allegations that IBM was a monopolist. You could push a button and show, for example, a whole pattern of instances in which we had allegedly offered preferential discounts to keep customers loyal. Control Data’s index marked the first time computer power had ever been applied to a lawsuit on a massive scale—and Norris made the system available to both the Justice Department and the other companies that were taking us to court.
One of the journalists covering these developments pointed out that without Control Data’s help the government wouldn’t have “discovered” very much about IBM or the computer industry at all. The Justice Department could only afford to assign twenty-five people to its case, so it relied on Norris and his staff to do most of the work. The government’s suit went along this way for three years or so, with papers going back and forth among the various parties, but little visible progress toward either a settlement or a trial. Finally in 1972 David Edelstein, the judge who had signed our consent decree in 1956 and who was now the chief judge of the Southern District of New York, assigned himself to the case. He wanted to push rapidly toward a trial, and he also seemed determined to make U.S. v. I.B.M. into a landmark of jurisprudence. Talking to reporters about the case, he said “its universality, its complexity, and its sheer volume of documentation beggar the imagination.… It’s not like ‘A’ suing ‘B.’ This case involves the world and the public.”
Our annual legal bill went up into the tens of millions of dollars. We had the best lawyers—within IBM we had both Marshall and Nick Katzenbach, who arrived just after the suit was filed; and for our outside counsel we used the firm of Cravath Swaine & Moore, led by Bruce Bromley, the seventy-nine-year-old retired judge who had guided us to the 1956 settlement. Perhaps I should have been more optimistic, but it depressed me to see IBM back in the lawyers’ hands. The antitr
ust case began to color everything we did. For years every executive decision, even ones that were fairly routine, had to be made with one eye on how it might affect the lawsuit. To keep damning evidence to a minimum, the lawyers even dictated what we could and couldn’t say at meetings. There were all sorts of code words and strange uses of language: for example, the executives in our computer divisions were told to avoid military metaphors when talking about beating the competition, and if IBM had more than fifty percent of a given market, they were supposed to use the phrase “market leadership” instead of “market share.” This sort of mealymouthing went against my instincts. I wanted IBM to be the best in everything and recognized as such, which meant making no apologies and capturing more market share than anybody else. Instead we were slowly tying ourselves in knots. In 1969 and 1970, because of the double drag of the lawsuits and a recession, IBM’s annual growth slumped to less than five percent—down from the nearly thirty percent we’d achieved in each of the two previous years.
All my life my answer to complicated circumstances has been to make a dramatic and decisive move, but this time there were no such moves to make. I remember agonizing over how we could settle the Justice Department suit and avoid the risk of going into court. The 1956 consent decree that ended our previous case had been reasonable, and my inclination was to make another deal. But all the Justice Department was offering us this time was the death penalty. They wanted to dismantle our company, so that instead of one $7 billion-a-year company, IBM would become seven $1 billion companies. That was something I couldn’t accept. At the beginning I was willing to split IBM in two, with one half devoted to making large computers and the other to small; we could have operated that way even though the split would have been traumatic. But the Justice Department wasn’t interested, and over the course of the suit, increasing competition from Japan weighed against any divestiture at all because we needed all of our mass and might to fight competitors such as Fujitsu and NEC. Our only choice was to let the case drag on.
Four years later I finally had an opportunity to move decisively. In late 1972, the pretrial maneuvering in the Control Data suit was finally coming to a close, and the lawyers at Cravath recommended that we’d be wise to settle out of court. Vin Learson negotiated the deal, which cost IBM a small fortune. It involved selling Norris our Service Bureau Corporation subsidiary for a fraction of its real worth. This was a $63 million-a-year business that did data processing jobs for customers whose own computers were overloaded or who didn’t have computers of their own. Control Data operated a pretty large service bureau of its own, and the addition of ours made them the world’s largest supplier of computer services. We also gave Norris a package of cash and contracts worth $101 million, including $15 million to cover his legal fees.
In spite of the expense, this settlement was a brilliant tactical stroke on the part of Cravath. Because of all the work Control Data had done in analyzing and indexing our documents, their suit was the master link in all the other antitrust suits against IBM. That computerized index now became the property of IBM, because it is customary in out-of-court settlements for the two sides to exchange their lawyers’ paperwork as a way of burying the hatchet. We got the so-called work product from their lawyers and gave them ours, and that night Bruce Bromley came to my office.
“You’ve got to destroy that index right away,” he said.
“My God, isn’t that illegal?” I said.
“It’s perfectly legal. We’ve spent millions of dollars to get them to drop the suit, and we get the work product as part of the settlement. It belongs to us now.”
“I can understand that we own the work product. But it seems to me there is something wrong about destroying evidence. Aren’t there laws against that?”
“Yes, but this isn’t evidence and it never was. Technically speaking, it’s nothing more than a set of files that used to belong to their lawyers and now belongs to you. You can do whatever you want with it.”
We both knew full well that the index could eventually be used to marshal evidence in the other lawsuits that were still pending. Business sense dictated my answer. If I didn’t take the lawyers’ advice, I would be placing IBM under a terrific handicap. I had no choice. “Burn it,” I said.
They destroyed the index that very night. A few days later there was a flurry about this in the newspapers, because the Justice Department and a lot of people in the industry felt that destroying the index was improper. But there was nothing they could do, and their investigations were hobbled after that. I never was totally comfortable about that decision. Even though I had a fiduciary responsibility to protect the investment of IBM shareholders, and even though the best lawyers money could buy assured me I wasn’t breaking any law, it made me feel uneasy.
Without Control Data around to help, the Justice Department’s case deteriorated into a courthouse mess. Judge Edelstein let things get completely out of hand. He refused to set limits—either on the issues the Justice Department was allowed to add as the case progressed, or the amount of evidence or witnesses we could use to respond. He left crucial motions undecided for months and made arbitrary rulings that shocked lawyers on both sides—such as insisting that depositions be taken in court rather than at the office or home of the witness as is customary. The lawyers had to spend months reading depositions into the court record with Judge Edelstein mostly absent from the bench. I sometimes thought he was overwhelmed and afraid to come to grips with the case. As time went by he became so hostile toward our lawyers and witnesses that we tried to have him removed.
The case dragged on for twelve years, until the Reagan administration finally dropped it in 1981. Looking back, I see a lot of sad irony in the whole affair. I think a lot of people would agree that at the outset the Justice Department’s complaint had merit. IBM was clearly in a commanding position in the market, and some of our tactics had been harsh. We eliminated many of these practices ourselves, and our overall record during the case was pretty clean. But I’ve always thought that if Judge Edelstein had carried the suit along rapidly, we’d likely have ended up with a consent decree in which we might have formally agreed to hold back from announcing our machines until we were a little further along with their development, to loosen our grip on the educational market, and so on. Instead, the case stretched on unresolved for so long that before it was over history showed my argument to Ramsey Clark to have been right. IBM kept growing, but the computer industry grew even more, and the natural forces of technological change etched away whatever monopoly we may have had.
Enough was going wrong by 1970 that I started to daydream about a very different kind of life. In the top drawer of my desk, mixed in with memos about key business issues and old letters from Dad, I had a secret list that I’d take out and look at when nobody was around. On it were adventures I wanted to have: climbing the Matterhorn was first, then learning to fly a helicopter, going on safari, sailing to the Arctic and around Cape Horn, and making a singlehanded voyage—to anywhere. I also wanted time to enjoy myself with my wife and my children. My zest for business was evaporating fast. We’d built IBM into a seven-billion-dollar-a-year giant, and in my heart of hearts I felt that I’d taken it as far as I wanted to go. I was fifty-six years old, and my life was a constant punch, punch, punch of making decisions and pushing IBM ahead—running from crisis to crisis, going to company dinners, visiting plants. Each year was filled with hundreds of meetings and speeches and public appearances, so that I was doing something practically every night and spending as much time on the road as at home. I’d lived this way for more than fifteen years, and Dad had kept it up all his life, but I was bone-tired and fooling myself that the fast pace didn’t bother me. During one hectic week I flew to Chicago to give a speech, and in the half hour I had to myself beforehand the thought suddenly came to me: “I can’t keep this up. Something is going to break.” But I pushed the idea out of my mind.
The Nixon recession, which started in mid-1969, was longer
and deeper than anybody had predicted, and IBM hit the first serious downturn of my career. Not only did our sales stay flat, but profits actually began to sink for the first time since the war, giving Wall Street the idea that IBM’s glory days might finally be over. In the first eight months of 1970 our stock fell and fell, losing almost half its value. Try as I might to get the company revved up, things looked worse and worse as the year progressed.
At the same time as the company struggled, my sense of isolation increased. Dick resigned from IBM in March to become U.S. ambassador to France. His good feeling about the company had never been restored after the events surrounding System/360. I knew the ambassadorship was something Dad would have been proud of—it took a weight off my heart to see my brother rise, despite my mismanagement, to the highest stature anyone in our family had attained. Still, at the same time, it was painful for me to think that when I stepped down, there would be no Watson to pass the business to. Meanwhile Jane, our sister, was slowly dying of cancer. It metastasized and in the spring she needed surgery to remove a tumor from her brain. From then on she was in a decline, in and out of the hospital every couple of months.
Father, Son & Co. Page 44