Only the Paranoid Survive Read online

  “New rules prevailed

  now—and they were

  powerful enough to

  cost us nearly half a

  billion dollars.”

  I teach a class strategic management at Stanford University’s business school as a part-time departure from my job as president and CEO of Intel Corporation. The way my co-teacher, Professor Robert Burgelman, and I normally grade the students is to go through the class roster right after each session and assess each student’s class performance while it is still fresh in our memories.

  The process was taking a little longer than usual on the morning of November 22, 1994, the Tuesday before Thanksgiving, and I was about to excuse myself to call my office when the phone rang. It was my office calling me. Our head of communications wanted to talk to me—urgently. She wanted to let me know that a CNN crew was coming to Intel. They had heard of the floating point flaw in the Pentium processor and the story was about to blow up.

  I have to backtrack here. First, a word about Intel. Intel in 1994 was a $10 billion-plus producer of computer chips, the largest in the world. We were twenty-six years old and in that period of time we had pioneered two of the most important building blocks of modern technology, memory chips and microprocessors. In 1994, most of our business revolved around microprocessors and it revolved very well indeed. We were very profitable, growing at around 30 percent per year.

  Nineteen ninety-four was a very special year for us in another way. It was the year in which we were ramping our latest-generation microprocessor, the Pentium processor, into full-scale production. This was a very major undertaking involving hundreds of our direct customers, i.e., computer manufacturers, some of whom enthusiastically endorsed the new technology and some of whom didn’t. We were fully committed to it, so we were heavily advertising the product to get the attention of computer buyers. Internally, we geared up manufacturing plants at four different sites around the world. This project was called “Job 1” so all our employees knew where our priorities lay.

  In the context of all this, a troubling event happened. Several weeks earlier, some of our employees had found a string of comments on the Internet forum where people interested in Intel products congregate. The comments were under headings like, “Bug in the Pentium FPU.” (FPU stands for floating point unit, the part of the chip that does the heavy-duty math.) They were triggered by the observation of a math professor that something wasn’t quite right with the mathematical capabilities of the Pentium chip. This professor reported that he had encountered a division error while studying some complex math problems.

  We were already familiar with this problem, having encountered it several months earlier. It was due to a minor design error on the chip, which caused a rounding error in division once every nine billion times. At first, we were very concerned about this, so we mounted a major study to try to understand what once every nine billion divisions would mean. We found the results reassuring. For instance, they meant that an average spreadsheet user would run into the problem only once every 27,000 years of spreadsheet use. This is a long time, much longer than it would take for other types of problems which are always encountered in semiconductors to trip up a chip. So while we created and tested ways to correct the defect, we went about our business.

  Meanwhile, this Internet discussion came to the attention of the trade press and was described thoroughly and accurately in a front-page article in one of the trade weeklies. The next week it was picked up as a smaller item in other trade papers. And that seemed to be it. That is, until that Tuesday morning before Thanksgiving.

  That’s when CNN showed up wanting to talk to us, and they seemed all fired up. The producer had opened his preliminary discussion with our public relations people in an aggressive and accusatory tone. As I listened to our head of communications on the phone, it didn’t sound good. I picked up my papers and headed back to the office. In fact, it wasn’t good. CNN produced a very unpleasant piece, which aired the next day.

  In the days after that, every major newspaper started reporting on the story with headlines ranging from “Flaw Undermines Accuracy of Pentium Chips” to “The Pentium Proposition: To Buy or Not to Buy.” Television reporters camped outside our headquarters. The Internet message traffic skyrocketed. It seemed that everyone in the United States keyed into this, followed shortly by countries around the world.

  Users started to call us asking for replacement chips. Our replacement policy was based on our assessment of the problem. People whose use pattern suggested that they might do a lot of divisions got their chips replaced. Other users we tried to reassure by walking them through our studies and our analyses, offering to send them a white paper that we wrote on this subject. After the first week or so, this dual approach seemed to be working reasonably well. The daily call volumes were decreasing, we were gearing up to refine our replacement procedures and, although the press was still pillorying us, all tangible indicators—from computer sales to replacement requests—showed that we were managing to work our way through this problem.

  Then came Monday, December 12. I walked into my office at eight o’clock that morning and in the little clip where my assistant leaves phone messages there was a folded computer printout. It was a wire service report. And as often happens with breaking news it consisted only of the title. It said something to this effect: IBM stops shipments of all Pentium-based computers.

  All hell broke loose again. IBM’s action was significant because, well, they are IBM. Although in recent years IBM has not been the power they once were in the PC business, they did originate the “IBM PC” and by choosing to base it on Intel’s technology, they made Intel’s microprocessors preeminent. For most of the thirteen years since the PC’s inception, IBM has been the most important player in the industry. So their action got a lot of attention.

  The phones started ringing furiously from all quarters. The call volume to our hotline skyrocketed. Our other customers wanted to know what was going on. And their tone, which had been quite constructive the week before, became confused and anxious. We were back on the defensive again in a major way.

  A lot of the people involved in handling this stuff had only joined Intel in the last ten years or so, during which time our business had grown steadily. Their experience had been that working hard, putting one foot in front of the other, was what it took to get a good outcome. Now, all of a sudden, instead of predictable success, nothing was predictable. Our people, while they were busting their butts, were also perturbed and even scared.

  And there was another dimension to this problem. It didn’t stop at the doors of Intel. When they went home, our employees had to face their friends and their families, who gave them strange looks, sort of accusing, sort of wondering, sort of like, “What are you all doing? I saw such and such on TV and they said your company is greedy and arrogant.” Our employees were used to hearing nothing but positive remarks when they said that they worked at Intel. Now they were hearing deprecating jokes like, “What do you get when you cross a mathematician with a Pentium? A mad scientist.” And you couldn’t get away from it. At every family dinner, at every holiday party, this was the subject of discussion. This change was hard on them, and it scarcely helped their spirits when they had to go back the next morning to answer telephone hotlines, turn production lines on their heads and the like.

  I wasn’t having a wonderful time either. I’ve been around this industry for thirty years and at Intel since its inception, and I have survived some very difficult business situations, but this was different. It was much harsher than the others. In fact, it was unlike any of the others at every step. It was unfamiliar and rough territory. I worked hard during the day but when I headed home I got instantly depressed. I felt we were under siege—under unrelenting bombardment. Why was this happening?!

  Conference room 528, which is located twenty feet from my office, became the Intel war room. The oval table there is meant to seat about twelve people, but at several times each d
ay more than thirty people were jammed in the room, sitting on the credenza, standing against the wall, coming and going, bringing missives from the front and leaving to execute agreed-upon courses of action.

  After a number of days of struggling against the tide of public opinion, of dealing with the phone calls and the abusive editorials, it became clear that we had to make a major change.

  The next Monday, December 19, we changed our policy completely. We decided to replace anybody’s part who wanted it replaced, whether they were doing statistical analysis or playing computer games. This was no minor decision. We had shipped millions of these chips by now and none of us could even guess how many of them would come back—maybe just a few, or maybe all of them.

  In a matter of days, we built up a major organization practically from scratch to answer the flood of phone calls. We had not been in the consumer business in any big way before, so dealing with consumer questions was not something we had ever had to do. Now, suddenly, we did from one day to another and on a fairly major scale. Our staffing first came from volunteers, from people who worked in different areas of Intel—designers, marketing people, software engineers. They all dropped what they were doing, sat at makeshift desks, answered phones and took down names and addresses. We began to systematically oversee the business of replacing people’s chips by the hundreds of thousands. We developed a logistics system to track these hundreds of thousands of chips coming and going. We created a service network to handle the physical replacement for people who didn’t want to do it themselves.

  Back in the summer when we had first found the floating point flaw, we had corrected the chip design, checked it out very thoroughly to make sure the change didn’t produce any new problems, and had already started to phase the corrected version into manufacturing by the time these events took place. We now accelerated this conversion by canceling the usual Christmas shutdown in our factories, and speeded things up even further by pulling the old material off the line and junking it all.

  Ultimately, we took a huge write-off—to the tune of $475 million. The write-off consisted of the estimates of the replacement parts plus the value of the materials we pulled off the line. It was the equivalent of half a year’s R&D budget or five years’ worth of the Pentium processor’s advertising spending.

  And we embarked on a whole new way of doing business.

  What happened here? Something big, something different, something unexpected.

  For twenty-six years, every day that we did business, we decided what was good and what wasn’t when it came to our own product. We set our own quality levels and our own specifications, and shipped when we decided a product met our own criteria. After all, we had designed and conceived these products, and along with the product came our implicit right—and obligation—to decide when the product was good and when it wasn’t. Nobody ever questioned that we had the right to do that and generally we were on target. Over twenty-six years, we pioneered one classic product after another: DRAMs, other types of memory chips, microprocessors, computers on a board. Our products had become the basic building blocks of digital electronics. But now, all of a sudden, we were getting strange looks from everyone that seemed to say, “Where do you get off telling us what’s good for us?”

  Furthermore, since we generally don’t sell microprocessors to computer users but to computer makers, whatever problems we had in the past, we used to handle with the computer manufacturers, engineer to engineer, in conference rooms with blackboards, based on data analyses. But now, all of a sudden, 25,000 computer users were calling us every day, saying, “Give me a new part, period.” We found ourselves dealing with people who bought nothing directly from us yet were very angry with us.

  What was the hardest to take was the outside world’s image of us. I still thought of us as a creative, dynamic start-up that had just grown a bit bigger than the other creative, dynamic start-ups. We could still turn on a dime. Our people still put the interests of the company ahead of their own interests and, when problems arose, employees from all different divisions would still rally around and put in incredible hours without anyone ordering them to do so. Yet now the world seemed to treat us like some typical mammoth corporation. And, in the public view, this corporation was giving people the runaround. That outside image didn’t jibe with my view of us.

  What had happened? And why now? What was different this time? Something was, but in the middle of those events it was hard to tell what.

  What Happened to Us

  A year or so later, as I reflect on it, I see two big long-term forces doing their work on us, creating the conditions in which a tiny flaw in a microprocessor’s floating point unit could mushroom into half a billion dollars’ worth of damage in less than six weeks.

  The first involved our attempting to change how our products were perceived. A few years back, we had introduced a major marketing campaign, the “Intel Inside” program. It was the biggest campaign the industry had ever seen—in fact, it ranks up there with big-time consumer merchandising campaigns. Its aim was to suggest to the computer user that the microprocessor that’s inside his or her computer is the computer.

  Like all good merchandising campaigns, this had the advantage of reinforcing the truth. Even before the campaign, when you asked someone what kind of a computer he had, the first thing he tended to say was, “I have a 386”—which was the microprocessor chip inside the computer—and then he would go on to identify the computer manufacturer, what kind of software it had and so on. Computer users knew instinctively that the identity and class of the computer were determined more than anything else by the microprocessor within. This was obviously very good for us. It gave us distinction, an identity, and helped build computer-user communities’ awareness of us and our products.

  We aimed our campaign at driving this point home to a wider consumer base and to future computer buyers. We created a distinctive logo and we worked with manufacturers who used our microprocessor in their product to display this logo in their advertising, often with a sticker on the actual computer. Hundreds of manufacturers, domestic and international, participated in this campaign.

  We spent a lot of money ourselves promoting this brand. We had billboards displaying the “Intel Inside” logo all over the world and we ran television commercials in many languages. We had even distributed many thousands of “Intel Inside” bicycle reflectors in China. By 1994 our research showed that our logo had become one of the most recognized logos in consumer merchandising, up there with names like Coca-Cola or Nike. So when problems developed with our flagship Pentium chip, our merchandising pointed the users directly back to us.

  The second fundamental factor in creating the conditions for the maelstrom was our sheer size. Over the years, we had become the world’s largest semiconductor manufacturer. We surpassed the large United States producers, companies we used to consider mammoth and gigantic compared to us just a few years ago, and we surpassed the major Japanese producers who just ten years ago had threatened to put us out of business (more about that in Chapter 5). And we were still growing fast—faster than most large companies. We had also become bigger than most of our customers, companies that I remembered from our earlier years at Intel to be monumentally large corporations. At some point along the way, like a kid who suddenly looks down at his father, our sizes reversed.

  This all happened relatively rapidly, in the past decade. And although the recognition of our size fleetingly touched us here and there, largely through the respect we got from other companies in our business, it was hardly something any of us dwelled on. It crept up on us; it just happened.

  And now we were dealing with a different and not nearly as pleasant consequence of our huge size and our strong identity. We had become gigantic in the eyes of computer buyers. Unfortunately, it took a blowup to make us realize this.

  Given the gradual nature of these changes, which over time added up to a very large change, the old rules of business no longer worked. New rules
prevailed now—and they were powerful enough to force us into actions that cost us nearly half a billion dollars.

  The trouble was, not only didn’t we realize that the rules had changed—what was worse, we didn’t know what rules we now had to abide by.

  Before this episode, we supplied our computer manufacturer customers well and we guarded the quality of our products as best we knew how. We marketed our products both to the engineers of these computer companies and to computer users. We were fast and agile, as all good start-ups are, we worked hard and everything had worked well in return. But all of a sudden this no longer seemed to be enough.

  What happened to us in the course of this event is something that happens to many businesses. All businesses operate by some set of unstated rules and sometimes these rules change—often in very significant ways. Yet there is no flashing sign that heralds these rule changes. They creep up on you as they crept up on us, without warning.

  You know only that something has changed, something big, something significant, even if it’s not entirely clear what that something is.

  It’s like sailing a boat when the wind shifts on you but for some reason, maybe because you are down below, you don’t even sense that the wind has changed until the boat suddenly heels over. What worked before doesn’t work anymore; you need to steer the boat in a different direction quickly before you are in trouble, yet you have to get a feel of the new direction and the strength of the wind before you can hope to right the boat and set a new course. And the tough part is that it is exactly at times like this that hard and definitive actions are required.

  Such phenomena are very common. Businesses are about creating change for other businesses. Competition is about creating change; technology is about creating change. The appearance and disappearance of regulations cause further changes. Sometimes these changes affect only a company, other times they affect an entire industry. So the ability to recognize that the winds have shifted and to take appropriate action before you wreck your boat is crucial to the future of an enterprise.