This case tells the story of Intel’s transition from memories to microprocessors — the event that cemented Andy Grove’s reputation as the most respected CEO in Silicon Valley.
Intel was founded in 1968. It was started by two alumni of Fairchild Semiconductor: Gordon Moore (of Moore’s Law fame) and Robert Noyce, with the backing of legendary tech investor Arthur Rock. Andy Grove joined on the day of incorporation as one of Intel’s first employees.
Decades later, Grove writes, in Only the Paranoid Survive:
Every start-up has some kind of a core idea. Ours was simple. Semiconductor technology had grown capable of being able to put an ever larger number of transistors on a single silicon chip. We saw this as having some promising implications. An increasing transistor count readily translates into two enormous benefits for the consumer: lower cost and higher performance. At the risk of some oversimplification, these come about as follows. Roughly speaking, it costs the same to produce a silicon chip with a larger number of transistors as a chip with a smaller number of transistors, so if we put more transistors on a chip, the cost per transistor would be lower. Not only that, smaller transistors are in closer proximity to each other and therefore handle electronic signals faster, which translates into higher performance in whatever finished gadget — calculator, VCR, or computer — our chip should be placed into.
When we pondered the question of what we could do with this growing number of transistors, the answer seemed obvious: build chips that would perform the function of memory in computers. In other words, put more and more transistors on a chip and use them to increase the capacity of the computer’s memory. This approach would inevitably be more cost-effective than any other method, we reasoned, and the world would be ours.
At the time, the most common memory in use was magnetic core memory. Intel was proposing that computer makers shift from magnetic core to semiconductors. This was not implausible, but it was untested. Semiconductor-based memories did not then exist as a market. In effect, Intel was attempting to create it.
Their first product was a 64-bit memory. Literally: it could store exactly 64 digits. How that first product came to be was as follows: at around the time that Intel was formed, Honeywell, then a major computer maker, called for proposals to build exactly such a device. Six companies had already bid for the job; Intel was late as the seventh bidder. The company’s engineers worked day and night to design the chip, and in parallel develop the manufacturing process for it. Decades later, Grove writes: “we worked as if our life depended on it, as in a way it did.”
Intel was the first to produce the chip, and therefore won that first proposal. They were off to the races.
The shape of the game for semiconductor manufacturing was to put out successively advanced products. Not too long after that first chip, Intel developed a 256-bit memory. This was a marvel of 1969 technology, and it seemed like “every engineer in every organisation of every computer maker — as well as every semiconductor maker bought one of each to marvel over.” But they didn’t buy production volumes. Intel continued to burn cash; in 1969, semiconductor memories were still a curiosity — and not used seriously by any of the major computer manufacturers.
And so two products and one year in, still in the red, Intel went on to develop their next chip. What else could they do? This time they aimed for a 1024 bit memory — four times as complex as their last one. Honeywell was the company that proposed the memory, and it had come up with a technological approach to get there (this was, in truth, a gamble — nobody knew if it was possible). Intel accepted the project. Grove writes:
[Making the 1024 bit chip] required taking some big technological gambles. But we plugged on, memory engineers, technologists, test engineers and others, a hard-working, if not always harmonious, team. As a result of the pressure we felt, we often spent as much time bickering with one another as working on the problems. But work went on. And this time we hit the jackpot.
The device became a big hit. Our new challenge became how to satisfy demand for it. To put this in perspective, we were a company composed of a handful of people with a new type of design and a fragile technology, housed in a little rented building, and we were trying to supply the seemingly insatiable appetite of large computer companies for memory chips. The incredibly tough task of development segued into the nightmare of producing a device that was held together with the silicon equivalent of bailing wire and chewing gum. The name of this device was the 1103. To this day, when a digital watch shows that number, I and other survivors of that era still take note.
Through our struggle with the first two technological curiosities that didn’t sell and with the third one that sold but that we had such a hard time producing, Intel became a business and memory chips became an industry. As I think back, it’s clear to me that struggling with this tough technology and the accompanying manufacturing problems left an indelible imprint on Intel’s psyche. We became good at solving problems. We became highly focused on tangible results (our word for it is “output”). And from all the early bickering, we developed a style of ferociously arguing with one another while remaining friends (we call this “constructive confrontation”).
This was the emergence of Intel’s self-identity, and the period that created its culture.
Robert Noyce on the left, Gordon Moore on the right in front of the Intel SC1 building in Santa Clara in 1970. (Source)
In the late 1960s through to the early 1970s, Intel had practically 100% share of the memory chip market. It went public on October 19, 1971. It was the first mover, after all, and had proven out the technology. But other competitors soon emerged following the popularity of the 1103 chip. These small companies had names like Unisem, Advanced Memory Systems, and Mostek.
By the end of the 70s, there were about a dozen competitors in the business, and they had caught up. Grove writes (bold emphasis added):
With each succeeding generation of memory chips, somebody, not necessarily the same company, got it right and it certainly wasn’t always us. A prominent financial analyst of the day used to report his observations of the memory business using boxing-match analogies: “Round Two goes to Intel, Round Three goes to Mostek, Round Four to Texas Instruments, we’re gearing up to fight Round Five” … and so on. We won our share. Even ten years into this business, we were one of the key participants. Intel still stood for memories; conversely, memories meant (usually) Intel.
And why wouldn’t it? Intel had created the market segment, after all.
The Strategic Inflection Point
By the late 70s, Intel was knee deep in the competitive fight for memories. Japanese memory producers were, at the time, helpful: Intel had outsourced production to them during a recession when the American company pulled back on production capacity. During that period, the Japanese manufacturers were considered a godsend, and not much of a competitive threat.
But by the early 80s, they started competing with the American memory producers, and they started winning. This didn’t happen immediately, of course. It became clear over the course of 1978 to 1984. Grove writes:
Things started to feel different now. People who came back from visits to Japan told scary stories. At one big Japanese company, for instance, it was said that memory development activities occupied a whole huge building. Each floor housed designers working on a different memory generation, all at the same time: On one floor were the 16K people (where “K” stands for 1,024 bits), on the floor above were the 64K people and on the floor above that people were working on 256K-bit memories. There were even rumors that in secret projects people were working on a million-bit memory. All this was very scary from the point of view of what we still thought of as a little company in Santa Clara, California.
Then we were hit by the quality issue. Managers from Hewlett-Packard reported that quality levels of Japanese memories were consistently and substantially better than those produced by the American companies. In fact, the quality levels attributed to Japanese memories were beyond what we thought were possible. Our first reaction was denial. As people often do in this kind of situation, we vigorously attacked the ominous data. Only when we confirmed for ourselves that the claims were roughly right did we start to go to work on the quality of our product. But we were clearly behind.
As if this wasn’t enough, the Japanese companies had capital advantages. They had (or were said to have) limitless access to funds — from the government? from the parent company through cross subsidies? or through the mysterious workings of the Japanese capital markets that provided nearly infinite low-cost capital to export-oriented producers? We didn’t know exactly how it all worked but the facts were incontrovertible: as the eighties went on, the Japanese producers were building large and modern factories, amassing a capacity base that was awesome from our perspective.
This dominance — so shocking in the early 80s — progressed over the next decade. It felt like an inexorable tide.
Intel responded by improving quality and bringing costs down. But they had a hard time with the Japanese producers, which seemed capable of producing ever higher quality product at increasingly lower prices. Grove mentions in Only The Paranoid Survive that at one point they got their hands on an internal memo sent out to the entire sales force of a large Japanese competitor. The key thing that stuck with him was the part of the memo that said “Win with the 10% rule … Find AMD and Intel sockets … Quote 10% below their price … If they requote, go 10% AGAIN … Don’t quit until you WIN!”
The competition was very discouraging. Intel tried all sorts of things during this period: they tried to retreat into niches of the memory market, they tried to invent special-purpose memories called value-added designs, they introduced more advanced technologies and built memories with them. At each step of the way, they were trying to earn a premium for their memories in a market where prices were being aggressively competed downwards. Grove says that there was a saying at Intel at the time: “If we do well we may get ‘2X’ (twice) the price of Japanese memories, but what good does it do if ‘X’ gets smaller and smaller?”
In truth, Grove says that many of these threats were clear in the early 80s. It was so painful for him to face that Grove distracted himself: he wrote and published his now-legendary book High Output Management in 1983. In Only The Paranoid Survive, written years later, Grove reflects: “Frankly, as I look back, I have to wonder if it was an accident that I devoted a significant amount of my time in the years preceding our memory episode, years during which the storm clouds were already very evident, to writing a book. And as I write this, I wonder what storm clouds I might be ducking now. I’ll probably know in a few years.”
There is another aspect to the story that matters: Intel continued to spend heavily on R&D throughout the period. Part of it was their history, and the way they won earlier in their life — they were a technology company, after all. They believed that every problem should have a technological solution.
During the 70s, Intel spent most of their R&D budget on memories, but a smaller team had been working on microprocessors since 1970 — a new technology they had invented (independently, it turned out, and at the same time as Texas Instruments). Intel’s first microprocessor was built to replace a number of ASICs in a Busicom calculator. The division had chugged along since, but was considered a “slower-growing, smaller-volume market than memory chips.” As such, the entire product line was mostly seen as a side show.
This was reflected in the facilities dedicated to the two divisions. The bulk of the memory chip development was done in a “spanking new facility in Oregon”. The microprocessor designers, on the other hand, were put in a corner of a production facility — and not even a new one at that — shared with manufacturing folks at a remote site.
In 1978, Intel released the 8806 processor, which was snapped up by a variety of computer makers. One of these computer manufacturers was the newly created IBM PC division. The IBM PC was a surprise hit, and demand quickly exploded far ahead of IBM’s expectation. Grove writes that these unexpected wins helped cover some of the losses in the memory division:
IBM (...) looked to us to help them ramp up production of the IBM PC. So did all of IBM’s PC competitors. In 1983 and the early part of 1984 we had a heated-up market. Everything we made was in short-supply. People were pleading with us for more parts and we were booking orders further and further out in time to guarantee a supply. We were scrambling to build more capacity, starting factory construction at different locations and hiring people to ramp up our production volumes.
Unfortunately, this turned in mid 1984. Business dried up. Intel’s order backlog “evaporated like spring snow” and the company was caught off-guard. They had been building production capacity into such strong growth for so long that they couldn’t wind down microprocessor manufacturing quickly enough; the company was still building inventory even as demand vanished.
And here their backs were finally up against the wall. During the period of growth Intel could afford to ignore its problems on the memory side of the house. It was losing money there, but that was ok for as long as microprocessors were selling. Now it was not ok. Grove writes: “We just kept at it, looking for the magical answer that would give us a premium price. We persevered because we could afford to. However, once business slowed down across the board and our other products couldn’t take up the slack, the losses really started to hurt.”
Intel’s leaders finally felt real urgency.
The debates in the company raged on for a year. Some argued that they should “go for it” — build a gigantic factory dedicated to memories and take the fight to the Japanese. Others argued that they should innovate their way out of their problem: develop some advanced technology that the Japanese couldn’t build. Others were still clinging onto the idea that Intel could develop some special-purpose memories, operating in a profitable niche that could be defended against copycats.
The problem was that the company was losing money every day that passed. Grove wrote: “During that time we worked hard without a clear notion of how things were ever going to get better. We had lost our bearings. We were wandering in the valley of death.”
And then, things came to a head. Grove recalls:
I remember a time in the middle of 1985, after this aimless wandering had been going on for almost a year. I was in my office with Intel’s chairman and CEO, Gordon Moore, and we were discussing our quandary. Our mood was downbeat. I looked out the window at the Ferris wheel of the Great America amusement park revolving in the distance, then I turned back to Gordon and I asked, “If we got kicked out and the board brought in a new CEO, what do you think he would do?” Gordon answered without hesitation, “He would get us out of memories.” I stared at him, numb, then said, “Why shouldn’t you and I walk out the door, come back and do it ourselves?”
And then they did. They figuratively walked out the office, stomped out their cigarettes, walked back in, and committed to getting themselves out of memories.
The Culture Revolts
Identity is a funny thing. It is easy to say “oh, let’s get out of memories, the very thing we started the company around” but it is much, much harder to do.
Grove experienced this immediately. He started out tentatively — too tentatively — when talking to people at Intel. Reflecting on this, decades later, he writes (all emphasis added):
As I got more and more frustrated that people didn’t want to hear what I couldn’t get myself to say, I grew more blunt and more specific in my language. The more blunt and specific I got, the more resistance, both overt and covert, I ran into.
So we debated endlessly. I remember at the end of a discussion asking one of our senior managers to write down what he understood our position to be on the subject; he was waffling as he struggled with the decision and I figured I could trap him with his own written memo. I failed. Months went by as we played these weird games.
In the course of one of my visits to a remote Intel location, I had dinner with the local senior managers, as I usually do. What they wanted to talk about was my attitude toward memories. I wasn’t ready to announce that we were getting out of the business yet because we were still in the early stages of wrestling with the implications of what getting out would mean—among them, what work would we have for this very group of people afterward? Yet I couldn’t get myself to pretend that nothing like that could ever happen. So I gave an ambivalent-to-negative answer, which this group immediately picked up on. One of them attacked me aggressively, asking “Does it mean that you can conceive of Intel without being in the memory business?” I swallowed hard and said “Yes, I guess I can.” All hell broke loose.
One major problem was that Intel’s employee culture held two beliefs very strongly — “as strong as religious dogmas” — and both of these had to do with the role of memories in the company. As a result, Grove reflected that an open-minded, rational discussion about getting out of memories was “simply impossible”.
These beliefs were:
“Memories are our technology drivers.” What this meant was that Intel always developed and refined their technologies on memory products first because they were easier to test. Once the technology had been debugged in memories, only then would they be applied to microprocessors and other products. “How can we lose our technology drivers?”, these managers would ask. “What would we use as our technology drivers?” they would plead, and of course Grove had no answer; he was as unsure of the journey ahead as they were.
The second belief was “We need to sell a full product line in order to do well with our customers.” If Intel got out of memories, it would no longer have a full product line. The belief — held collectively by the hundreds of folks who made up Intel’s sales division — was that if they didn’t have a full product lineup, their customers would turn to competitors who did.
The discussion with these senior managers went nowhere because they kept circling back to these two issues. As the night wore on, both Grove and his managers got more and more frustrated with each other.
But it eventually turned out that Grove himself was lured by his emotions. He writes:
The senior manager in charge of our memory business couldn’t get with the program even after months of discussions. Eventually he was offered a different job and accepted it, and I started with his replacement by spelling out exactly what I wanted him to do: Get us out of memories! By this time, having had months of frustrating discussions under my belt, I no longer had any difficulty making myself clear. Still, after the new person got acquainted with the situation, he took only half a step. He announced that we would do no further R&D on new products. However, he convinced me to finish what his group had in the works. In other words, he convinced me to continue to do R&D for a product that he and I both knew we had no plans to sell. I suppose that even though our minds were made up about where we were going our emotions were still holding both of us back from full commitment to the new direction.
Initially, Grove reasoned that the move out of memories had to be done one step at a time. But after a few more months of this he gave up. Intel had to rip the bandaid: it had to get out of memories, once and for all.
Intel’s leadership told the sales force to notify all of their memory customers. This was a scary moment. How would the customers react? Would they stop doing business with Intel now that Intel was no longer ‘full product line’? Surprisingly, the reaction was ‘a big yawn’: most of Intel’s customers had already figured out that they were not a big player in the memory market, and had made alternative arrangements with other suppliers. In fact, some of them responded to the announcement: “It sure took you a long time.” Grove observed later that “people who have no emotional stake in a decision can see what needs to be done sooner.”
So they were finally out of memories. But what to do next? The answer here was relatively clear: by this point in Intel’s life, they had been supplying the key microprocessors for IBM-compatible PCs for nearly five years; they were the largest player in the PC market. And Intel’s next mainline microprocessor, the 386, was ready to go into production. There was one, final step — getting Intel’s best developers in that ‘spanking new facility in Oregon’ on board with 386 production.
So I went up to Oregon. On the one hand, these developers were worried about their future. On the other hand, they were memory developers whose interest in and attachment to microprocessors was not very strong. I gathered them all in an auditorium and made a speech. The theme of the speech was, “Welcome to the mainstream.” I said that Intel’s mainstream was going to be microprocessors. By signing up to do microprocessor development, they would be bearing the flag for Intel’s mainline business.
It actually went a lot better than I had expected. These people, like our customers, had known what was inevitable before we in senior management faced up to it. There was a measure of relief that they no longer had to work on something that the company wasn’t fully committed to. This group, in fact, threw itself into microprocessor development and they have done a bang-up job ever since.
Intel was a microprocessor company for the rest of Andy Grove’s life.
Their transition went more smoothly than it otherwise might have in part because the 386 was a huge hit — the most successful microprocessor in history up to that point. It took only a year to return to profitability, though it took a few years for the public to cotton on to Intel’s new identity.
That repositioning, post exit, happened deliberately. Grove notes that Intel began calling themselves a “microcomputer company” in public statements, literature, and advertising (‘microcomputer’ was the term used to describe desktop computers of the time, what we now know as Personal Computers, or PCs). As the 386 became a phenomenal success, the identity took hold in the hearts and minds of Intel’s management, and eventually most of its employees. And then the outside world began to see Intel as a processor company — so much so that it is barely recognised for products other than processors today.
How long did the entire transition take? In Only The Paranoid Survive, Grove lays out the key years in this transition as he saw it — something he calls a ‘strategic inflection point’ (bold emphasis added):
I learned that the word “point” in strategic inflection point is something of a misnomer. It’s not a point; it’s a long, torturous struggle.
In this case, the Japanese started beating us in the memory business in the early eighties. Intel’s performance started to slump when the entire industry weakened in mid-1984. The conversation with Gordon Moore that I described occurred in mid-1985. It took until mid-1986 to implement our exit from memories. Then it took another year before we returned to profitability. Going through the whole strategic inflection point took us a total of three years. And while today, ten years later, they now seem compressed to one short and intense period, at the time, those three years were long and arduous—and wasteful. While we were fighting the inevitable, trying out all sorts of clever marketing approaches, looking for a niche that couldn’t possibly exist in a commodity market, we were wasting time, getting deeper into red ink and ultimately forcing ourselves to take harsher actions to right things when we finally got around to taking action at all. While the realization of what we were facing was a flash of insight that took place in a single conversation, the work of implementing the consequences of that conversation went on for years.
But, finally, the most important bit lay at the end of all his reflections. Grove pointed out that with the benefit of hindsight, he now realised that Intel’s employee culture had long been preparing for the strategic shift. This is perhaps the most powerful section of the book (all bold emphasis added):
… while Intel’s business changed and management was looking for clever memory strategies and arguing among themselves, trying to figure out how to fight an unwinnable war, men and women lower in the organization, unbeknownst to us, got us ready to execute the strategic turn that saved our necks and gave us a great future.
Over time, more and more of our production resources were directed to the emerging microprocessor business, not as a result of any specific strategic direction by senior management but as a result of daily decisions by middle managers: the production planners and the finance people who sat around the table at endless production allocation meetings. Bit by bit, they allocated more and more of our silicon wafer production capacities to those lines which were more profitable, like microprocessors, by taking production capacity away from the money-losing memory business. Simply by doing their daily work, these middle managers were adjusting Intel’s strategic posture. By the time we made the decision to exit the memory business, only one out of eight silicon fabrication plants was producing memories. The exit decision had less drastic consequences as a result of the actions of our middle managers.
This is not unusual. People in the trenches are usually in touch with impending changes early. Salespeople understand shifting customer demands before management does; financial analysts are the earliest to know when the fundamentals of a business change.
While management was kept from responding by beliefs that were shaped by our earlier successes, our production planners and financial analysts dealt with allocations and numbers in an objective world. For us senior managers, it took the crisis of an economic cycle and the sight of unrelenting red ink before we could summon up the gumption needed to execute a dramatic departure from our past.
Were we unusual? I don’t think so. I think Intel was a well-managed company with a strong corporate culture, outstanding employees and a good track record. After all, we weren’t quite seventeen years old, and in those seventeen years we had created several major business areas. We were good. But when we were caught up in a strategic inflection point, we almost missed it; we nearly became another Unisem, another Mostek, another Advanced Memory Systems.
Sources
Only The Paranoid Survive: How to Exploit the Crisis Points that Challenge Every Company and Career, Andrew S. Grove.
https://archive.computerhistory.org/resources/text/Oral_History/Karp_Joel/102658274.05.01.pdf