Interview with Gordon
March 3, 1995
Los Altos Hills, California
RW: We're here today
in Woodside, California with Gordon and Betty Moore. Gordon is a senior statesman
in semiconductors, having started at Shockley Labs, been a founder of Fairchild,
and then a founder of Intel Corporation. He's presently Chairman of the Board
of Intel Corporation and Betty and he are going to take us on a little tour
[Garden tour segment omitted , approximately seven minutes.]
... Well, let's go do an interview.
[ Start of Interview: 7:46]
GM: OK, fine.
RW: Well, Gordon, tell me about your early days. I guess you were born right here in California?
GM: That's right. I was born actually in San Francisco only because it was the nearest hospital to Pescadero, a little farm town kind of community. Pescadero was really my home town. I stayed... its only distinction is that it's the only town I know of in California that's smaller now than it was sixty years ago. The main street used to go through it and then the moved the highway onto the coast and Pescadero became a bit of a backwater.
I moved to Redwood City when I was ten years old. My father was deputy sheriff and he got a promotion and had to move into the county seat. And I lived in Redwood City, less than five miles from where we are right now, until I went away to school. In fact, my parents lived there until they died.
I went through grammar schools here locally, Sequoia High School, two years at San Jose State, where I met my wife, to, be, and then transferred to University of California at Berkeley for my junior and senior years. I got a bachelor's degree in chemistry in 1950 and went on to Caltech to do graduate work. Got a Ph.D., it was awarded in 1954.
Actually at that time, it was hard to find technical jobs in California. I had my new Ph.D. in chemistry and physics and started looking around and decided I really had to look east in order to find a job. So my first job was at the Applied Physics Laboratory of Johns Hopkins University in Silver Spring, Maryland, just outside of Washington, DC. I went to work there in September of 1953, just after I had finished the requirements for my graduate work, about a year before the degree was actually awarded. There I did basic research on such things as flames, some shapes of spectral lines... what caused them to behave the way they did. And I found myself calculating the cost per word to the taxpayer in the published articles coming out of that work, and I wasn't really sure that society was benefiting sufficiently from what I was doing!
It was time I should get to something a bit
more practical. That was helped by the fact that the group I was working
in was kind of breaking up, it was a little research group within the Applied
Physics Laboratory. So I started looking for another job and put out feelers
in several places. One of them happened to be the Lawrence Livermore Laboratory
interviewed in Livermore and the work they wanted me to do wasn't something I found very attractive. But that actually led into my getting into the semiconductor industry. Because Dr. William Shockley knew the people at Lawrence Livermore and got access to their records, the people to whom they had made offers and had turned them down. And he got my name, thought he needed a chemist in a new operation he was setting up, and gave me a call one evening to see if I would be interested in talking to him about that possibility.
And that's really how I got into semiconductors.
RW: Well, what was Bill Shockley like?
GM: Well, he was an unusual fellow. First of all, he was extremely competitive and controversial. If there were two ways of stating things, one of which was controversial, and one of which was straightforward, he'd pick the controversial one every time. He just thrived on stimulating controversy. He had phenomenal physical intuition. One of my colleagues once said he thought Shockley could see electrons, he had such a good idea of what was going to happen. He was very competent in the solid state physics area, actually had been in other technical areas before that.
But he had some peculiar ideas on how to motivate people. This was the first time he really took on a major management responsibility. At Bell Laboratories, he had run a relatively small research group. But here he was trying to set up a new enterprise and some of his ideas, frankly, didn't work out too well for the success of that enterprise.
RW: Did he do polygraph tests on people?
GM: Well that was one of the things that happened. We had an incident in the laboratory where, actually a little pin point was left in one the doors and a lady cut her hand on it a bit. And Shockley decided that was malicious and started trying to track down who had put this point there in order to hurt this lady. And it got to the point where he was going start going through the whole staff with a polygraph test. He didn't get very far with that one however, we all kind of rebelled and that one died.
RW: I also heard a story that a visiting professor came in and was giving a presentation in the morning at eight o'clock or so and Shockley came in and the technical staff was there, and he said: "How come you guys aren't at work?" Is that one that you recall?
GM: I don't remember that one specifically, but there are a lot of Shockley stories. For example, the people he hired were typically young scientists in those days and he'd ask us we were in our twenties he'd ask the group one day what he could do to make their job more interesting and more rewarding and a couple of them commented, well, gee, they'd like to be able to publish some papers. So he said "OK." That night he went home and he worked out the theory for some effect in semiconductors, I don't exactly remember what, brought it back, handed it to him the next morning and said "Here, flesh this out and publish it."
Typical of the feeling he had for what was really motivating people.
RW: Yeah. So what caused you guys was it eight? eight of you, "the treacherous eight?"
GM: Yeah, well, we've been called a variety of things.
RW: What caused you to leave?
GM: Well, there were several of these incidents
where Shockley's management was really disruptive rather than helpful. For
example, we had a secret project we were a very small group, forty or fifty
people, total, in a single buildingand Shockley came up with a new
idea and all some of us could know was that it was "potentially as
important as the
invention of the transistor." Half the group was working on it and the other half couldn't tell what it was very disruptive to the organization.
Anyhow, Arnold Beckman, the founder of Beckman Instruments, was the financing behind Shockley Semiconductor Laboratory, and Arnold came up to talk to us one time and after he was done Shockley stood up and made a few comments ending with, "If you're not happy with what were doing, I can take this someplace else." Which was completely uncalled for. Anyhow, we used that as a reason to actually call Beckman and say, "Hey, that's not true. If Shockley tried to move here, he'd have to go almost by himself."
And Beckman used that to say, "Well, I gather things aren't going completely right there. What can we do about it?" We agreed to have a dinner meeting with Beckman. In fact, we ended up having several dinner meetings. I would guess four, in retrospect, where Beckman came up from southern California and a group of us met with him to see what steps could be taken to improve the management situation at Shockley. The kind of thing we were looking at was something that put Shockley in a consulting position and brought in somebody who was an experienced general manager. In fact, I remember, Beckman had someone in his organization by the name of Joe Louis who kept getting taller and taller every time he talked about him he sounded like just the kind of person we needed for the operation.
Anyhow, on the last of these dinner meetings with Beckman, you now, we had previously been exploring such things as getting Shockley a professorship at Stanford but keeping him as a consultant. By that time, Shockley had won the Nobel Prize and it wasn't very difficult to find him a professorship someplace.
Anyhow, someone had gotten to Beckman... I have heard a rumor that it was someone from Bell Laboratories... and told Beckman that if he made this move, it would just ruin Shockley's career. So Beckman's attitude toward us changed and he essentially said, "Hey, Shockley's the boss, you guys like it or... whatever." At that time we felt we felt we had burned our bridges so badly that there was no way we could stick around working for Shockley, after having gone around him to try to straighten out this problem.
So, one of the group that had been meeting with Beckman, by the name of Gene Kleiner, had a friend who worked at Hayden Stone, the investment banking firm, in New York at that time. And it said, essentially, "there's a group of us who like working together, we're all going to leave Shockley, do you think there's a company that would like to hire the group?" The response was: "Well, wait a minute," and one of the senior partners, by the name of Bud Coyle, and a young Harvard MBA by the name of Arthur Rock, came to California to meet with the group.
After spending an evening with us talking about the kind of things we wanted to do and the like, they concluded that rather than finding a company that would like to hire us, what we ought to do is set up our own company and sold us on the idea of establishing a new company. Now, we found that fairly easy to accept as an idea because we all lived here, we all owned houses in the area, and it would clearly be a lot less disruptive to our personal lives than any other solution we were likely to come up with.
So, Hayden Stone agreed to take on the job of raising money to start a company and the company we were looking at was to pursue the goal that Shockley had earlier abandoned. He was initially going to make a silicon transistor and then he changed to deciding to make a four, layer diode, a rather specialized and obscure device. Well we still thought the transistor was the way to go and Fairchild, finally, was financed, started to pursue that transistor.
Anyhow, Hayden Stone said they would find financing and that was before
the days of readily available venture capital. So we sat down with a copy
of the Wall Street Journal and literally went through all of the companies
on the New York Stock Exchange trying to identify everyone we could think
of that might like to start a semiconductor operation. We
identified some thirty, five companies and the Hayden Stone team visited every one of them and got turned down by every one of them. None of them even getting to the point where they wanted to talk to the group. The reason, presumably, for several of them, was because they didn't see how they could support a group on the outside while they had a lot of engineers on the inside doing similar things who wouldn't be getting the same kind of a deal.
Anyhow, by accident, the people from Hayden Stone were introduced to Sherman Fairchild. Sherman Fairchild was intrigued by technology and gadgets. He had previously set up Fairchild Camera and Instrument Corporation and Fairchild Aviation when he was pursuing aerial surveying. Sherman Fairchild introduced them John Carter who was the chief executive of Fairchild Camera and Instrument.
John Carter sent his executive vice, president, by the name of Richard
Hodgson, out to meet with us and after spending an evening with us... there
may have been more than the one meeting, I don't really remember that too
well... anyhow, Fairchild agreed to support the group of us, and there were
eight of us in the group at that time, to try to establish a
semiconductor company and that was the beginning of Fairchild Semiconductor Corporation.
RW: What were some of the technical breakthroughs at Fairchild?
GM: Well, Fairchild did a lot of pioneering work. The company was really in the right place at the right time. In the first place, we pursued this idea of a diffused silicon transistor that Shockley had been initially going to do. It was something that had been made in the laboratory at Bell Labs but was not a commercial device at all. We were the first ones to bring to the market what's known as a "mesa" transistor and it was a silicon mesa transistor at that time. It was quite a successful device on the scale that we were working at least, but that was only the first of several devices.
That was the first silicon transistor that was built by the batch process where you made a lot of them on a wafer and then cut them up individually. And it was the first device in manufacturing that used photolithography to produce the structures. So these were fairly important early developments.
RW: Let's break right now.
GM: Speaking of
lithography reminds me that when Fairchild started, we split the major processes
that had to be developed among the participants there and Bob Noyce had
responsibility for setting up the lithography capability. He went to San
Francisco to a large camera store and dug through their supply of 35mm movie
camera lenses excuse me,
16mm movie camera lenses and picked out the three that were best matched in so far as focal length was concerned. And those were the optics in the step, and, repeat cameras we built to make the first transistor structures.
RW: In other words, there was no... you had to build everything in those days... there was no industry that built this...
GM: That's right. We were the first ones getting into that technology. I had responsibility for the fusion furnaces for example. I'd built furnaces previously, so we designed furnaces and actually, one of the very first spin, offs to start a new company out of Fairchild was a result of that. Art Lash was my technician... we had him doing a few things but initially he was working with some of the assembly operations and we had to make little glass capillary tubes... to bond the gold wires onto the transistors. We had a scheme where the gold wire was fed through the capillary and then you heated it with a flame so it made a ball on the end and then you pushed down the capillary and it would stick to metal. Anyhow, those tended to get plugged so we needed a lot of capillaries, so we encouraged Art to go into business nights and evenings making capillaries to sell to Fairchild. And then he was helping me also with the furnaces we were designing and those two products were the basis of the formation of Electroglass Corporation. Art finally made that a full, time job rather than continuing to work at Fairchild. A lot of the infrastructure companies developed more or less like that during those early days.
Anyhow, while we were making these first mesa transistors, completing development of the process and putting them into production, we had a person whose background was as a theoretician, as part of the original group, by the name of John Hoerni. And, particularly when we setting up the initial equipment, John was writing in his notebook and coming up with ideas of things to try, and he came up with a proposal... of instead of making a "mesa," which exposes the sensitive area of the transistor to the outside world, that one should just do more of these diffusions oxide mask diffusions and leave the oxide over the top of the junction, the sensitive part. Well, that was something that previously had been considered a bad idea because Bell Labs' conventional wisdom was that the oxide was dirty and you wanted to get rid of it.
But we couldn't try John's idea right away because it took four index masking operations in order to make the structure he was proposing and Bob Noyce only bought three lenses!
So we couldn't make enough masks in order to make the full structure at first. So, the idea lay dormant for well over year approaching two years before we could get it to the point where we could actually try it. And when we did, it turned out to work beautifully. This protected the transistors in the regions where they were really sensitive and that was a major step forward that came out of Fairchild. In fact, when I look at the development of the integrated circuit, I always measure it from the first planar transistor rather than from the first integrated circuit.
RW: Well, IC's today are built the same way, are they not? With an oxide layer?
GM: That's right. Yeah, it's very much the same technology today. Now, when the patents for the planar transistor were being filed, Noyce was working with the patent attorney and the patent attorney suggested: "Now, have you looked at all the ramifications of this technology?" And Bob, who was director of research and development at Fairchild at that time went back... actually had a meeting of the senior staff there and during that meeting, he invented the two things that were needed to go from the planar transistor to an integrated circuit: the idea of using thin, film interconnections over the top of the silicon oxide, and the idea of using extra junctions in order to isolate one transistor from another. And he came up with both of those during the same meeting.
So, fortunately, we were really in the right path of the technology to do these things. Texas Instruments... Jack Kilby at Texas Instruments, had already built an integrated circuit but his was very much a laboratory device that was... had etched thin areas to make resistors, it had flying wire bonds that weren't really practical for anything like a production device. But the technology we had at Fairchild was the path to make the practical integrated circuit.
Noyce and Kilby are often given credit as co, inventors of the integrated circuit but what they contributed was dramatically different: Kilby made a laboratory model by hook and crook, Noyce took the planar technology and extended it so you could make a complete structure using the material processing operations we had developed. So you could cover a whole wafer with identical structures again and cut them apart and package them individually.
RW: Which again, is what is done today.
GM: Absolutely, this was the step along that path. So Fairchild really got a lot of that going.
RW: Well, at some point in time, you took over R&D and you built the facility at Palo Alto.
GM: Yeah, well that's another story. To digress a bit, when we set up Fairchild the first thing we knew was that none of us had any experience at all in running a company. And we'd seen how difficult that was at Shockley. So, we set out to hire our own boss. The eight of us went looking rather broadly for a general manager to come in and run the company. We advertised in the Wall Street Journal, looked around, and we found a fellow by the name of Ed Baldwin from Hughes Semiconductor. He'd been engineering manager at Hughes and he knew a lot of things about operating an enterprise that we didn't. So we hired Ed as our boss.
Ed never really felt Fairchild was his company. I still don't understand why he came in very early he had the same equity participation that the eight of the founders did... or at least he had access to it... in fact, he never put in his $500 and never got it. But about a year after he arrived and after we'd put the first products into production, he and several of the people he had brought in announced one morning they were leaving and they went down the street and set up Rheem Semiconductor. That was the time I became director of R&D. Until that time, I had had a position responsible for engineering the new processes and products and putting them into production and the quality organization, and Bob Noyce had a parallel position running research and development, looking at new devices.
When Ed left, the eight of us sat down to discuss what should we do should we go out and look for somebody else? And we decided that after being betrayed by the first guy we brought in that we would risk the fact that none of us knew much about running a business, although we knew more now than we had a year previously. And we decided that Bob should become the general manager. Then I moved over and took the research and development responsibility, and I had that at Fairchild until I left in 1968.
RW: Under your leadership, there was a tremendous amount of developments made.
GM: Well, Fairchild had the right technology at the time and as a result, I think, we were the most productive laboratory in the business for about a decade there at least. The integrated circuit developments principally came out of there, we made more and more complex circuits, extended the capability to linear circuits as well as to digital circuits, we did the basic work on making stable MOS devices. Some parallel work was done at RCA laboratories but most of it really came out of the Fairchild labs.
We tackled such problems as how one makes small quantities of quite complex circuits economically and came up with both of the approaches that have proven to be useful: the standard cell approach and the gate array approach at that time. Well, neither of those became practical applications while I was still at Fairchild. Subsequently they have become...
RW: Yes, you find that 1980 at LSI Logic is a clone of what we did at Fairchild. With a more modern semiconductor technology, to be sure, but the techniques at Fairchild... we developed logic simulation, place and route, we developed VLSI testers... which later became the Sentry series... so all that was done under your leadership.
GM: Yeah, well it done in the laboratory at least during that time... I don't know about my leadership... One thing we did that I don't think it's generally recognized, is the first CMOS circuits were made there. In fact, I remember this because my first trip to Europe in 1963 was to describe the advantages of CMOS circuitry for low power electronics. I went to several European countries with a group headed by Ed Keanjan, that was under some kind of NATO sponsorship... there's a book on micropower electronics that came out of that.
RW: Well, it was Phillips that developed LOCOS which was the forerunner of modern CMOS, is it not?
GM: Yeah, Phillips did LOCOS which really allowed for a somewhat denser structure. Essentially, you could put things closer together using LOCOS. And there had been a tremendous number of variations on the technology since. But Fairchild really was in the right place at the right time. Not only did we have a lot of technical contributions, that was a time period where it seems like every new idea that came along spawned one to five new companies. It really was the period of time when the "silicon valley effect" of all the spin, offs really blossomed. There's a genealogy chart that Don Hefler published in a couple of editions that shows a lot of the companies that ... can trace their origin back to Fairchild.
RW: Yeah, lots. Now when did you come up with "Moore's Law?"
GM: Moore's Law has been applied to ... one graph in an article I published in 1965. Electronics magazine's 35th anniversary edition asked me to predict the course of component technology for the next ten years and I looked at what we had done in integrated circuits. Integrated circuits then were about four years old. We had just gotten some of the first families out, making some a bit more complex, and I looked at what was happening on those and saw that the number of components that is the number of transistors or resistors in an integrated circuit was about doubling every year. So I just took that and said "What's gonna happen in components is going to continue to happen for the next ten years, so things will be a thousand times as complex in 1975 as they were in 1965."
And I think the most complex circuit we had around was 64 components when I did this so I was predicting 64,000 components in an integrated circuit by the mid, 70s. And amazingly enough, we stayed almost exactly on that curve for ten years.
RW: Did Carver Mead have anything to do with that, with his scaling work?
GM: Ah, no, it was completely independent of... we've talked... I certainly have talked with Carver off and on but he wasn't involved at all in that prediction.
Now, I modified that in 1975, suggesting it was going to slow down to more like a doubling every two years and I was a little bit too pessimistic then we've actually beat that. It doubles something between 18 months and two years.
RW: Yes, and when will it approach the number of atoms in the universe?
GM: I haven't extrapolated it that far. [Laughter.] That's one thing: any exponential like that predicts a disaster if you extrapolate it far enough.
RW: Ah, one of the questions that has come out of a number of lawsuits as of recently, that here on the west coast, in silicon valley, we didn't patent circuit designs or computer aided design for that matter. And as a result, we never patented at Fairchild the ROM and the RAM. Intel never patented the microprocessor... and others did. East coast companies or TI in particular, took our work and patented it. How come we never recognized the importance of that... circuit development?
GM: Well, it was probably a different attitude about patents. One thing that happened in the semiconductor industry... semiconductor processes are a long series of steps and the patents had gotten pretty broadly spread because all of the people working on the technology had some of them. And the net result was in order for any of us to operate we had to be cross, licensed so the participants tended to all cross, license one another. So, there was not a tremendous advantage to having more patents... with a couple of exceptions, there wasn't much net benefit from it.
What we never anticipated, I guess, was a lot of other participants were going to enter the business later on. So, at Fairchild we tended to patent relatively few things, typically the ones that we thought we could police most easily and were the most difficult to get around, you know, the more fundamental things. But, I was responsible for a lot of those decisions. I remember one in particular that, in retrospect, is kind of funny. In the early days of the integrated circuit, Bob Norman, one of the people who were involved there, suggested the idea of semiconductor memory... the whole idea of how semiconductor flip, flops could be used as a memory structure, and I decided it was so economically ridiculous, it didn't make any sense to file a patent on it.
You recognize that a few years later, semiconductor memory was the basis of...
RW: of Intel...
GM: ...why we founded Intel. It shows how hard it is to predict what's going to be happening.
RW: Now, you know, a guy that never gets any credit for his contributions is Lee Boisel. And Lee in his early days at Fairchild felt a lot of things... but what I found in my research is that he, in fact, developed the first microprocessor. It was never commercially sold but it was, essentially, an 8080 and it came before the 4004!
GM: Ah, I didn't realize that. Lee had done some complex circuits at Fairchild. Unfortunately, Lee was kind of a loner and, in particular at Fairchild, he wasn't in the laboratory, the group that doing all the MOS stuff. He was off in an applications group and there was very little cooperation between the two so, the stuff he was designing never really got into the mainstream. When he went off to set up Four Phase, he extended some of the stuff he was doing and, you know, I couldn't confirm that he did something before the 4004. The timing would have been pretty tough on that.
RW: Ah, he did and it was in... they still exist. They're in his early systems. He also did the first 1K DRAM which was... which again they kept proprietary to use in their systems and they never sold. So, it's quite interesting... I've come across the documentation that convinces me that that's indeed the case.
GM: How early was it?
RW: His essentially 8080 was in '69 and is still in use today by the IRS, unfortunately... as you know, they're not too swift on...
GM: ...modern technology...
RW: They and the FAA don't seem to upgrade too quickly.
GM: Well, you
ask about patenting on the microprocessor and frankly, we didn't think the
microprocessor per se was that patentable. What we had done was take a computer
architecture and make it all on one chip instead of on several chips. And
that was kind of the direction that the integrated circuit technology was
pushing in anyhow, always putting
more and more of the system on a chip. What TI did was then start saying: 'Well, a microprocessor with a keyboard is an "invention," and I'll admit, I never would have thought of filing patents on those things that TI got issued patents for.
RW: Yeah, well I mean it was really... minicomputers were around, controllers were around.
RW: The organizations were well, established and we all in our minds knew that someday we'll put it on one chip and it's just a question of having the process.
GM: Yeah, and you know Ted Hoff happened to recognize what the right time was, was very familiar with hardware efficient devices so was able to design one that was within the realm of being able to produce it economically.
GM: And he also saw that it was useful for much more than just a family of calculators, the particular problem he was solving at the time. So, I always give Ted full credit for the vision of where that was going to go.
RW: And it went a long way...
GM: It sure did. Well, it probably went further than his vision but it went further than any of our visions.
RW: It was incredible. I mean, he
talks about trying to convince the folks at Intel to do a microprocessor...
and the marketing folks and they said: "well, there something like
10,000 minicomputers in the world and, you know, we're starting late, the
most we could get would be 20% share of market that's only 2,000. It doesn't
make sense to go into
the business for just 2,000."
GM: Well, you know, everybody's memory of these things tends to be a little different, I think. I didn't really have any trouble at all going along with Ted's idea there. And the marketing people weren't really controlling the decisions on that. He may have thought he had to convince more people than he did... but once we got the Japanese calculator customer convinced to take that approach then we were off and running.
RW: Yeah. Well, we're already into Intel and we haven't left Fairchild really. Why did you guys leave and found Intel? What did you not like about Fairchild?
GM: Fairchild went through a peculiar period and I don't know all of the history, but they fired John Carter as CEO and put Dick Hodgson in and six months later almost six months to the day Hodgson was out as chief executive. You ought to get his story about what happened there... I've always been wanting to ask him but haven't.
GM: Yeah. My impression was he also got fired but I heard from someone else that he resigned, that he just didn't want to do the CEO job, which kind of surprised me. Anyhow, they had fired two CEOs... or two CEOs were gone within a six month period, and they were trying to run the company with a three, man committee of the board of directors, while looking on the outside for another CEO. The likely internal candidate was Bob Noyce... you know, certainly well, qualified by any measure, he'd run the most successful part of the company it was a world, leading operation, but they were going to bypass him. That kind of ticked Bob off and he decided that he didn't like that very much. And knowing Bob wasn't happy and was going to leave, and that we were going to have somebody else coming in from the outside, who'd probably want to make major changes in the operation I headed... I said "Ah, I think I'd rather leave before than after." So the two of us decided to leave and then went out and got financing to set up a company to look at new technology and new product areas in semiconductor memory.
RW: Was... your first product at Intel was bipolar...
GM: That's right. We set out to develop three technologies actually for memories: Schottky bipolar using aluminum Schottky diodes which hadn't been done previously; silicon gate MOS that had only been used for individual transistors and we had no idea if it could be done in production; and a multichip assembly where we were gonna to place several memory chips on a ceramic substrate and the like.
It turns out that the bipolar technology worked out extremely easily. The aluminum Schottky diodes worked better than we had any reason to believe and the net result was that product went into production first, so we got our 64, bit bipolar memory out... I think it was in August of the year after we started, and the MOS device followed about a month later so we had a 64, bit bipolar in August and 256, bit MOS in September. Those were our first two products. We never did succeed in getting a multichip product out... still trying... we have one in the P6 now, it's two chips.
RW: OK, let's break here.
RW: Intel corporate culture... and I worked for both Intel and Fairchild... Intel corporate culture was markedly different than Fairchild... night and day. Was that a conscious decision?
GM: Ah, I think it's more a case of having a chance to start over again. At Fairchild, everything was a surprise, it was really on, the, job training for everybody. At Intel, we had a chance to take advantage of what we learned at Fairchild and set things up a bit differently. Somewhat different people in cHRGe. It started out in the beginning... well, in the beginning, Noyce and I kind of running it as a team. And then I think we shared more or less philosophical concepts about how some of the things ought to get done. You know, certainly the egalitarian nature if you want a close parking place you get to work early...
RW: Right. No mahogany row.
GM: Yeah, in fact, no office partitions these days... It was just, I guess, a kind of a west coast approach to a new company.
RW: Ah, I can name some others. Openness. Remember Momars?
GM: Oh yeah, communications was very important, we saw some of the Shockley problems that... we didn't have that... we were fairly open at Fairchild too... certainly in the laboratory I always thought we were open...
RW: And seminars once a week...
GM: Yeah. And broad distribution of the progress reports. We had quite a few things going on there that wasn't intended to be different. Intel had a much narrower focus than something like the laboratory at Fairchild when we got started. We were all aiming in a fairly narrow direction to solve some particular problems.
RW: But you also didn't have a separate R&D.
GM: No, that was definitely by design. One thing that was frustrating to me at Fairchild was the lab was being very productive but we were having an increasingly difficult time to get the new things we did into production. The production people were busy but also I found that the more technically competent the production people became, the more difficult it was to transfer something new to them. When the laboratory was obviously the source of the technical information, we could tell them what to do and they would do it. The more technically competent they became, the less willing they were to accept it that way. They wanted to re, develop it rather than take what had already been done.
My biggest frustration was MOS, where we had stable devices in the laboratory in 1962, maybe 1961, and by the time I left in 1968, Fairchild was not yet successful in MOS. In spite of the fact there was even a spin, off of one of our spin, offs that was making MOS devices by then using what Fairchild had learned. So when we set up Intel, we decided we would avoid that split between R&D and manufacturing. Wed be willing to accept less efficient manufacturing for a more efficient transfer process and make the R&D people actually do their development work right in the production facility and we have continued that with some variations ever since.
Now, we have gotten to the point where we can give the development people a full, blown factory to set up from the beginning and tell them their development job isn't done until they're showing high yields at high volumes.
RW: Umm, you know, one of the other interesting things about Intel is the number of failed products... that have come all the way from MICROMA to micropower to the 432. Umm, to the 2, bit bipolar slice. More failed products than many successful companies. And yet, Intel is tremendously successful. What do you attribute that to?
GM: Well, in the first place, I probably wouldn't agree with you completely on which were failed products... I still wear my MICROMA watch! [Laughter.] This is my $15 million watch! It still wakes me up every morning...
We did misunderstand the electronic watch business. We were the first ones to market with a liquid crystal display watch, we were one of the first ones out of the business. Our initial view was that this was the beginnings of a personal electronic system that would grow in complexity and capability. In fact, digital watches, at least for a while, went quite the other way to be the cheapest way to tell time.
Ah, the 432, which was a very aggressive shot at a new microprocessor. I guess, to a significant extent, I'm personally responsible for that in that when we'd completed the 8080 which was a landmark microprocessor I sat down with some of our designers and said "look, we probably have one more chance to start over and do this right." So unfettered by anything we've done in the past, compatibility or anything, go out and make the right microprocessor.
Now the group, or one of the leaders of the groups was a young computer
scientist who knew what all the new thinking was in the computer science
departments, and every bell and whistle you could imagine was put in there.
This was to execute object, oriented code only, it had objects embedded in
the hardware, it was fault, tolerant in that it was designed
for transparent multiprocessing if your system wasn't going fast enough you could just plug in a new one and make it go faster. Anyhow, the functionality was extended so far that in order to make a product, they had to take out everything that related to performance. The net result was when we finally made the chips, while they gave the functionality, the performance level was so far below that of conventional microprocessors that it didn't hit any market at all.
But in fact, it was also the wrong concept for the time. It was really doing another closed system, since the hardware and the software were intimately combined and the like, and the whole market was really moving toward open systems at that time. So, we took way too big a step. It was the subject, in fact it resulted in a textbook that we used in many computer science courses around the country because it did embody all these advanced architectural features. But from a product point of view it was at that time a failure.
Now, it has been reincarnated a couple of times since then and much of the stuff is still embodied in our 960 family of microcontrollers, but not at all the full, blown stuff that was initially involved in the 432 project.
RW: I guess my point was not to assess blame here but to say that how successfully Intel has recovered from products that didn't work and a lot of companies don't do that. So, there must be some mechanism where you're ready to jettison an idea when it proves it doesn't work instead of just hanging in there with it for ever.
GM: Well, none of these were probably undertaken at a time when they were set up in a manner that would kill us, you know. While we were doing the 432, for example, one of our people Ed Gelbach really convinced us we needed a plain, ordinary 16, bit processor, preferably compatible with the 8080. This resulted in the 8086 and its companion product, the 8088, which was the basis of the first IBM PC. So we had kind of a parallel path going that was conventional. But I think, in my career, I've always been a sucker for a new computer architecture. I supported one at Fairchild you may remember Rex Rice's [?] project, Symbol? That was a major new step to a new computer architecture that probably would have been revolutionary but never was successful.
RW: It did result in the DIP [Dual Inline Package] though.
GM: That's right.
RW: Which was tremendously successful.
GM: That's true... these things tend to have fallout. But, when you're a startup company you have to bet the company on a lot of the programs you do. When you get bigger you only want to bet half the company if you can do that. You don't want to do something that if it doesn't work, will put you out of business completely.
On the other hand, you gotta keep reaching. If everything you do works, you're probably not trying hard enough.
RW: Well, I think thats one nice thing we can say about silicon valley is it's OK to fail. We've all done it...
GM: That's right. I used to say that when you know there was really... people weren't risking anything when they went and tried to start a new company because if they failed they could still get jobs for 20% more because they're experienced. There was really no stigma at all to failing and that's been an important part of this area... just, all the new companies that have formed with relatively no concern about the risks.
RW: Yeah. Well, I'd like to end up with your ideas on the economics of wafer fabs as they're moving now.
GM: Well, this is something that has changed dramatically over the last several years. When Intel was started in 1968, we developed our first products and put our manufacturing facility in place and developed the processes for the $3 million that we started with. At that time, a piece of equipment a bank of furnaces or an evaporator, or a lithography machine each cost about $12,000.
Now when we build a plant, each of those pieces of equipment costs in the $3, 5 million range. It's just been a huge escalation in the cost of equipment. Just hard to think that for the cost of one piece of equipment now we started Intel. The net result is, when you combine that with the increase in process complexity so there are more process stepsa modern factory is a billion dollars plus. In fact, Intel is building now factories that when they are filled with equipment will be in the $2.5 billion range, and I suspect we'll be building 3 or $4 billion factories before the end of the decade.
Fortunately, the industry is growing fast enough that it can afford to do this. But, the equipment cost has certainly been a rapidly growing portion of the total cost and I'm concerned that the economics of the equipment are more likely to limit progress in the future than any fundamental physical problems.
RW: Well these billion dollar fabs... what's their effective lifetime?
GM: We generally figure a fab will be good for two generations of technology, which means about six years.
RW: You have to amortize this thing over six years?
GM: Yeah, well... Intel typically writes off its equipment in four years. That's roughly its economic lifetime. At the end of the two generations of technology, hopefully we can refurbish the building and use it for another couple of generations. In fact, the ones we've abandoned have generally been because they're too small. Our whole idea of what an economical fab is has just changed dramatically.
RW: But amortizing a billion dollars a year is kind of difficult, isn't it?
GM: Well, it's hard for a small company!
RW: That's for sure.
GM: It is, and it's taken a while to get used to these new numbers. I remember when we were first looking a $100 million years for capital. I really had a tough time convincing myself that that was going to happen. This year Intel will spend $3.5 billion dollars on capital buildings plus equipment and probably more than that next year.
RW: Don't you have some deal where there are floating municipal bonds, or something?
GM: We do in, at least in New Mexico, and when you're buying a billion dollars worth of equipment, if you have to pay sales tax on it, for example, it's a huge amount of money. By working a deal with the local authorities, the equipment is actually owned by the county, that doesn't pay any sales tax, and we finance that by buying municipal bonds from them. So it's actually a scheme to give us a tax incentive to put our plant in New Mexico rather than someplace else.
RW: You're getting bigger than some companies... countries. Are you not, in the gross national product of some small...?
GM: Well, we've been very successful. This year, Intel's revenues will be about $16 billion.
RW: That's probably bigger than Zaire... or some of the others... Well thanks Gordon...
RW: I appreciate it. This has been great.
[an error occurred while processing this directive]