Gordon Moore on:
Imagine being in the room to witness the genesis of an invention that would profoundly change human civilization. Gordon E. Moore was there; he supervised the project 25 years ago that put the first "computer on a chip."
As the co-founder of a Silicon Valley startup known as Intel Corp.. Moore has seen feat after feat of technical prowess sustain a geometric growth in computer power so regular you can practically set your watch by it. The growth is so predictable and important, it has been canonized as a law--"Moore's Law." It was Moore who noticed as early as 1965 that microchips were doubling in circuit density (and thus in their potential computational power) every year or so.
Moore retired as chairman of Intel in 1989. In 1990, he received the National Medal of Technology from then-President George Bush. Now 70, he still works several days a week in his old cubicle at Intel. In September 1997, Moore took some time to speak with Scientific American's west coast editor, W. Wayt Gibbs. Their wide ranging discussion is presented in four parts.
In part one, Moore describes how he narrowly missed studying nuclear bombs instead of microchips, how he helped found a multibillion-dollar company, and how he snookered Japanese investors out of rights to microprocessors. Part two recounts the discovery of Moore's Law and his take on the formidable technical obstacles it is beginning to face.
In Part three Moore describes the technological tricks that Intel engineers are developing to keep the computer revolution humming. In the final section, Moore reveals his predictions for the next 10 years of computing.
PART 1: THE BIRTH OF THE MICROPROCESSOR
SCIENTIFIC AMERICAN: When you were younger were you more interested in science than in engineering?
GORDON MOORE: Yes, from the time I was in junior high school I decided I wanted to be a chemist. I didn't quite know what a chemist was, but I kept it up and got my Ph.D. in physical chemistry. My first job out of school was to do basic research at Johns Hopkins University's applied physics lab. Then William Shockley [co-discoverer of semiconductors] caught up with me.
SA: So, tell me about Johns Hopkins. Were you working on weapons research and military research?
GM: I wasn't in weapons research. In the applied physics lab there was what they called the research center, which was essentially doing basic research on . . . well, on whatever we were interested in, almost. I was looking at the shapes of infrared absorption lines and some spectroscopic study of flames.
SA: Why were you dissatisfied with that job?
GM: First of all, the team I was with was breaking up, and my two bosses were moving out. I started calculating the cost per word in the published articles [that emerged from the lab] and decided that at $5 a word, I wasn't sure that the government was getting its money's worth. I didn't know whether anybody was even reading them.
I decided I really wanted to get closer to something with practical application. I was looking for some technical stuff that would lead to a real product. I didn't know quite what.
I interviewed at several places, one of which was the Lawrence Livermore Lab out here, and considered thermonuclear devices as a practical application. That's where Shockley got my name, actually. They made an offer to me, and I decided to take the position of Inspector of Nuclear Explosions.
SA: That's what the job involved?
GM: Essentially. Shockley got permission to go through Livermore Lab's files of all the people to whom they had made offers or who had turned them down. He thought he needed a chemist in his new operation; chemists had proved useful things in his group at Bell Laboratories. So that's how he came up with me. I had no background whatsoever in semiconductors. I did at least know who Shockley was, because I had heard him talk in Washington.
SA: So that background wasn't necessary for the job?
GM: No, there weren't an awful lot of people around in those days who did know much about semiconductors.
SA: What happened with Shockley? That didn't work out for very long either, did it?
GM: I was there for about a year and a half. We fiddled around, trying to make some devices. Then Shockley changed direction. When I first went to work for him, he was thinking of making a transistor. But then he decided he wanted to make a rather obscure device called a four-layer diode. Mainly it was just that while Shockley was a technical genius, he really didn't understand how people worked very well. He stirred up things internally.
A group of us finally went around Shockley to try to get that straightened out. We went to Arnold Beckman, the source of his funding. We thought we were making considerable progress. But Beckman finally told us that Shockley was the boss, and we'd just have to learn to live with him.
SA: Sounds like a mutiny of sorts.
GM: Well, it was, I mean we had burned our bridge pretty badly. So after doing that, eight of us felt that we really had to leave and go someplace else. We didn't want that kind of situation there long-term. We were later called the Traitorous Eight and a variety of other things. We founded Fairchild Semiconductor.
SA: Now there was already a Fairchild making other things, right?
GM: Fairchild Camera and Instrument supported it. To be more complete, we sat down with a copy of The Wall Street Journal and went down the companies in the New York Stock Exchange to see which ones might like to start a semiconductor operation. We identified something like 35 companies. The investment bankers that we were dealing with, one of whom was named Arthur Rock, a Harvard business school graduate at the time and a senior partner, went and contacted all the companies we identified, and they all turned it down without even talking to the group.
Then they caught up with Sherman Fairchild, who was the founder of Fairchild Camera and Instrument and also of Fairchild Aircraft, and he liked technology. He introduced them to the person who was then running Fairchild Camera and Instrument. They sent a representative out to talk with us and decided to support us in founding a new company.
SA: So now you had your own company, yet that didn't work out perfectly either.
GM: There was a lot of on-the-job learning there. It was quite technically successful, and we developed a fairly large business out of it. It became a division of Fairchild a few years later when they exercised an option to buy it out. Our division was something like 30,000 employees and $150 million. So it was a fairly successful operation.
But then in the parent company there were some changes that I never have understood. They fired two chief executives within six months. They fired the first and put in the person with whom we had dealt most. And then he was out, and they started trying to run the company with a three-man committee of the board of directors, while starting to look on the outside for a new president.
Robert Noyce [a co-founder of Fairchild Semiconductor and later of Intel] was the logical internal candidate, but he wasn't too enthused about the way that was going. I could see that the company was going to change quite a bit, because they went outside for someone.
So I decided I'd rather leave before than after the changes. And I'd been a bit frustrated with what I was doing. I was running the laboratory, and it was getting harder and harder to transfer stuff to manufacturing.
SA: The walls were growing higher and higher?
GM: Yes, the wall grew higher and higher. The more technically competent the manufacturing people became, the less willing they were to accept the things that had to be done to the technology coming out of the lab. So, out of the convergence of those things, Bob decided to resign, and I decided that I would, too.
SA: Were you two pretty close at that point?
GM: We'd been working with one another since the early Shockley days. He came to work for Shockley on Friday, and I arrived the following Monday. At Fairchild, at the beginning we were peers, then he became general manager of the division, so he became my boss for a long time.
SA: Legend has it that Bob typed up a simple one-page business plan for Intel on his home typewriter. Is that actually true?
GM: There is a copy of it downstairs. It says absolutely nothing. It is completely and utterly vague. Our plans were a little more concrete than that suggests!
The idea we had for Intel was to try to make complex integrated circuits. The problem was that when you define complex integrated circuits, they tended to be unique--you know, used once in a computer system or something. And we saw semiconductor memory as an opportunity to make something complex and sell it for all kinds of digital applications.
So that was the first thing we went after. That was really the basis of our business plan. And then we were looking for other products that had the same kind of characteristics, namely complex chips that could be made in large volume.
The idea of the electronic calculator was really just getting going then. We started looking for a calculator company that we could deal with. But we were a little late--all of the calculator companies you'd heard of had already tied up with a semiconductor company.
So we caught up with Busicom, which was a Japanese start-up--a peculiar operation in itself, not very well financed. But they wanted to build a business in scientific calculators, and they were looking for a semiconductor partner. They came in, actually, with all of the logic done for their family of calculators, something like 13 chips with considerable complexity.
SA: They had done all the design work?
GM: They had done all the design work on those. We had a small engineering group, and most of the people were up to their eyeballs in memory circuits, so we didn't have a lot of engineering to put on something like this. To do 13 different complex custom circuits was far beyond what we could tackle.
Then one of the guys we had looking at this, Ted Hoff, looked at what they were trying to do and told us that with a general-purpose computer architecture, he thought he could do all of their calculators. Beyond that, I remember him suggesting elevator controls and traffic light controls as two specific things--this would be a general-purpose controller, too.
His real insight was seeing that this could be done with about the complexity of the MOS [metal oxide semiconductor] memory circuit we were making then. So the idea of a single chip computer was something to talk about in the industry, way in the future, a "someday" kind of a deal. But Ted saw that we were at the point where we could actually do that.
SA: Was his insight to see that you didn't actually need as many elements as the industry guys had thought you would need or that in fact you had enough elements now?
GM: Maybe a little bit of each. He'd been working with the old DEC PDP-8, which was a relatively hardware-efficient computer, and he knew it very well. So he knew some of the techniques used to make things hardware efficient. And he just recognized that with about the same complexity as the memory chip you could make a simple processing unit. That cut the project down to bite size because you had one special chip to design, instead of a dozen or so, and a couple of memories, which were just variations of what we were doing anyhow. So this knocked it down to something we could actually try.
Then our problem became selling this to Busicom--making them throw away all of the design work they had done and start all over with a little start-up company here in Silicon Valley. I remember we faced that meeting with a lot of trepidation. We had the chief technical guy from Busicom visiting us with one of his engineers. We went in there and gave a pitch on this, expecting a lot of push back, but he said, "Fine, we'll do it your way."
SA: He didn't have any doubts that you could pull this off?
GM: Not at all, and more than that he didn't offer any alternatives. He immediately agreed, which really shocked me. I thought we were going to have a really tough selling job.
SA: Did you already in the back of your mind think: "Gee, this is a great way to boost our business--really to fund an R&D project for us, because we can then crank out different versions of these things and sell them to others?"
GM: Well, initially we were looking at it as a way to get into the calculator business. And we knew it had potential beyond that because Ted had pointed that out first. But it was just another one of these chips that we could make in fairly large volume.
In fact, the way the development got done, Busicom paid a portion of the development costs and therefore owned the rights to the design. So in the beginning, we weren't able to sell it for these other applications.
Then Busicom was meeting a lot of cost pressure in the calculator business and wanted to get lower prices for the chips. We started shipping these chips in early February 1971. Busicom wanted lower prices, and we wanted higher volume. We negotiated a deal to give them lower prices if we could have the rights to sell this chip for other applications.
So we got the rights for noncalculator applications by giving them lower prices on the things they had. And then eventually, when Busicom got into deeper financial trouble, we essentially gave them back their $65,000 and got the rights to the chips back for all uses. So the Japanese initially owned all the rights to microprocessors, but sold them for 65 grand. In retrospect, it was kind of like the purchase of Manhattan [from the Native Americans, for $24].
SA: I've read that the 4004 [Intel's first commercial microprocessor] took nine months to design and create. Do you remember how many engineers worked on that?
GM: In those days all chips took about nine months. It wasn't very many; something like four. In fact, one problem we had was we didn't have a team ready to take on the design right away. We had to go out and hire some more people. I don't remember the exact timing, but one of the Busicom engineers, named Shima, was due to come over here to meet with Intel and check our progress.
We had just hired Federico Faggin, who ran the design team, a week before the guy got here. So he arrived, and it was obvious that we hadn't done anything! Shima ended up staying over here as a Busicom employee and then later worked for Intel.
SA: It would be interesting to contrast the manpower that went into the 4004 with what went into, say, the Pentium II.
GM: My recollection is that about four engineers worked on the 4004 as well. Now to design one of our chips we have more like 400 engineers, often spread around several different sites. And today it takes more like four years. It's a much bigger deal.