The Intel 4004 Microprocessor:

What Constituted Invention?

WILLIAM ASPRAY

This paper investigates the context for the development of one of the earliest
microprocessors, the Intel 4004. It considers the contributions Intel employ-
ees, most notably Marcian E. “Ted” Hoff, Jr., and Federico Faggin, made
and contributions other people made to this development who are not gen-
erally known, most notably Tadashi Sasaki and Masatoshi Shima. The paper
represents a case study of how corporate and national cultures affect tech-
nological development and of the many aspects of invention, including con-
ceptualization, logical design, engineering, fabrication, capitalization, and
marketing.

Introduction
T he microprocessor is among the most ubiquitous and power-
ful technologies of the late 20th century. It is not surprising,
then, that credit for this invention has been widely discussed—
By considering corporate and even national cultures, we can gain
a new and deeper perspective on this important invention.

especially in the past several years, since an obscure inventor Background

named Gilbert Hyatt was awarded a U.S. patent, to the consterna- During the 1950s, the basic control element in computers, widely
1
tion of the semiconductor industry. With many inventions, there known as the logic gate or switch, was implemented using vac-
is a tendency, after their significance becomes widely appreciated, uum tubes or discrete transistors. Integrated circuit technology,
to reconstruct the history, making the story simpler, more rational, using layers of metal and oxide on a polished silicon chip, began
and more heroic. This is the case with the microprocessor, which to be used to implement logic components for computers, replac-
is widely credited to the engineering genius of Marcian “Ted” ing transistors, diodes, and resistors on printed circuit boards. In
2
Hoff at Intel. the early 1960s, the scale of integration, a measure of the number
It is not my intention to disparage the accomplishments of ei- of logic components that could be placed on a chip, was a few
ther Hoff or Intel. Hoff is a gifted and accomplished engineer who dozen; and chips were used mainly to implement individual logic
played a fundamental role in the development of Intel’s first mi- devices. By the mid-1960s, with continuing improvements in the
croprocessor, the 4004; and Intel has repeatedly demonstrated its semiconductor field—mainly, better etching and manufacturing
capabilities while becoming a dominant force in the global semi- techniques and improved circuit design—more functionally com-
conductor industry. But the story of the invention and develop- plex logic devices such as adders and shifters could be imple-
ment of the microprocessor is not as simple or straightforward as mented on a chip. The scale of integration continued to rise rap-
is generally told. idly, approximately doubling every year. By the mid-to-late
This paper aims to tell a more interesting and complex story by 1960s, it was practical to begin considering semiconductor de-
examining the historical context in which the 4004 was devel- vices to replace magnetic cores as the primary storage device for
3
oped, both outside and inside Intel. I will show, for example: mainframe computers.
• that the conceptualization of the microprocessor, which was Technological conditions were favorable for the development
Hoff’s principal contribution to the 4004 project, was inde- of the microprocessor in the mid-to-late 1960s. The increase in
pendently conceived in other companies and Hoff was scale of integration made it possible to place ever more elements
aware of some of this work; of a computer circuit onto a single chip, leading engineers to
• that Hoff had a relatively minor role in the hard work of speculate about building a computer on a chip. Federico Faggin,
making the 4004 a commercial reality (i.e., the detailed who led the engineering work on the 4004 project at Intel, ac-
logical design, engineering, applications development, and knowledged this fact:
marketing); By the mid-1960s people were building single-board micro-
• that the stimulus and financing for the 4004 project came computers, using MSI [medium-scale integration] as a side
from a Japanese company, not from Intel; and function. It didn’t take geniuses to figure out that this pat-
• that Intel did not originally embrace the microprocessor as tern of continuing integration and combining functions was
an important part of its product line. going to happen. To people in the art and inside the industry,

1058-6180/97/$10.00 ” 1997 IEEE

4 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

 it was the natural thing to do. The question was, “When will During his years with Kobe Kogyo, Sasaki was a leader in es-
4
we have the technology that will allow it economically?” tablishing the Japanese semiconductor industry. He met John
In an interview with the author, Hoff indicated that, at the time Bardeen on a tour of Bell Laboratories in 1946. During 1947, just
he conceived the microprocessor that became the Intel 4004, he before Bardeen and his coworkers invented the transistor, Sasaki
was aware of similar ideas engineers had expressed at several had several communications with Bardeen about the limits of
5
other organizations, including SRI, IBM, and RCA. Additional miniaturizing vacuum tubes. As Sasaki explained it, while he
work was being conducted at Fairchild, Rockwell, General In- worked to make the distance between the grid and the cathode
struments, and Texas Instruments at about the same time, although progressively smaller, Bardeen made the conceptual leap to em-
6
there is no reason to believe that Hoff was aware of it. As I shall bed the grid in the cathode, which resulted in the point-contact
show in the next section, a similar idea had also been discussed at transistor. Bardeen personally sent Sasaki news of his invention
the Sharp Corporation in Japan. within a few weeks after it occurred. Sasaki recognized the sig-
The economic climate was also favorable to the development nificance of Bardeen’s contribution and promptly initiated a re-
of the microprocessor. In a thoughtful and widely read review search program at Kobe Kogyo on transistors, later supported by
article, Hoff and Intel President Robert Noyce made this case: the Ministry of International Trade and Industry. Leo Esaki, who
won the 1973 Nobel Prize in physics for his research on semicon-
By the late 1960s, the semiconductor industry was becom- ductors, was the first researcher to work on transistors under
ing aware of a serious design problem. The complexity of Sasaki’s direction. Sasaki’s interest in semiconductors has contin-
random logic designs was increasing steadily. If this contin- ued up to today.
ued, the number of circuits needed would proliferate beyond
the available supply of circuit designers. At the same time,
the relative usage of each circuit would fall. IC [integrated Bardeen made the conceptual leap to
circuit] cost effectiveness would suffer; increased design embed the grid in the cathode, which
cost and diminished usage would prevent manufacturers
from amortizing costs over a large user population and
resulted in the point-contact transistor.
7
would cut off the advantages of the learning curve.
In 1962, the British company Sumlock Computer began manu-
Hoff and Noyce went on to argue that “the microprocessor facturing a desktop calculator, known as the Comptometer Mark
was a very necessary invention—and that its rapid acceptance IV, which became a major commercial success in Japan. Japanese
8
was in many ways predetermined.” This article disputes their companies were struggling to make calculators using vacuum tube
claim for autonomous technological development. Technologi- components, and, spurred by this foreign threat, Sasaki sought to
cal and economic forces set a context for the development of the replace vacuum tube components in Japanese calculators with
microprocessor, but it was human choices by the semiconductor transistors. At the time, Kobe Kogyo was supplying electronic
manufacturers, systems manufacturers, users, and perhaps oth- components to Sharp Corporation, a leading manufacturer of do-
ers that shaped the technology. In fact, on the same page as their mestic appliances. Sasaki persuaded Sharp, which had no experi-
assertion about the necessity of the microprocessor, Hoff and ence in the calculator business, to send several of its young engi-
Noyce suggest some of the alternative approaches that might neers back to the university for additional training in computing
have been taken to solve this technicoeconomic problem, such as a first step toward entering the semiconductor-driven calculator
as computer-aided design, discretionary wiring, and master business.
slicing. Because of management restructuring within Fujitsu, Sasaki
decided to leave for a senior management position at Sharp in
1964. His arrival at Sharp coincided with the return of the engi-
Tadashi Sasaki and the Early Japanese neers from the university where they had received their extra
Semiconductor Industry computer training. Unlike Kobe Kogyo, Sharp was in a financial
Although it is little known outside Japan, one important line in the position to provide adequate development funds, and Sasaki was
story of the microprocessor’s development began with electrical able to put the engineering staff to work developing semiconduc-
9
engineer Tadashi Sasaki. Because Sasaki is not widely known in tor-based calculator products. He believed that if he could replace
the West, I will take some space to detail his career. He was born the large mechanical calculators then in common usage with
in 1915 in Taiwan but grew up in Japan. In 1938, he graduated smaller electronic ones, the new calculators might find a market in
from Kyoto University with a distinguished record and a bache- the home and build on Sharp’s experience in domestic electrical
lor’s degree in electrical engineering. He worked for a short time appliances. Under Sasaki’s sponsorship, Sharp soon produced the
on circuit design at the Electrotechnical Laboratory—a preemi- world’s first transistorized calculator.
nent research laboratory sponsored by the Ministry of Telecom- Sasaki closely monitored the U.S. semiconductor industry,
munications—before moving to the aircraft manufacturer Ka- with the intention of serving both his own company and the entire
wanishi Kogyo, which was integrated into Kobe Kogyo after Japanese electronics industry. He observed the U.S. move from
World War II and absorbed by Fujitsu in 1963. Sasaki worked small- to medium-scale integration and persuaded several of his
on antiradar devices during World War II for Kobe Kogyo’s Japanese colleagues to follow suit. He argued, with less success,
vacuum tube division. Dr. Sasaki, who received his PhD in for the move from bipolar to majority-carrier devices. In what
electrical engineering from Kyoto University in 1961, remained seems to be a characteristically Japanese management style,
with Fujitsu until 1964, serving in research and management Sasaki used analogy to explain and partly justify his preference
positions of increasing responsibility. for single-carrier devices:

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 5

The Intel 4004 Microprocessor

At the time the Sharp Corporation hired many, many young tion, but the other engineers did not find particular value in it.
girls for the assembly operation, for production. Sharp Cor- Sasaki made his decision on the future research direction in a
poration had built many dormitories for several-person oc- customary Japanese way, on the basis of the majority opinion,
cupancy. They also built housing complexes for married which he later acknowledged to be a mistake. He did arrange for
couples. Then I realized that what is more efficient and cost the company’s shops to experiment with a program he called
effective is singles’ dormitories; they are much better than Components on Silicon, but the major thrust of research was
the couples’ dormitory because in the case of the singles’ along the lines of the majority opinion, without any particular
dormitory, you don’t have to think up any design. It’s very effort to build a complete central processing unit on a chip.
simple, and you don’t have to worry about any separations However, Sasaki would not readily let go of the idea of the
between them. Then I put that idea into semiconductor de- fractional chip. The U.S. manufacturer Rockwell had an exclusive
vices. People have used the p-n junction, and they had to contract to supply the semiconductor devices Sharp used in its
put some insulator with it. It’s no good for the future appliances. Sharp was an important customer to Rockwell, and
miniaturization of devices. So I recommended utilizing the there were mature working relations between them. Sasaki asked
10
single type of transistor structure without insulator. Rockwell to produce these four-division chips, but Rockwell re-
In 1966, Sharp began work on p-channel MOS technology, fused—according to Sasaki because Rockwell was already earn-
which it learned about from a technical report prepared by ing high profits with its other semiconductor devices and did not
Fairchild Semiconductor. Two years later, Sasaki began to worry want this distraction.
that calculators built with p-channel MOS technology would con-
sume too much power for battery operation and would operate too Thus, while the other Sharp engineers
slowly, so he introduced work on complementary metal-oxide
semiconductor technology in his company. However, this created focused on the functionality of the
a new problem: Too great a percentage of the power was being calculator, Murakami focused on the
consumed by the display unit. In response, he initiated what
turned out to be one of the first and most successful research pro-
functionality of the chip.
grams in liquid crystal displays.
Sasaki’s role in the development of the microprocessor came In 1968, in the first several months after Intel was founded,
about as part of his effort to keep Sharp at the forefront of the Noyce visited Sasaki at the Sharp offices in Nara, Japan, hoping
Japanese calculator business. During the mid-1960s, every two to sell Intel’s manufacturing semiconductor devices to Sharp.
years the number of chips required to make a calculator was Sasaki felt beholden to Noyce because of the important use he had
halved because of the rapid pace at which the semiconductor scale made of Noyce’s earlier results at Fairchild on planar-type semi-
of integration was progressing. As a result, Sharp anticipated conductor devices; so he asked Rockwell if it would allow Intel, a
product life cycles for calculators of two years or less. By 1968, new and struggling firm, to produce a small percentage of Sharp’s
some calculators were being built from as few as four chips. Pro- semiconductors. Rockwell refused, pointing to its exclusivity
jecting a continuation of this trend, Sasaki organized a series of agreement with Sharp.
brainstorming sessions with about 10 of his engineers and circuit At that point, Sasaki decided to take entrepreneurial action be-
designers to plan for the future generations of calculators. hind the scenes. Yoshio Kojima was the president of a young and
Most of the engineers suggested an incremental increase in the unproven calculator company named Busicom, short for Business
overall functionality of the calculators by taking advantage of the Computer, that was experiencing some financial problems. He and
increasing scale of integration to continue to put more on each Sasaki had graduated from the same university department
chip. Sasaki rejected this idea as simply the kind of conservative (although at different times), and university ties being strong in
thinking that is ingrained by the Japanese university system—to Japan, Sasaki did not want to see Kojima’s business fail. Sasaki
make incremental extensions rather than technological leaps. had previously given technical advice to Busicom, which is per-
However, one member of the team, Ms. Murakami, who was a missible in Japanese business culture since Busicom was small
software engineering researcher who had recently graduated from and posed no serious threat to Sharp. Thus he decided surrepti-
Otani University in Kyoto, focused instead on increasing the tiously to provide 40 million yen to Kojima, with the stipulation
functionality of individual chips. Noting the reduction in the that Busicom would front a contract with Intel to manufacture the
number of chips needed to build a calculator over the previous four-division chip. This funding was used to pay for the develop-
several years, she suggested that if the trend continued, the num- ment contract with Intel that led to the development of the 4004
ber needed would decrease to two, then one, then a fractional microprocessor.
number.
What she meant by a fractional number is not entirely clear, Masatoshi Shima and Busicom
but it seems she meant that the calculator comprised multiple At this point, the story shifts to Busicom and a young engineer
functional units and that one could divide a chip into multiple employed there named Masatoshi Shima (Fig. 1). In 1967, Busi-
regions, placing in each region a complete functional unit, with com was manufacturing mechanical and electronic desktop busi-
buffers to bypass. This leads one to emphasize the implementation ness calculators, selling Mitsubishi mainframe computers, devel-
of an entire functional unit, such as a central processing unit, on a oping operating systems and applications software for Mitsubishi,
single chip (or portion thereof). Thus, while the other Sharp engi- and importing business computers from France. Busicom calcu-
neers focused on the functionality of the calculator, Murakami lators (Fig. 2) were sold in the United States under NCR’s trade
focused on the functionality of the chip. Sasaki liked her sugges- name and were also exported to Europe and other parts of Asia.

6 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

 Shima, who was born in Shizuoka, Japan, in 1943, graduated learn as much about the physical properties of transistors as about
with a major in organic chemistry from Tohhoku University in the logical design of the calculator as a system (main processor,
1967. That year, there was a downturn in the chemical industry. memories, display, keyboard, and other peripherals) and about the
Not able to find a good job as a chemist or chemical engineer, he architecture of integrated circuits. Thus he found he was not par-
accepted a position as a software engineer at Busicom. Computers ticularly handicapped by his lack of traditional grounding in elec-
were just beginning to be used in Japan in the chemical analysis trical engineering.
of organic compounds when he went to college, and hoping for a The Japanese calculator industry was then undergoing rapid
career in scientific programming, he joined Busicom. transformation, driven by the rapid innovations in the microelec-
11
Some of Shima’s first work at Busicom involved programming tronics components used in calculators. A particularly landmark
the Mitsubishi MELCOM 3000 computer, even though he had event was Seiko’s introduction of a fast, compact impact line
never had any formal training in either computers or electronics. printer for the 1964 Tokyo Olympics, which all the Japanese cal-
However, he did not find sufficient challenge in the programming culator manufacturers wanted to incorporate in their products. By
tasks assigned to him, especially since most were business rather this time, virtually all calculators were being made in Japan. U.S.
than scientific applications. Busicom had a factory in Osaka, Ja- and European firms had dominated the mechanical and electro-
pan, that manufactured electric desktop calculators, and he was mechanical calculator businesses earlier in the century, but they
granted permission to transfer there to work as a hardware engi- had been slow to make their products electronic, and market share
neer after completion of six months of employment with the com- had shifted rapidly to Japan. Only one U.S. company, Monroe,
pany. He had no training in hardware engineering either, but he was still manufacturing calculators.
taught himself some rudimentary electronics by reading textbooks Busicom was not immune to these competitive forces. Tadashi
and attending a one-month course on automation. Tanba, Shima’s supervisor in Osaka, had used a hardwired ap-
proach—the way in which calculators had traditionally been de-
signed —to design several calculators for Busicom. But because
of the competitive pressures for more rapid product development,
Tanba decided to draw on his experience in the computer industry,
as an engineer for Control Data Corporation, to design a calcula-
tor using a programmed approach, blending computer software
technology with desktop calculator hardware. He believed this
would allow changes to be incorporated into the product line more
rapidly. He assigned Shima to this project because of his pro-
gramming experience. This occurred in 1968, about one year after
Shima joined Busicom.
During 1968 and 1969, using a program logic method instead
of hardwired logic, Shima developed the Model 141-PF, a
desktop calculator with printer that Busicom produced and sold
Fig. 1. Masatoshi Shima. successfully for small-business applications. Consciously mim-
Photo courtesy of M. Shima. icking the design of the main block of a computer, Shima incor-
porated in his calculator design read-only memory, entry regis-
ters, accumulation registers, multiplication registers, and two
arithmetic units (adders). He also defined and designed the
macro-instruction set for the calculator, based on what was be-
ing used in decimal computers. All of this work was done, he
claims, in complete isolation from people outside his small de-
velopment group in Osaka.
A new business challenge to Busicom came in late 1968, when
Sharp, with assistance from its U.S. partner Rockwell, introduced
a desktop calculator using large-scale integration (LSI) to squeeze
the 200 or so basic components in a calculator onto only four
chips. The competition was surprised that Sharp could achieve a
design with so few chips.
Busicom’s response was two-fold. It began to develop its own
LSI-based calculator line and seek out U.S. semiconductor com-
panies that could be its partner in the same way that Rockwell
aided Sharp by preparing logic schematics and circuit designs for
chips to be used in the calculators. Busicom had two factories
Fig. 2. A Busicom calculator. manufacturing calculators—Osaka’s Nippon Calculator Machine
Corporation, manufacturing small-business machines in high
Shima was fortunate in that he joined the company when it was volume, and Tokyo’s Electro Technical Industries Corporation,
moving from using discrete transistors to integrated circuits in its manufacturing scientific calculators and specialized office equip-
desktop calculators. He soon discovered that he did not have to ment such as billing machines and teller machines. The company

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 7

The Intel 4004 Microprocessor

sought a basic LSI design that could be used across the entire ready to run as soon as you are. Having something like that
13
range of products at both factories. you hope will get around that cash flow problem.
This technical objective was partly achieved. In the end, fol- In June 1969, Busicom sent three employees to Intel’s offices
lowing the recommendation of a U.S. consulting firm, Busicom in Santa Clara, California, to describe the work they wanted done:
settled on two U.S. semiconductor partners, Mostek and Intel, and project manager Mr. Masuda, senior engineer Mr. Takayama, and
two design strategies. Mostek was contracted to provide the LSI Shima, who was the junior member of the team but who never-
technology for the small-business calculators built in Osaka. Intel theless made the presentation. The contract task Shima outlined
was contracted to provide the integrated circuits for the products was to design a set of chips using LSI technology to run a family
manufactured in the Tokyo factory. Sharp’s small-business calcu- of high-performance programmable calculators. Read-only mem-
lator with four chips had been implemented in metal-gate technol- ory chips were to be added to customize the basic design for each
ogy, so Busicom chose silicon-gate technology, which had the model in the family. The meeting was followed on 16 Sept. 1969
prospect of providing higher-scale integration than Sharp’s ap- by a letter presenting a formal offer of contract.
proach—which was important since Busicom did not want to Intel originally gave responsibility for the Busicom project to
build simple calculators like Sharp was building for the home Ted Hoff (Fig. 3), manager of the Application Research Depart-
market but more powerful calculating machines for professional ment, with assistance from another engineer, Stanley Mazor
use. Mostek and Intel were chosen because they were the two (Fig. 4). Hoff had a strong technical background: an undergradu-
U.S. semiconductor manufacturers that worked with silicon-gate ate degree in electrical engineering from Rensselaer Polytechnic
technology and were not already aligned in business relationships University in 1958, a PhD from Stanford University in adaptive
with other Japanese calculator manufacturers. systems (what we would now call neural networks) in 1962, and
The Osaka factory brought to Mostek the logic and circuit six years’ further experience at Stanford as a researcher, continu-
schematics, and Mostek manufactured the chips to Busicom’s ing his work on adaptive systems. He had joined Intel in 1968 and
specifications. These calculators were first produced using two already completed some important work for the company on
chips, but in 1970 Busicom was able to introduce a desktop cal- metal-oxide semiconductor (MOS) random-access memory before
culator using a single LSI Mostek chip and hardwired random this assignment, which he received because of his systems and
logic. The collaboration with Intel was not as straightforward. applications experience, even though by his own admission he
14
was not a chip designer.
Busicom and Intel: Defining a Product
The history of Intel is well-known and need only be summarized
12
here. Intel was one of the so-called Fairchildren, the companies
started by former employees of Fairchild Semiconductor. It was
incorporated in July 1968, after Noyce and Gordon Moore left
Fairchild to exploit the commercial opportunities of the emerging
LSI technology. Intel’s primary goal was to produce a semicon-
ductor computer memory to replace magnetic cores. The com-
pany’s first big success came in late 1970, when it introduced the
1103, with one-kilobit dynamic random-access memory. By 1972,
this product was replacing core memories in large numbers and
was the most successful semiconductor memory on the market.
Already, Intel was supplying memory chips to 15 of the 18 largest
computer manufacturers.
When Busicom first contacted Intel about developing the chip
set for its new line of computers, Intel was less than a year old.
Despite the fact that Busicom’s project was outside of Intel’s ma-
Fig. 3. Marcian E. “Ted” Hoff, Jr.
jor line of business in semiconductor memories, Intel welcomed
the business, both to build up its client list and to provide it with
some contract work to ease its cash-flow problems while it devel- Despite the value of the collaboration to both companies, the
oped its proprietary memory chips. As Hoff explained: project did not progress smoothly during the first year. The Japa-
I was involved in some of [the contract negotiations with nese team had already gone a long way toward completion of its
Busicom].... Most of Intel’s other products were proprietary design, having the logic schematics 80 percent to 90 percent com-
15
with the potential for fairly long design cycles. When you plete in Shima’s estimation. They expected, not unreasonably,
develop a new memory chip of your own design, you finally that with all of this work completed, the chip set could be pro-
announce it to the world, and then you sit back for a year or duced rapidly. Shima was at first optimistic about the collabora-
two and wait for people to design it [into their products] and tion, but his confidence soon faded.
get their production going.... You are ready to make them by [Hoff] had good abilities in many, many areas, including
the millions and people are just buying them by the ones computers, software, circuit design, logic design, and
and twos to try them out. So the business of trying to get simulation. In the beginning of the meeting[s] we thought
around that is doing custom silicon. When you do custom that once we showed the logic schematic to them, they
silicon, you are working with a design team that you hope is could understand what we wanted to do. But after several

8 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

 meetings, we found out that they didn’t have a logic engi- tage of Intel’s strength in low-cost semiconductor memories to
neer to understand our logic schematics and they didn’t make the same design trade-off in the Busicom design. He be-
have any circuit engineer to convert our logic schematics to lieved this approach had other advantages, only some of which
circuit schematics.... were of importance to Busicom, which restricted its business in-
terests to calculators:
Also they didn’t know about desktop calculators them-
selves. For example, Seiko’s printer was a line printer Reducing complexity of the elemental instructions could
[unlike the slower serial printers they were familiar with also make the resultant processor a more general-purpose
from Monroe calculators].... Therefore, we had to explain machine. In the calculator, a program stored in ROM could
what these desktop calculators were, what the line printers utilize sequences of more general instructions not only for
were, and how to control them. After a couple of months of arithmetic, but for keyboard scan and debounce, display
discussion, we were not able to reach agreement. Also, we maintenance, and other functions as well. With the flexibil-
had to explain the function and control of the keyboard, dis- ity, Hoff thought, a more general-purpose processor might
19
play, and card reader. It was quite a difficult job for us to find applications quite apart from calculators.
explain because they had never seen them. In August 1969, Hoff proposed to Intel management this ap-
proach of building a calculator by making a computer that was
Ted Hoff asked me many questions. I brought some other programmed to act like a calculator. Intel management endorsed
things. For example, ... the flow chart for the desktop cal- 20
the idea, and Hoff began to develop it. Intel management was
culator’s program. Many times I explained the desktop cal- relieved not to have to design a larger chip set, such as Busicom
culator’s program and the macro instruction set. He had a 21
proposed, because it would have taxed Intel’s small staff. In a
lot of questions about this flow chart, then he asked me, later interview, Noyce corroborated and commented on this point:
“why don’t you explain the instruction set?” I explained the
list of instructions for the desktop calculator, showing the And it was a little difficult because we had at the time four
hardware block diagrams and flow charts and program ... or five circuit designers. We didn’t have the spare resources
16 to apply to this project, which was good because then that
based on the macro-level decimal instructions.
led Ted Hoff to the idea of trying to simplify this whole
thing by finding an easier design path to their needs, and
their objectives. It’s the old story: necessity’s the mother of
invention. I can’t do what you’re asking me to do, so I’ve
got to figure out a simpler way to do it, and of course that
was the origin of the idea of having a programmable unit
there, which is really the essence of the microcomputer: do
the design job, have it programmable so that you can use it
in the many, many different applications.

There was discussion prior to that of the number of new de-

signs that you’d have to do everyday in order to realize the
potential of large-scale integration. I recall a discussion
some years earlier at Fairchild, trying to extrapolate and
guess what the world would look like if large-scale inte-
grated circuits came in to effect, and basically what you’re
saying is we would have to design 10 circuits a day. This
was at a time when designing a circuit took a year and the
Fig. 4. Stanley Mazor.
question [was] how you would ever accomplish that. The
answer, of course, is you don’t; you find another way
Hoff, however, had a different perspective. He regarded the 22
around it.
Busicom design as too complicated and too expensive to imple-
ment, if one wanted to sell the calculators at a marketable price. However, Shima was not convinced: “The basic idea from Ted
He had been working with a Digital Equipment Corporation Hoff was quite nice. But to use his idea for my applications, the
PDP-8 computer in his research and was struck by the difference detail was not so good ... lack of system concept, lack of decimal
between the PDP-8’s lean architecture and the relatively compli- operations, lack of interface to the keyboard, lack of real-time
23
cated program logic in the Busicom design. He felt, in the Japa- control, and so on.” Shima had other problems with Hoff’s ap-
nese design, “that hierarchical structure [of using a building block proach as well. It required using the older transistor-transistor
17
over and over] seemed to be lacking.” (In particular, Hoff was logic technology for system interface, whereas Busicom had ap-
concerned that the control logic for the various peripherals was proached Intel in the first place with the hope of avoiding all use
done by separate structures; the shift register memory required of transistor-transistor logics, replacing them with the LSI com-
18
complex timing; and the elemental instructions were intricate. ) ponents that were supposed to be Intel’s expertise. Shima was
The PDP-8 achieved its considerable power by making the design concerned about the great expense for the extra read-only memory
trade-off of employing a spartan instruction set but at the cost of a Hoff’s design required. He was also frustrated that Hoff did not
large program memory. Hoff saw an opportunity to take advan- work out the details of his implementation.

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 9

The Intel 4004 Microprocessor

From August to October 1969, Hoff pursued his approach, ing.... When I went to Intel, I found out they had not done
while Shima worked (also on-site at Intel) to improve and com- anything! ... We fought it quite strongly because Busicom
plete his original design. As Hoff remembers the situation: paid money and Busicom sent all the documents on each
[The Japanese engineers, Shima in particular] had a great LSI product.... And at that time when I went to Intel, noth-
28
reluctance to change the design. I can understand it; they ing was done. I was mad!
had a big investment. They had already done a fair amount Intel management had pulled Hoff from the project to work on
of coding, so it would be very reasonable that he would not another contract, to develop a chip to serve as the central proces-
want to. At a conference a few years later, Shima said, “I sor for an intelligent computer terminal (the Datapoint 2200) for
had to write all my code over again!” The interaction was Computer Terminal Corporation—a project that ultimately re-
that I was going in one direction [and the Japanese in an- sulted in another, but more powerful microprocessor known as the
other]. Basically they already had a design that was pretty 8008. (Shima was not informed of the fact that Hoff had trans-
well along. As we explained what we felt some of the ferred to another project until he arrived in the United States in
problems [with their design] were, they felt that they really April 1970. He took out his anger on the project’s new chip de-
did understand those problems and accepted them ... but signer, Faggin, until he learned that Faggin had joined Intel only
they felt they could stay with the spirit of the original design one week before and did not even know any of the details of the
... to the logic, and doing so they were cutting the number of project yet.)
transistors and the number of chips necessary [thus ad-
dressing Hoff’s concern that the implementation of the
24
Busicom design would be too expensive]. Intel did not regard the Busicom
In October 1969, Busicom executives visited Intel, and the two contract as a high priority and did not
approaches were described to them. Hoff presented a set of four
chips: a two-kilobit read-only memory to store the program in- then see a place for the microprocessor
structions (known as the 4001), a 320-bit random-access memory in Intel’s future.
to store data (the 4002), a 10-bit shift register (the 4003), and a
four-bit central processing unit (the 4004). All of these were on
Hoff’s redeployment was a reasonable management decision
chips with 16- or 18-pin packages, which was the standard in the
for Intel. Hoff was manager of the Applications Research De-
semiconductor industry (unlike the packages that Shima designed
partment and was skilled at the applications aspects of the devel-
with larger numbers of pins, which were technically better but did
opment process but not in the details of chip layout and fabrica-
not conform with industry standards—in fact, Shima did not know
tion. When Hoff was reassigned, the setting of the design in sili-
enough about the semiconductor industry to know that there were
con for Busicom was turned over to Les Vadasz, head of Intel’s
standard packages). The 4004 microprocessor, which was the
MOS design team. However, this created several problems.
most complex of the four chips, as proposed involved integrating
25 Vadasz and his staff were too preoccupied with Intel’s principal
an estimated 1,900 transistors on a chip. Shima had reduced his
business, the design of memory chips, to do much work on the
design somewhat, but it still required 12 chips, averaging more
Busicom chip set; moreover, their experience was in memory chip
than 2,000 transistors apiece, using nonstandard 36- or 40-pin
rather than logic chip design. It was clear they needed to hire
packages. Although these comparisons seem to favor Hoff’s ap-
someone with appropriate experience to carry out the Busicom
proach, Hoff was surprised at the decision: “I always thought it
project, but it took them six months to do so. Partly it was because
was a coup that we managed to persuade the Japanese managers
26 there was a shortage of qualified chip designers available for hire;
to choose our design over [the one of their own engineers].”
apparently it was also partly because Vadasz was reluctant to hire
the most qualified available person, Faggin, because of the fric-
Implementation and Marketing tion between them when they had worked together at Fairchild.
29
Shima accepted the decision gracefully and began to work with Another reason may have been that Intel did not regard the Busi-
Mazor to solve what he regarded as the shortcomings of Hoff’s com contract as a high priority and did not then see a place for the
design. This work was completed by the end of 1969, and Shima microprocessor in Intel’s future.
returned to Japan in December to finalize the programming and Eventually, Intel did hire Faggin (Fig. 5), who proved to be an
27
documentation. Over the next several months, he worked on excellent choice for the next stage of the project. He had received
some of the programming detail, a product manual, product speci- 30
a PhD in physics in 1965 from the University of Padua. He lec-
fication, and some of the hardware architecture. tured in physics for a year at the university before holding posi-
By the end of March 1970, Shima had completed this work and tions in several Italian companies, where he helped develop an
returned to California in April to see how Intel was progressing at early process technology for MOS devices. He came on exchange
turning the architectural design into silicon products, as well as to to Fairchild in the United States in 1968, where he played a major
finalize the last few details in the contract between Intel and Busi- role in developing new silicon-gate technology for fabricating
com. He was distressed to learn that virtually no progress was high-performance, high-density MOS integrated circuits.
being made. This delay at Intel seriously undermined Busicom’s Intel hired Faggin in April 1970 expressly to work on the Busi-
schedule to build its calculator line and jeopardized its market com project, with the understanding that the architecture and logic
competitiveness. design were already completed and only some circuit design and
I visited Intel once again. That time I was told from Busi- the chip layouts remained to be done. However, he found the
com that my main job was just to check what Intel was do- project much less advanced than he had been led to believe: Many

10 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

architectural and logical design issues were unresolved, and he Probably [the general-purpose capability of the microproc-
found the Intel staff to be little help in resolving these matters. essor had] not that much [economic value to Busicom]. ...
With some assistance from Shima and Mazor, Faggin worked The only advantage [in the calculator business] was in a
furiously for the next year on the project, often putting in 12- and smaller number of designs, higher volume and lower cost of
16-hour days. He has indicated that “I wanted so badly to do a production, getting down the learning curve faster. [On the
good job that I almost worked myself to death to meet the sched- other hand] there is no guarantee that the original [Busicom]
31
ule.” He resolved the remaining architectural problems, pre- design would have been any better, and maybe it would
36
pared the final logic and circuit designs, and developed a new have been a lot worse.
process for laying out the four chips. The 4001 (read-only mem-
ory) was completed in October 1970 and the 4002 (random-access TABLE 1
memory) and 4003 (shift register) the following month—with THE 4004 MICROPROCESSOR
very few flaws appearing. There were minor problems with the
The 4004 was a four-bit microprocessor, measuring 0.11 ×
4004 (central processing unit), and the masks were not perfected
0.15 inch and containing 2,300 MOS transistors. It incorpo-
until February 1971. By March, full chip sets were sent to Busi-
rated a four-bit adder for doing additions, an accumulator
com for testing—a set consisting of four 4001s, two 4002s, two
for keeping track of partial sums, and 16 registers for tem-
4003s, and one 4004. The speed at which Faggin was able to
porary storage. The 4004 could perform 60,000 executions a
complete his work and the high quality of the results he produced
second and address 1,280 half-bytes of data and 4K bytes of
are great testaments to his ability and effort.
programmed instruction. Two four-bit numbers could be
added in about 11 millionths of a second.

Gathered from S. Augarten, State of the Art. New Haven,

Conn., and New York: Ticknor and Fields, 1983.

More interesting is the fate of the microprocessor at Intel.

During the development of the chip set, Faggin had felt that he
always had to struggle for resources:
One possible reason for having to fight for those resources
is that there was a lack of understanding at a more senior
management level. Possibly the company just didn’t have
the resources to give. Maybe this project was not considered
37
as important as the memory chips they were working on.
There was a recession in 1970, chip orders had not increased as
rapidly as Intel had expected, and Intel was strapped for cash. The
company laid off workers and delayed moving into its new head-
Fig. 5. Federico Faggin.
quarters building. Memory, rather than logic chips, remained the
company’s main interest. In any event, under the terms of the
Shima remained at Intel from April until November 1970. contract, Intel could manufacture only the chip set, including the
During this time, he worked on a detailed logic schematic that he 4004 microprocessor, for Busicom’s use; there was no opportunity
believes to have been instrumental in the chip layout. He also at this time for Intel to build a business manufacturing and selling
worked on logic simulation and on a test program, which was microprocessors.
32
used to verify that the set worked properly. Between April and The situation changed rapidly, however. Other Japanese manu-
October, Shima’s colleague, Masuda, developed a breadboard of facturers were producing calculators in high volume, and Busi-
the calculator based on the logic schematic that Shima had pre- com felt that it had to lower its calculators’ prices in order to re-
pared while at Intel. In December 1970 and January 1971, Shima main competitive. As early as April 1971, just two months after
used this breadboard to build a prototype of the printer desktop Faggin had produced a complete working chip set, Busicom man-
calculator. In March 1971, he was able to rebuild the prototype agement approached Intel with a request to renegotiate the con-
into a fully functional model using a 4004 chip made at Intel (see tract. Faggin recommended to Intel President Noyce that Intel
33
Table 1). renegotiate, giving price concessions in exchange for abrogation
Starting in April, Busicom began the manufacture of calcula- of the exclusivity clause. Hoff made a similar request:
tors, billing machines, cash registers, and teller machines based on
the new technology. Most of these machines were manufactured I talked to marketing people who were going to Japan. ...
on an original equipment manufacturer basis, with NCR as the Our marketing people did not seem terribly enthused [when
34
main customer. Sales reached 100,000 units. Busicom made I asked them to negotiate for those rights] but they did in-
some use of the flexibility of the microprocessor approach by deed negotiate for those rights, and they came back in May
38
using microprogramming to introduce new features into several of with the right[s].
the calculator models—a flexibility that customers apparently These rights came in two stages, each with a price concession
35
appreciated. But Hoff has argued that the microprocessor design to Busicom. At first, Intel had limited rights to sell the chip set as
did not have major importance to Busicom: long as it was not sold to other calculator manufacturers, and later

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 11

The Intel 4004 Microprocessor

42
Intel gained universal rights to sell the chip set, which it called the reluctant to make a decision.” Noyce and Hoff indicate that this
39
MCS-4 (short for Microcomputer System four-bit). uncertainty was paralleled in Intel’s board of directors:
Faggin disputes Hoff’s role: There was debate, though, as to whether Intel should exer-
I started pushing management to go into the open market cise its new option. Some members of the board of directors
with that chip. Now, Ted Hoff is saying he was the one do- were not certain the company should venture into the sys-
ing it, but actually he was against it. He was telling every- tems business. In the end, the view of Arthur Rock, chair-
body that it was only good for calculators, so I took it upon man of the board, prevailed and the board endorsed the
43
myself to show management that I could actually solve venture.
other mechanical problems using the 4004. ... When it was
time to develop the production tester for the 4004, I used the
4004 as a controller ... and used that to say, “now tell me
that it is only good for calculators.” I built a lot of momen-
40
tum inside the company to actually go market the 4004.
Although Intel had gained the right to sell the microprocessor
(Figs. 6 and 7), the company did not do much to exercise that
right. At first, the company did not seem to recognize the poten-
tial for the microprocessor. Marketing viewed the microprocessor
narrowly, as only a direct competitor to the minicomputer.
But nothing was done [to follow up the negotiated rights to
market the MCS-4], and there was a lot of negative feeling
within Intel. Primarily in marketing. One marketing guy Fig. 7. The Intel 4004 in its package.
said, “the total sales of minicomputers is 20,000. We are Photo courtesy of Intel.
latecomers to the business, so we will be lucky to get 10%
[of the business]. 2,000 computers is not worth all this.”
The change took place at about the time when Intel hired a new
The other position was that, even if we announce [the prod-
marketing director, Ed Gelbach. He had previously been at Texas
uct], computers take a lot of [staff support] and we have no
Instruments, which had shown an interest in logic devices. In fact,
way [to do this]. Some of these strategies were discussed.
work by Gary Boone and others on a microprocessor was already
[The argument was given that] we can hand-hold a few key
under way at Texas Instruments when Gelbach joined Intel. Gel-
customers, and the rest of them will be on their own—but a
bach swept away the concerns formerly expressed by marketing
lot of people who buy minicomputers are on their own.
and helped push Intel into the microprocessor business. One of
They get a manual, and we could put together a manual ...
the most compelling arguments inside the company at that time
with examples how to program it. But basically [customers]
41 was that customers would need memory devices to attach to the
were [to be] on [their] own.
microprocessor, so the microprocessor was regarded as a way of
enhancing the market for Intel’s primary business interest.
In the 15 Nov. 1971 issue of the trade journal Electronic News,
Intel ran a large advertisement, “Announcing a new era of inte-
grated electronics.” At the Fall Joint Computer Conference held in
Las Vegas, Nevada, that same month, Intel staffed a suite where it
distributed information about the MCS-4. “Customers came in
and really inundated them, [wanting to know] what is this com-
44
puter on a chip thing.” Intel received 5,000 responses to the
Electronic News announcement, far greater response than to any
previous product announcement. In 1971, Hoff and Mazor went
on the road in the United States and Faggin in Europe, giving
talks about the MCS-4. Hoff remembers standing-room-only
crowds and on many occasions having to relocate these meetings
to larger lecture rooms.
Certainly by the end of 1971, Intel marketing and senior man-
agement recognized the value of the new product. By late 1972,
the technical press was taking notice, and companies such as Na-
tional Semiconductor and Rockwell were beginning to build their
own microprocessors. By the mid-1970s, the business media were
Fig. 6. The Intel 4004 microprocessor.
paying attention to the microprocessor, led by a major article in
45
1975 in Fortune. By the end of the decade, public interest had
The senior management was no more certain than the market- been captured.
ing department of the value of the new product. Hoff remembers Faggin has pointed out that marketing a microprocessor is
having a meeting with Noyce and others. Hoff said, “They had a more difficult than marketing conventional electronic compo-
tiger by the tail and did not know how to handle it. They were just nents. In order to market the microprocessor, Intel had to produce

12 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

not only the data sheets giving the product’s specifications but the application, the funding to carry out the development, and the
also “a programming manual, application notes showing how to stimulus to Intel to complete the project after the funding had
use the components, a development board capable of implement- been supplied. The behind-the-scenes role of Sharp as technologi-
ing a functional prototype of the hardware, and a cross-assembler cal entrepreneur—a fact that was completely unknown to the
(i.e., a program running on a minicomputer that enabled the con- other principles—should also be remembered. (In fact, Sasaki also
46
version from instruction mnemonics into machine language).” provided start-up funding to Faggin when he founded the rival
50
Most difficult of all, one had to teach applications engineers to microprocessor firm Zilog.)
change their approach to design, to seek software rather than
hardware solutions. For a while, Intel spent more on printing and
mailing manuals (trying to interest applications engineers) than it For a while, Intel spent more on printing
47
gained in revenue from microprocessor sales. The company was and mailing manuals (trying to interest
able to turn this marketing difficulty into a business opportunity,
however, by developing design tools, such as prototyping boards
applications engineers) than it gained
and software assembler and simulators, which it sold to applica- in revenue from microprocessor sales.
tions developers. In the first few years, these design tools were
more profitable than the sale of the microprocessors themselves. A second piece of lore that merits historical reinterpretation is
The difficulty of teaching application engineers to program Intel’s foresight in developing the microprocessor. The evidence
rather than hardwire their logic functions was compounded by the indicates that the Busicom project was attractive to Intel not be-
limited capabilities of the 4004 microprocessor. The application to cause of the important new technology the project would develop,
the calculator was obvious, but for what else was the 4004 appro- but primarily because it was a cash cow that would enable the
priate? The answer, Intel decided, was embedded control, and company to continue its operations while it pursued its principal
Intel set about to create a market, concentrating first in business interest in semiconductor computer memories. Even after the
machines, aviation and medical applications, and test equip- MCS-4 chip set had been successfully completed, the company
48
ment. For example, the 4004 was placed in test equipment so did not at first fully recognize the potential of microprocessors.
that engineers would not burn out their equipment by setting the Marketing reaction ranged from unenthusiastic to hostile, senior
dials incorrectly. Intel also saw opportunities for using microproc- management was indecisive in promoting the technology, and the
essors as controllers in durable consumer goods such as washing technical people had only an inkling of the range of applications
machines and microwave ovens. The company solicited applica- for which microprocessors might be used. It took no more than a
tions suggestions from its customers, including one who told them year for these initial positions to change, but it was only with its
of attaching microprocessors to cows to monitor water and salt third microprocessor project, the 8080, that Intel made sufficient
49
consumption and milk production. commitment to this field to invest its own money in development.
A third piece of common lore is to assign the invention of the
Conclusions microprocessor solely to Hoff. This is intended in no way to dis-
Intel has had an impressive history as both technical innovator parage Hoff, a truly remarkable engineer. But it must be remem-
and profitable business, and it has every right to be proud of its bered that Hoff’s contribution was primarily to have one good
role in the development of the microprocessor. However, several idea about how to reconceptualize an existing project, making
claims that seem to have become accepted lore in the engineering economical use of the microprocessor, and then to work for a
community do not hold up under historical scrutiny. One is that short period of time to flesh out the architectural design—without
Intel invented the microprocessor. If one means by this statement completing it. Much of the hard work in logic design, chip layout
that the company was the first to conceive of the microprocessor and fabrication, and customization to application were accom-
and introduce the concept to industry, then it is quite clear that this plished by Shima and Faggin with little, if any, of Hoff’s assis-
claim is false. The stored-program computer, which was used tance. Faggin remembers that Hoff seemed to have lost interest in
prominently as a calculator in the 1950s and 1960s, was one of the Busicom project after he moved on to the 8008 project and
the most lauded achievements of the postwar era and was familiar that there was little contact with Hoff as Faggin worked feverishly
to every engineer working in the semiconductor industry. That to make the chip set a reality.
industry itself had the computer as its most prominent application There may be many reasons why Hoff has been given so much
area. With the scale of integration on chips doubling every year, credit for the microprocessor. People need to have heroes and
by the 1960s the technology was reaching a point where it was simple stories to tell and remember, and a sole inventive genius
technically feasible to build a central processing unit on a chip. conforms to this need. Western culture appreciates scientific feats
We have seen, not surprisingly, that a number of different groups more than engineering ones, so it is not surprising that greater
independently came to Hoff’s concept of a computer on a chip. credit is awarded to conceptualization than to implementation. It
It is clear that what was important in the “invention” of the mi- has also been suggested that Intel’s powerful public relations ma-
croprocessor was not the conceptualization but rather the imple- chine has done little to disabuse the notion that people who re-
mentation of an economically sound product. On these grounds, main close to the company (Hoff) deserve more credit than others
Intel has a stronger claim. It was the first company to successfully who moved on to rival firms (Faggin)—although it is beyond the
51
build and commercialize a microprocessor product. It recognized scope of this paper to evaluate this claim.
an application where it made both technical and economic sense It is difficult to apportion credit fairly to the other principal
to use a microprocessor, and it carried forward the project suc- participants in this story. This has been a contentious issue for
cessfully. But it must also be remembered that Busicom provided many years, and Faggin, Hoff, and Shima have decidedly dif-

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 13

The Intel 4004 Microprocessor

52
ferent viewpoints about this matter. A lot of the controversy Hoff, especially when Shima’s English language skills were weak.
may be explained by the different interests and backgrounds of Gelbach seemed to have turned around the attitude toward micro-
the actors, which led them to have different sets of priorities and processors in the marketing department and to have instilled some
values. of the interest Texas Instruments had in logic devices into the
It is clear that Faggin rescued the Busicom project and worked memory-oriented Intel culture. The contribution of Noyce as
long and hard to make it a success. He seems to have made major president remains unclear, but it is apparent that he played an
contributions to the architectural, logical, and circuit designs and important background role throughout this story. In Japan, Mu-
to have been almost single-handedly responsible for the chip lay- rakami and Sasaki were important background figures. The story
out and fabrication. Over the next several years, Faggin managed is most appropriately told as one of many individuals, Japanese
Intel’s growing microprocessor business, which amounted to ap- and American, working sometimes together and other times at
proximately 30 percent of the company’s sales by 1974. He was cross-purposes, to socially construct the microprocessor.
the cofounder and chief executive officer of Zilog in 1974, which
was a very important manufacturer of semiconductor devices for Acknowledgments
the early personal computer industry and a formidable competitor This paper is an abbreviated version of “The Social Construction
to Intel for a while. His entrepreneurial, managerial, and technical of the Microprocessor: A Japanese and American Story,” in A.
abilities in semiconductors have continued to show in his current Goldstein and W. Aspray, eds., Facets: New Perspectives on the
start-up firm, Synaptics. History of Semiconductors (New Brunswick, N.J.: IEEE Center
It is much more difficult to evaluate Shima’s contributions. for the History of Electrical Engineering, forthcoming). In the
Shima has written his personal account of his role in the devel- absence of good archival sources (especially in English), this
opment of the microprocessor, in which he assigns himself a cen- article was written largely from published sources and oral history
tral position in the development of the 4004, 8080, Z80, and interviews. Depositions in relevant court cases, such as Texas
Z8000 microprocessors. However, in interviews, Sasaki, Faggin, Instruments v Zenith (No. 3-89-2960-R), were also not available
and Hoff have strongly disputed Shima’s claims. for examination. I appreciate the helpful comments received from
Shima correctly claims that he was the designer of the original Ross Bassett, Thomas Misa, and Yuzo Takahashi on preliminary
Busicom plan, important elements of which were incorporated drafts of this paper. Rachel Stewart and other staff of the Intel
into the chip set Intel built. Shima also correctly claims that Intel Museum Archives Collection were extremely cooperative in
did not have organizational capability in logic chip design (at least finding archival sources.
until Faggin was hired) and that he provided this skill to the proj-
ect at a critical time, even if his capability was of a self-taught References
variety. Shima was the principal conduit for the knowledge of the [1] See, for example, D. Clark, “High-Stakes War Over Chip Patents,”
application, which shaped many aspects of the design. He was an San Francisco Chronicle, 8 Sept. 1990, pp. b1-b3; M. Antonof,
“Gilbert Who?” Popular Science, Feb. 1991, pp. 70-73; and J. Gem-
important catalyst in persuading Intel to follow through on its perlein and P. Carey, “If Hyatt Didn’t Invent the Microprocessor,
contractual commitment to Intel, and later he played an important Who Did?” San Jose Mercury News, 2 Dec. 1990, p. 27.
role in designing the chip set into a line of Busicom products— [2] The standard historical works on the history of semiconductors pro-
even though they were ultimately not enough to prevent Busicom vided limited information about the history of the semiconductor: E.
Braun and S. MacDonald, Revolution in Miniature, 2nd ed. New
from going out of business.
York: Cambridge Univ. Press, 1982; P.R. Morris, A History of the
Shima was much in demand. Intel courted him to work on the World Semiconductor Industry. London: Peter Peregrinus, 1990; H.
redesign of the clunky 8008 microprocessor into the much more Quiesser, The Conquest of the Microchip. Cambridge, Mass., and
successful 8080 microprocessor. Faggin hired Shima away from London: Harvard Univ. Press, 1990. Also see Intel’s official history,
Intel to work for Zilog, where he played a useful role in the de- A Revolution in Progress: A History of Intel to Date (Intel, 1984),
written by M. Real, Oral History Associates. Useful biographical
velopment of the highly successful Z80 and Z8000 microproces- sketches are given in “The 30th Anniversary of the Integrated Cir-
53
sors. Later, when he decided to return to his native country, Intel cuit: Thirty Who Made a Difference,” Electronic Eng. Times, Sept.
accommodated him by creating a new laboratory in Japan. 1988, pp. 14-24.
However, there is another side to the Shima story. Hoff and [3] See the differing perspectives given by the principal participants:
Noyce and Hoff, 1981; F. Faggin, “The Birth of the Microproces-
Faggin credit Shima with being an excellent detail man, but not
sor,” Byte, Mar. 1992, pp. 145-150, especially p. 146; M. Shima, The
with being someone who made any essential contributions to the Birth of the Microcomputer: My Recollections. Tokyo: Iwanami
conception or design. Faggin referred to Shima as a Shoten, 1987 (in Japanese); Shima, “The Birth and Growth of the
“supertechnician” with machine-like virtues: Microcomputer,” Nikkei Electronics Book, Electronics Innovation,
pp. 159-185 (in Japanese); and Shima, “History and Future of the
Shima is almost like a computer in the sense that he does Microprocessor,” Johoshori, vol. 34, no. 2, pp. 135-141, Feb. 1993.
not make mistakes. He can teach you talents. You can al- Also see T.S. Perry, “Careers/Profile: Marcian E. Hoff,” IEEE
most always rely on him when he says, “Yes, I am fin- Spectrum, pp. 46-49, Feb. 1994.
[4] Faggin oral history interview conducted by the author, 1 Sept. 1994,
ished.” You will find no mistakes in that work. He was an
IEEE archives.
incredibly strong right hand. With Shima I would say to do [5] M. Hoff oral history interview conducted by the author, 1 Sept.
it this way and then it would be. I would say, “check my 1994, IEEE archives.
work; here is the way I want the logic to be done. Clear the [6] Faggin, 1992, p. 146. Some of the references in the published litera-
54 ture to similar ideas include E.A. Sack, R.C. Lyman, and G.Y.
tables, do this, do that.” And I would rely on him.
Chang, “Evolution of the Concept of a Computer on a Slice,” Proc.
Other people also had a role in the development of the micro- IEEE, Dec. 1964, pp. 1,713-1,720; E. Fubini and M. Smith,
processor. Shima credits Mazor as an important go-between to “Limitations in Solid State Technology,” IEEE Spectrum, May 1967,

14 ∑ IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997

 pp. 55-59; Editors, “Integrator on a Chip,” Electronics, 22 Aug. 167-168, 14 Nov. 1974; L.J. Mandell, “Pitfalls to Avoid in Applying
1967, pp. 38, 40. µP’s,” EDN, pp. 22-26, 20 Jan. 1975.
[7] R.N. Noyce and M.E. Hoff, “A History of Microprocessor Devel- [50] Faggin interview; Sasaki interview.
opment at Intel,” IEEE Micro, vol. 1, pp. 8-21, especially p. 8, Feb. [51] Faggin interview.
1981. [52] See the letter to the editor by Faggin’s wife, Elvia, in the 3 Oct. 1986
[8] Noyce and Hoff, 1981, p. 8. edition of the San Jose Mercury News and Hoff’s reply printed in the
[9] The information in this section is taken from a resume provided by newspaper nine days later.
Sasaki and from an oral history interview conducted by the author in [53] On the design process for the Z8000, see M. Shima, “Demystifying
Tokyo on 24 May 1994 (IEEE archives). Microprocessor Design,” IEEE Spectrum, July 1979, pp. 22-30. Re-
[10] Sasaki interview. printed as “Design Case History: Z8000 Microprocessor,” Design
[11] For background on the Japanese industry, see T. Sasaki, “The Role Studies, vol. 2, pp. 97-106, Apr. 1981,.
of Government in the Formative Stage of the Japanese Electronics [54] Faggin interview.
Industry,” lecture given in May 1986 at the Electronics Show in To-
kyo, printed in Summary of Speech[es] in 1986/1987, Sharp Corpo-
ration, especially p. 104; Y. Takahashi, “Progress in the Electronic William Aspray is executive director of the
Components Industry in Japan After World War II,” pp. 37-53, W. Computing Research Association. His ear-
Aspray, ed., Technological Competitiveness. New York: IEEE Press,
lier positions included faculty posts at Wil-
1993.
[12] For a history of Intel, see Real, 1984; G. Bylinsky, “How Intel Won liams College and Harvard University, asso-
Its Bet on Memory Chips,” Fortune, pp. 142-147, 184, Nov. 1973. ciate director of the Charles Babbage Insti-
[13] Hoff interview. tute, and director of the IEEE Center for the
[14] Noyce and Hoff, 1981, p. 9; Hoff interview. History of Electrical Engineering. His his-
[15] Shima interview.
[16] Shima interview. torical writings on computing include John
[17] Hoff interview. von Neumann and the Origins of Modern
[18] Noyce and Hoff, 1981, pp. 9-10. Computing (MIT Press, 1990) and (with Martin Campbell-Kelly)
[19] Noyce and Hoff, 1981, p. 10. Computer: A History of the Information Machine (Basic Books,
[20] Noyce and Hoff, 1981, p. 10.
1996). He is presently writing a history of the academic discipline
[21] G. Bylinsky, “Here Comes the Second Computer Revolution,” For-
tune, vol. 92, no. 5, pp. 134-138, 182, Nov. 1975. of computer science and engineering in the United States to 1980.
[22] Noyce interview, pp. 38-39.
[23] Shima interview. The author can be contacted at
[24] Hoff interview. Computing Research Association
[25] The best technical review of this microprocessor is given in F. Fag-
gin, M. Shima, M.E. Hoff, Jr., H. Feeney, and S. Mazor, “The MCS- 1875 Connecticut Avenue NW, Suite 718
4—an LSI Micro Computer System,” IEEE ’72 Region Six Conf. Washington, DC 20009, U.S.A.
Proc., pp. 1-6. The bibliography in their paper lists the major techni- (202) 234-2111
cal presentations. e-mail: aspray@cra.org
[26] Hoff interview.
[27] Shima interview.
[28] Shima interview. Faggin, 1992, p. 146, corroborates this story.
[29] Hoff interview.
[30] Most of this biographical information is taken from Faggin, 1992,
and the Faggin interview.
[31] Faggin interview.
[32] Shima, personal communication, 10 Dec. 1994.
[33] Shima, personal communication, 10 Dec. 1994.
[34] Real, 1984, p. 12.
[35] Shima interview.
[36] Hoff interview.
[37] Faggin interview.
[38] Hoff interview.
[39] Private discussion with Hoff.
[40] Faggin interview. The author has no way of reconciling these two
conflicting accounts.
[41] Hoff interview.
[42] Hoff interview.
[43] Noyce and Hoff, 1981, p. 13.
[44] Hoff interview.
[45] Bylinsky, 1975.
[46] Faggin, 1992, p. 148.
[47] Real, 1984, p. 13.
[48] Intel, The MCS-4 Story, film, circa 1974 (Intel archives).
[49] Hoff interview. The product literature in the Intel archives gives
some indication of the wide applications of the 4004, including fan-
ciful uses such as described in “Microprocessor Computes and
Checks Life Cycle Tendencies in Coin-Operated ‘Biorhythm’ Ma-
chine,” 20 Oct. 1975, p. 5. Technical publications regularly de-
scribed how to use the 4004, e.g., R.H. Cushman, “What Can You
Do With a Microprocessor?” EDN, pp. 42-47, 20 Mar. 1974; Cush-
man, “How to Get Acquainted With a µP”, EDN, pp. 46-51, 20 Sept.
1974; B. Cole, “4-Bit Controller System Upgraded,” Electronics, pp.

IEEE Annals of the History of Computing, Vol. 19, No. 3, 1997 ∑ 15