Chapter 2 - Background
2.18 The Transistor - 1947
The transistor was the first of three technological discontinuities to radically alter the dynamics and structure of the computer market-structure.56 Everyone knew an alternative had to be found to vacuum tubes if computers were to be made more reliable, faster, smaller, and consume less power and generate less heat. Driving and funding the search for an alternative was the military.
A transistor is a device that accomplishes the two necessary tasks of switching on-offand amplification. Switching on-off made possible the conversion of an analog signal, electricity, to the 0’s and 1’s of a digital signal. Amplification had two purposes: to let a very small current control a very large one, and to boost signals to overcome circuit noise so signals with information could be detected. And profoundly different than a vacuum tube, a transistor worked not by electrons flowing through voltage gradient, but by channeling them in semiconductor materials.57
But investigators lacked pure enough materials to create semiconductors. Once again the government played a major role in pushing theory into practice. During World War II, the government significantly increased funding of semiconductor research at Bell Labs, universities and industrial companies, and created MIT Radiation Laboratory to coordinate the research.58 The war time scientific culture also encouraged communication and cooperation among all researchers – an example was a “Meeting on Germanium Crystals” held at Bell Labs on April 9, 1945 among university and industrial laboratory scientists.59 Three months later, Bell Labs issued an “Authorization for Work” to investigate making transistors from solid-state materials.
It would take over two years, but on December 23, 1947, the first transistor was demonstrated at Bell Telephone Laboratories. Walter H. Brattain and John Bardeen, experimentalist and theorist respectively, demonstrated a crude, but working, amplifying transistor made from germanium and wires. Their demonstration motivated William B. Shockley to work out the seminal principle of a solid-state transistor over the following five weeks.60 (For their achievement, all three received the Nobel Prize.) Public announcement was delayed, however, until later in 1948.
From public announcement onward, Bell Labs, and thus AT&T, consistently acted to effect wide disclosure and use of transistor technology. For example, the military was told only one week before the public announcement to prevent them from blocking its release on the grounds of national security; they held seminars in the early 1950’s when they shared all they knew about transistor technology; and in 1952 they licensed the technology to all comers for a minimum royalty of 5% of sales. Those analyzing Bell Lab’s important open policy conclude Bell Labs knew the transistor was too important to keep to themselves; knew they had more to gain than lose from wide-spread use; might have suspected that the government would have forced such a policy in any case, especially in light of the Justice Department’s antitrust case filed in 1949; and such a policy was consistent with their past practice.61 In AT&T’s case, the fact that it was a monopoly mattered, based on historical practice, but more importantly, it was because AT&T executives truly saw themselves as a service business. This issue will be extensively explored in the next chapter.
If the transistor had been perfected by an organization committed to appropriating every advantage from its innovation, not in making the innovation “public property,” then the subsequent growth in the semiconductor, and all related, including computer industries, would certainly have been very different.
The transistor was not the only possible alternative to the vacuum tube; and again, the government funded competitive investigations. Three of the most promising alternatives were thyratribs (hot filament gas tubes), tronodes (miniature bi-stable neon gas tubes) and magnetic amplifiers (iron-cored or ferrite-cored inductors). Magnetic amplifiers were pursued most aggressively by Remington Rand and subsequently Sperry Rand, although also by IBM, Burroughs, Raytheon and Logistics Research Corporation. In a press release in 1956, Sperry Rand hailed their perfection of high-speed magnetic amplifiers, brand named FERRACTORS, and stated: “The computer opens up an era in which filament tubes and transistors will be outmoded by devices of this kind.”62 The computers introduced by Sperry Rand in 1960 (UNIVAC SS80/90) and 1961 (UNIVAC SS80/90 II) were made with magnetic amplifiers. Not until December 1961 did Sperry Rand introduce a transistorized computer (UNIVAC 490), two years after IBM had introduced theirs (IBM 7070). Sperry Rand’s protracted commitment to magnetic amplifiers proved to be a failed strategy. In their case, it contributed to their continuing inability to re-establish market leadership.
By 1952, Western Electric (and a few other firms) manufactured approximately 90,000 point-contact transistors, which were sold primarily to the military.63 Data from 1955 to 1960 clearly shows the importance of government purchases. See Table below. Two important sources of demand were the early commitment of the Air Force to use semiconductors in the Minuteman Missile in 1958, and the growth of IBM – the largest customer of every semiconductor company. (Note: Table is all devices not just transistors where the percent government purchases was much higher.)
Exhibit 2.22 Government Purchases of Semiconductor Devices 1955-1960
|Year||Total semiconductor shipments ($ millions)||Shipments to federal government ($ millions)||Government share of total shipments (%)|
Source: Richard R. Nelson, Government and Technical Progress: A Cross-Industry Analysis (Pergamon Press, 1982), 60.
In addition to manufacturing transistors for sale, AT&T’s Bell Labs built the first transistorized computer, the TRADIC, finished in January 1954 and funded by the Air Force. Bell Labs disclosed TRADIC in a paper presented at a conference in December 1954.
How did IBM master transistors? The answer, predictably enough – with the help of the government, but with a twist.
In September 1955, IBM lost out to Remington Rand in the bid to build a super fast computer – intended to be 100 times faster than the UNIVAC I – for the University of California Radiation Laboratory (UCRL) [now the Lawrence Livermore National Laboratory]. By then Remington Rand was a division of Sperry Rand.64 Code named LARC, Sperry Rand expected to deliver the computer in February 1958. (It would take until 1960, and $19 million, not the estimated $2.85 million.65 A key mistake was using magnetic amplifiers – they had to redesign again using transistors.)
If LARC was to be the next, newest, fastest, state-of-the-art computer architecture, an architecture based on a component other than vacuum tubes – how could IBM stay competitive if it also did not invest to learn and master the required technological innovations? What would be the best computer architecture using transistors? And how well would it work?
IBM set out to sell a computer design to be delivered under contract to a competitive government laboratory, one doing the computationally intensive work of nuclear research, Los Alamos Scientific Laboratory, now known as Los Alamos National Laboratory (LANL) in New Mexico. A year later, in November 1956, LANL accepted their proposal. The project would be named STRETCH, meaning to extend the boundaries of state-of-the-art computing. It was an opportunity to “rethink” and “redesign” a computer’s architecture using transistors, without needing to be compatible with their existing architectures. The objective: a computer 100 times more powerful66 than the 704.
Meanwhile, the first operational transistor computer, the Burroughs Atlas Mod 1-J1 Guidance Computer built for the Air Force, had been delivered in April 1955, although not operational until September 1957. The first available commercial transistor computer was the General Electric 210 in June 1959. IBM announced theirs, the 7070, in September 1958; RCA, theirs, the 501, in December 1958.
The market-structure of First Generation mainframe computers (1950-1959) consisted of only seven companies and 31 computer models. Other companies developed computers, but they did not sell them commercially. Research and development funding came almost entirely from the U. S. Government. Over 100 computers were “completed on an experimental or contract basis for special uses.”67 Although a commercial computer market existed, it was far from clear what its economic potential might be.
The transistor, as a technological discontinuity, as the economist Joesph Schumpeter might describe it68 , would strike “not at the margins of the profits and the outputs of the existing firms, but at their foundations and their very lives.”69 In the case of transistors, once firms started making computers with transistors, they never again used vacuum tubes.
The assumption of three technological discontinuities is not an assertion that there were only three discontinuities, only that for the purposes of the present argument, three are sufficient to explain the central economic dynamics of computers.
See Ibid. for an excellent discussion of semiconductor technologies and developments.
Ibid., p. 199
Crystals were critical to both radar and computers. To computers they became the clock, the system signal that synchronizes actions.
T. R. Reid, “The Chip,” Simon and Schuster 1984, p. 43-53
Nelson, p. 195
Ibid., p. 59
The merger created internal organizational conditions that would cause it to be both slow and wrong in its actions. Fortune article citation.
Ibid., p. 188
Ibid., p. 189
Only Schumpeter saw”a decisive cost or quality advantage,”not technology as cause.
Schumpeter, 1942: 84