Engines That Move Markets (2nd Ed)

by Alasdair Nairn

  The negotiations between ECC and the Census Bureau resulted in a contract with NBS which provided for payment of $300,000, less 15% for NBS. As Mauchly and Eckert had budgeted for expenses of $400,000, even from the outset the project was scheduled to lose approximately $125,000 (over $1.5m). This was not necessarily commercial suicide, as the cost figures included the research and development costs which could be amortised over future sales of UNIVACs. The real key remained perceptions. Potential customers had to be confident of the solvency of the company and its ability to deliver the specified product on a timely and financially sound basis. This made it vital that the initial cost estimates were accurate and contained a contingency reserve for cost overruns. Unfortunately, scepticism over the prospects for the UNIVAC meant ECC had little bargaining power. It had to settle for whatever it could get. Perhaps ECC could have negotiated a cost-plus contract, but since these contracts typically involved ceding control of any patents they would effectively have reduced ECC to a subcontractor in the work it was itself creating.

  Scepticism about the UNIVAC stemmed from the same source that Mauchly and Eckert had encountered in their work on ENIAC. The head of NBS referred the project to the National Research Council for an opinion. It was not supportive. This might have been because the group giving the opinion was itself involved in competitive work or had good reasons to delay the project.⁸² While the NBS remained committed to the project, the report of the NRC was damaging. The net result was that the project began life undercapitalised and with insufficient cash flow to sustain it. As a consequence, the company struggled continually to find ways of obtaining funds to continue the development process. Eventually the principals lost control of their company.

  Mauchly’s first step was to agree to build a small computer to operate a guidance system for a new military aircraft being built by Northrop. While the contract for this computer, the BINAC, brought in additional funds, it also distracted attention from the hard-pressed team at ECC. Worse still, the final cost of completing the contract was over two and a half times the $100,000 in revenue it generated. If this was not bad enough, as the focus of the company then shifted back to the UNIVAC, there was no follow-up on the BINAC and therefore no sales of similar machines to offset the development cost. The BINAC project was not totally without value as it did generate publicity. However, the publicity was limited and restricted to short commentaries on the machine, rather than highlighting its commercial prospects. As a consequence, BINAC failed to deliver either visibility or cashflow.

  The funding need remained pressing and a degree of salvation for the company, now named the Eckert-Mauchly Computing Corporation (EMCC), came with the injection of $0.5m by the American Totalisator Company, a group that exercised an effective monopoly on racecourse betting and required the computational power for obvious reasons. It took a 40% equity stake in return. The additional funds sustained EMCC for a period; further funds were received in the form of advance payments from new customers, including A. C. Neilsen, the market research company, and Prudential Insurance. The financial situation remained stretched, though, as research and development consumed capital. The desperation for funds was obvious to new customers, who demanded further fixed-price contracts on less than advantageous terms. The firm remained precariously in operation until the tragic death in an aircraft explosion of their principal supporter at the American Totalisator Company. At this point, the principal equity shareholders demanded either a refinancing from external sources or a sale of the business.

  After repeated efforts to raise new funding and in increasing desperation, Mauchly and Eckert sought to find a buyer for the business. A meeting with Thomas Watson at IBM failed to provide a home, partly due to IBM’s concern over the antitrust implications, and partly because of their belief that they could develop the resources internally. Although both NCR and Remington Rand showed an interest, the latter moved the quickest. The administrative head at Remington Rand, General Leslie Groves, had been aware of the role of ENIAC in the atomic project during the war and was thus familiar with the work of the two scientists. In 1950 EMCC was acquired for a payment of not much more than $500,000 ($5m) and with it the patent rights of the work that had been completed, as well as any future patents which would be forthcoming. The new owners then sought as far as was possible to renegotiate the outstanding contracts for UNIVAC machines. In March 1951 the UNIVAC was finally completed and passed its acceptance tests for the Census Bureau.

  Success for the UNIVAC

  The UNIVAC proved a great success, undoubtedly helped by the publicity it received in correctly predicting a landslide Eisenhower victory at the 1952 presidential election. In 1952 Remington Rand also purchased the Engineering Research Associates Company (ERA) to advance its computer capability. ERA had been formed after the war by William Norris and Howard Engstrom, two naval engineers, to manufacture electronic cryptology machines and had moved into general purpose computing equipment. Remington Rand therefore found itself in a position where it had the leading technology in a market that was beginning to expand rapidly.

  Competition certainly existed. IBM, for one, recognised its vulnerability to the new technology and moved a large part of its R&D budget to combat the problem. Despite this, its rival to the UNIVAC was not available until four years after the first UNIVAC had been delivered. The UNIVAC retained its technological lead. IBM’s sales of its 701 and 702 machines survived largely because the unanticipated speed of market growth left Remington Rand unable to meet demand for the UNIVAC. The IBM machines were a stopgap measure to prevent UNIVAC dominance. Further models were quickly brought to the market and IBM began to close the technological gap.

  The growing market attracted many other potential competitors. These included existing companies working in related fields, such as RCA, GE and Honeywell, together with spin-offs from defence companies including Computer Research Company from Northrop and Datamatic, a joint venture between Honeywell and Raytheon. The growth also caused the creation of a number of companies started by employees within larger groups who then split off to start their own operations.

  The growing threat from IBM was not lost on Remington Rand, and it merged with Sperry in 1955. The problem for the merged company was that its competitive position was at risk from more than just technology. Sperry-Rand had not moved to integrate the sales and marketing of its ‘old’ punch-card technology equipment with the ‘new’ UNIVAC machines. It also lacked sufficient capital to fund all the various operations that it had acquired, and a consistent method of prioritising needs. Overlaid on top of this were serious cultural differences between the merged companies. In 1957 Norris left to form Control Data and in 1959 Mauchly departed to set up on his own again.

  The financial benefits to be had from the company’s technology depended on the degree of patent protection. With the telephone, Bell had fought repeated battles to maintain his claims. Bell’s personal relationship with the patent attorney and his meticulous record keeping had been particularly crucial.

  Unfortunately for Mauchly and Eckert, they received no such advice and as a consequence eventually lost their patent protection. Their first reverse came in 1952 when Remington Rand signed an agreement with IBM to cross-license technology that required IBM to make payments if patents were issued on the ENIAC and UNIVAC applications. IBM got the better part of the bargain. Then, in 1972, in a judgement over patent infringement between Sperry-Rand and Honeywell, the patents were overturned and Atanasoff was credited with the first electronic computer. Had Mauchly and Eckert paid more attention to the patent process, and been better advised on their submissions, they might have obtained the same protection that the Bell Companies achieved in the early years. The debate as to who built the first electronic computer still rages on, just as it does on the telephone. Whether it was Atanasoff, the scientists at Bletchley Park, or Mauchly and Eckert is largely irrelevant in an investment sense, because the main point is that as the new industry emerged the original ent
rants could not rely on patents to sustain their operations. As a consequence the battle came down to questions of technology, customer service and funding.

  World War II had caused a massive acceleration in the new technology of the computer. In Britain, Bletchley Park had seen the construction of the Colossus. In America ENIAC had been created, spawning EDVAC and subsequently UNIVAC. Other academic sites had been sponsored by the government to provide assistance with the war effort. At MIT, Project Whirlwind had been inaugurated in 1943 by the Special Devices division of the US Bureau of Aeronautics to produce a flight simulator. This project evolved to become a full-scale attempt to provide a real-time computing capability. The initial estimates of cost and duration were $200,000 and two years later these escalated to $8m (over $100m) and eight years. The project was only saved by the increased tension of the Cold War occasioned by the demonstration of Russian nuclear capability. The demands for a national air defence system produced SAGE, the Semi-Automatic Ground Environment, in which the computer from Project Whirlwind was an important element.

  The commercial significance of this was that IBM was commissioned to take the machine developed at MIT and produce the system for the armed forces. The company was given both the technology and the funding for its future development. Both were important. It has been estimated that in the 1950s the gross revenue on the project was in excess of $0.5bn (nearly $4bn) and that nearly 20% of the IBM workforce was employed on the project.⁸³ The project allowed IBM to bring its real-time capabilities to the forefront and directly provided the basis for the successful airline reservations system which was to be developed for American Airlines.

  The arrival of the transistor

  The limiting factor on analogue computers had been the fact that the physical accuracy of the model determined the accuracy of the resulting calculations. Electronic digital computers avoided this problem but had their own practical constraints. A machine built with relay switches was superseded by one which used vacuum tubes and could operate at much faster speeds. The computers of the 1950s employed a technology which essentially stretched back to Edison’s incandescent lamp, with updates from Fleming and De Forest. Just as previous models had been constrained, so too the computers based on vacuum tubes were limited by the characteristics of the tubes themselves. As the tubes were large and got hot, this meant that the more powerful the machine, the larger it needed to be, the more electrical power it consumed and the more heat that had to be dissipated.

  The prevailing view on how big the market for these computers could become was based on these constraints. The cost of the machine and its maximum power could be estimated, and given this the potential market could be approximated. Under this scenario computers seemed likely to be profitable, but with applications limited to large private and public sector users. All of this was absolutely correct – except that it was predicated on an assumption that the technology would remain static. If the vacuum tubes could be replaced, then all the working assumptions fell by the wayside – which is exactly what happened. The vacuum tubes had come out of the lighting industry, but had been developed further to build the radio industry. The ability of the triode to amplify signals caused intense interest at AT&T and the study of their properties became part of the scientific agenda at Bell Laboratories. Potential alternative media that could perform the same task without encountering the reliability and heat problems associated with the tubes were also studied. Out of this work was to emerge the transforming device of the transistor.

  The invention of the transistor and its adoption by the computer industry did not happen overnight. First it had to become reliable and cost-effective when compared to existing tube technology. The transistor was invented in 1947 at Bell Laboratories by William Shockley, Walter Brattain and John Bardeen. The work had begun in the 1930s with the search for a replacement for the somewhat unreliable valves and physical switches then integral to telephone exchanges. Bell Labs had been working on the purification of silicon and Shockley had also been looking at the potential use of copper oxide as a rectifier. This led in turn to the exploration of semiconductors, i.e. material that conducted current in only one direction. This work was put on hold during the war, but the body of knowledge in the area increased as the development of RADAR created a demand for improvements in the availability and use of semiconducting materials.

  After the war the scientific group at Bell was reassembled and work began again on the task of investigating semiconductor amplifiers. After three years’ work, the ‘transistor’ was announced to the outside world, which welcomed it with a muted response. The initial perception was that what had been invented was a solid-state replacement for the valve. The replacement, while smaller and free from the heat and power problems of the valve, was less reliable and difficult to produce. As a consequence it did not immediately displace the valve, which retained its position as the main component of computers through the early development of the mainframe. The inventors of the transistor were to receive the 1956 Nobel Prize for their efforts, but by then Shockley had already left Bell Labs to set up his own company, Shockley Semiconductor, to exploit the technology.

  The Shockley Semiconductor Laboratory was funded by Rockefeller and established in Shockley’s hometown of Palo Alto, California in 1954. It comprised Shockley, plus a large number of scientists he had brought with him. The principal target of the group was to take the transistor and increase its reliability and reduce its cost by improving the manufacturing process, although this was frequently compromised by Shockley’s desire to find further new inventions. The group at Shockley did not last long as differences over policy and personality conflicts with Shockley himself contributed to a split in 1957. Eight of the key employees left as a group to be funded by Fairchild Camera and Instrument in a new venture called Fairchild Semiconductor. This company was to play a pivotal role in the development of the computer industry, and in many ways can be said to have spawned California’s Silicon Valley.

  In the early years, the company’s focus was a scientific race against Texas Instruments to overcome the drawbacks in the manufacture and reliability of hardwired devices. Their solution was to adapt the manufacturing process so that the electronic circuits could be reduced in size and placed on small pieces of semiconductor material. Texas Instruments was the first to achieve this and in 1959 filed for a patent on its semiconductor ‘chip’. This chip still required hardwiring and TI found themselves leapfrogged by Fairchild when Robert Noyce devised a method known as ‘planar’ manufacturing, whereby the wiring was actually embedded in the silicon. These two advances effectively combined to overtake the transistor and condemned the vacuum valve finally to obsolescence. The transistor had been slowly encroaching and IBM was forced to develop a successor to its valve-based 800 and 650 models. Its next machine, the 1401, was intended to replace these models and by incorporating transistors rather than valves was a big step forward in computing capability.

  Computer wars

  The business imperative for IBM was to make it cheaper for companies to replace old accounting equipment with new computers. In the early years, the decision to switch to computing implied the need for programming resources and a big change in working practices. IBM’s solution was to provide a total solution for the customer. In other words, it would provide both the hardware and either applications software (such as accounting or payroll packages) or a simple, easy-to-learn function which would allow the customer to do programming in-house. IBM had a massive customer base over which to amortise the cost of writing these applications and long practical experience of working with customers to allow them to do it. The 1401 was also launched with a much-improved printer, nearly four times faster than its predecessor. The logic for customers was compelling and the 1401 became almost an overnight success. By this time the structure of the industry was beginning to resolve itself both in terms of the number and identity of the participants and the segments of the market they were seeking to address. Effe
ctively, however, it was IBM against the rest. The rest comprised a range of entrepreneurial start-ups, divisions of large companies in related businesses, and companies with obsolescent technology who were seeking some way of surviving.

  At this point, only two categories of company really counted for much. There were the manufacturers of computers and the suppliers of components. Software remained the preserve of the equipment manufacturers or their customers. Such was IBM’s dominance of the industry that its competitors had to make a decision which competitive strategy they were going to adopt. Since software was such an important issue for most clients, it was extremely difficult to persuade would-be purchasers to take the risk involved in moving away from IBM equipment. You either produced IBM-compatible equipment focused on a specialist niche where software was not an issue for the client, or you took on IBM across the board. Companies such as Control Data and later Cray took the niche route, supplying top-end powerful computers to technically sophisticated clients, capable of developing their own software anyway. Others such as Honeywell went down the compatibility route; the ‘200’ machine it launched in late 1963 provided almost four times the computing power of the IBM 1401 for the same price. This strategy was also adopted by RCA, which took the approach of waiting for the new IBM models to be announced and then seeking to produce a lower-cost version. Others, such as Burroughs and NCR, stayed with their own systems. They relied on their specialist expertise with their particular client groups, particularly in the banking and retail sectors, to retain their customer base.