TECHNOLOGY

Perspective

Semiconductors face the '80s Seven leaders of the boomingest business in the U.S. prognosticate the problems supergrowth could lead to

Gordon Moore, president of Intel, put it succinctly: "By 1986, the semiconductor industry could be producing something of the order of 101 4 functions per year. Who will use it all?" He, and 800 other participants in a recent colloquium in California's Silicon Valley, organized by the Santa Clara Chapter of IEEE's Electron Devices Society, heard some attempts at answers from industry leaders. But Dr. Moore was not satisfied. The rate of growth is such, he told Spectrum in a follow-up interview, that if the industry stays on its present upward curve of producing functions and tries to solve its problems as it always has in the past—by lowering prices—there will come a time in the 1980s when it could face disruption and a significant drop in revenues. "And that," he says, "despite exciting new markets on the horizon, in an oth-erwise blossoming economy, could be a disaster."

Nonetheless, the tone of the semiconductor industry panelists, and the packed ballroom of the Cabana Hyatt in Palo Alto, site of the meeting, was bullish." We've done it before, and we can do it again," seemed the predomi-nant theme of the discussion. In all, it was an "insider's" view of an aggressive industry looking for ways to help develop or force-feed markets that, in general, seemed too timid.

Addressing the applications, markets, and other issues of the 1980s were, besides Gordon Moore: Edward Davis, vice president, Systems Products Division, IBM Corpo-ration; John Welty, vice president and general manager, Motorola Semiconductor Group; Richard Soshea, general manager, Microwave Semiconductor Division, Hewlett-Packard; Wilf Corrigan, chairman and chief executive, Fairchild Semiconductor Company; Charles Clough, vice president and marketing manager, Semiconductor Di-vision, Texas Instruments; and Floyd Kvamme, vice president, National Semiconductor, standing in for the company's president, Charles Sporck. James Riley, for-merly president of Intersil and Signetics and now of Da-taques, chaired the meeting.

In addition to the elite of the semiconductor and Bay Area industries, the audience included financial analysts and venturers hoping to catch clues to the markets of the '80s. Following are some of the highlights of what the semiconductor industry leaders believe to be in the cards for the next decade.

Moore on the problem of growth "Anytime you make extrapolations that depict straight

lines on log paper," according to Gordon Moore, "you come up with a catastrophe... such as the Club of Rome. But the main utility of performing the task is to show what kind of catastrophe to look for so that you can start

Nilo Lindgren Contributing Editor

Gordon Moore

soon enough to make the necessary adjustments to live with it."

Moore's extrapolations and projections of usage of circuit functions are shown in Fig. 1 and Tables I and II. He notes that the prices of semiconductor products tend to decrease rapidly, whereas the total dollar volume must increase. Thus, the industry must produce an ever-in-creasing number of circuit functions. As Fig. 1 shows, the semiconductor industry grew from approximately 5 × 108

circuit functions in 1960 to 101 2 functions in 1976. These figures are based on industry statistics of shipments, starting with transistors in 1960 and moving into inte-grated circuits by the '70s. (Moore defines "function" as either a gate or a memory bit.)

Where are the largest chunks of these semiconductor functions being used? Table I is the result for 1976. Moore estimates that some 2000 mainframe computers are sold per year, each using about 50 000 functions. One may quibble with these figures—assigning perhaps 500 000 functions per mainframe—but the difference is trivial in this scheme of things.

Semiconductor memory is the most significant con-tributor to the market, at something like 2 × 101 1 func-tions, and 30 million pocket calculators, at an estimated 3000 functions each, add another 101 1 functions. An in-teresting observation that emerges from these estimates is that pocket-calculator circuit usage is now a thousand

IEEE spectrum OCTOBER 1977 42

or so times larger than computer mainframe usage. In any event, based on just these major markets, Moore arrives at approximately 3 × 101 1 functions, corresponding to the lower limit in Fig. 1 for 1976. Then extrapolating in what he regards as a conservative manner, as shown in Fig. 1, he arrives at his figures for 1985 production if the industry continues at its current rate—in the range of 101 3 to 101 5

functions, or a quarter million functions for every person on earth. By contrast, it is 30 to 300 times the total amount of electronics produced in 1976 for all applica-tions.

To attempt to answer his own question as to who will use all these functions, Moore makes some estimates re-garding the large foreseeable markets of the mid-'80s. Table II shows that, even with larger data-processing systems with significantly larger numbers of functions, and with video games added to ten million television sets each year in addition to 200 000 circuit functions in each of ten million new automobiles, the total demand still will be of the order of 101 3 versus the industry's extrapolated ability to produce 101 4 or 101 5 functions per year.

"Given these numbers," concludes Moore, "for the stability of the industry, we face either or both of two chores: practicing some kind of 'birth control/ which is contrary to our mentality, or starting to develop new uses, new consumers, and new markets."

To Moore, and to others, the answer for the '80s may be more electronics in the home. With about 50 million family units in the U.S., Moore calculates that a million electronic functions at work for each of these units would begin to match the slope of the semiconductor industry growth. The potential for the home market, he says, is tremendous—logic devices for energy conservation, mi-croprocessor controls in appliances, smart burglar alarms, and so on, almost ad infinitum.

However, with one or two exceptions, Moore feels the semiconductor industry is shortsighted in its pioneering and development of these potential new markets. "Ev-eryone, up through top management, is hung up on the problems of getting this month's shipments out." And,

I. Circuit-function usage—1976

Functions No. of Total per Unit Units Functions

Computer mainframe 50 000 2 000 0.001 X 1 0 1 1

(logic) Minicomputers 10 000 20 000 0.002 × 1 0 1 1

Semiconductor memory 2.0 X 1 0 1 1

Pocket calculators 3 000 3 × 10 7 1.0 X 1 0 1 1

Electronic watches 1 000 3 × 10 7 0.1 X 1 0 1 1

Total 3.0 X 1 0 1 1

II. Estimated circuit-function usage—1985

Functions No. of Total per Unit Units Functions

Data-processing systems (logic) 10 6 10 5 0.001 X 1 0 1 3

Semiconductor memory 0.6 X 1 0 1 3

Calculators, including scientific computers 3 × 10 4 3 X 10 7 0.1 X 1 0 1 3

Watches 2 X 10 3 2 X 10 7 0.04 X 1 0 1 3

Video games 10 5 10 7 0.1 X 1 0 1 3

Automobiles 2 X 10 5 10 7 0.2 X 1 0 1 3

Total 1.0 X 1 0 1 3

he adds, they are ever ready to exploit an opportunity that someone else has developed. "We need more com-panies in this industry willing to do more pioneering, to open and develop new markets . . . to exhibit more ma-

. turity." Moore concludes that his principal mission is to make more people in the industry realize that life is not going to go on in the same way forever. "The way we look at things," he says, "is going to have to change."

Davis on data processing Could the data-processing industry—historically, the

big customer of the semiconductor industry—consume more than the share of digital circuit functions that Moore projects for the 1980s? This was one of the ques-tions addressed by Edward Davis of IBM. Noting that in 25 years the industry had grown from a "standing start" to a multibillion-dollar enterprise, Davis is sanguine about its continued prospects of growth, and therefore about its capacity to devour vast amounts of semicon-ductor functions.

io° I ι ι ι ι 1

I960 '65 70 75 '80 '85 [1] Gordon Moore's extrapolations and projections of usage of circuit functions indicate that by 1985, if the industry continues at its present rate, it will be producing some quarter million functions for every person on earth. (The two curves bound the probable margin of error.)

Edward Davis

Lindgren—Semiconductors face the '80s

The semiconductor industry has, in fact, been highly responsive to the needs of the computer industry, espe-cially in the supply of memory. In the six years from 1971 to 1977, semiconductor memory size has gone from 1000 to 16 000 bits/chip, an increase of 60 percent per year. In measuring the actual productivity improvement, Davis notes that chip size and wafer size have grown, and cal-culates a net productivity, from the "bit business point of view," of 55 percent per year.

Looking at the technological innovation that supports this kind of growth, Davis asks whether or not that kind of mastery or "pushing of molecules" can continue for the next ten years at the same rate as in the past decade. His answer, in basic agreement with Moore, is that there will be no visible reduction in technological innovation. Davis cites the following examples: continuing refinement and cleverness in circuit design using photolithography; the use of electron-beam tools; an advance in micro-type processing that could push control to 2.5 ìéç by 1980 and to less than 1 ìðé by 1985; changes in fabrication processes with plasma sputtering; dry etching as opposed to chemical etching; and the use of low-temperature oper-ations. Davis also projects that there will be better than a factor-of-10 improvement in defect densities within the next five to ten years, and that wafer size will continue to increase—to 4 inches by 1980 and to 6 inches or more by 1985. Moreover, he says, wafer cost will remain about the same, regardless of size.

1 000 000

1965

[2] Edward Davis' figures indicate a tremendous increase in the number of bits per chip over the next decade.

[3] Davis also projects a great increase in the number of logic circuits per chip, wi th bipolar logic becoming competitive with field-effect transistor memories.

100 000

10 000 h

1000

The implications of these advances for the data-pro-cessing industry, in Davis' figures, are on the same track as Moore's. Figure 2 projects for semiconductor memories a tremendous increase in bits per chip by 1985. The growth of logic circuits per chip (Fig. 3) is similarly large, with bipolar logic becoming very competitive with FET for numerous reasons. And, as Fig. 4 indicates, the aver-age price per bit will continue to decline, to around 30 millicents by 1985.

Thus, Davis comes to the same problem as Moore: Who is going to use it all? Where will all these bits be applied, and will the demand be such that these projections will be borne out?

But Davis draws a more "bullish" model of the future, as far as data processing is concerned, based principally on its character as a growth industry. In developing his argument for an increasing use of semiconductors by the data-processing industry, Davis makes two assumptions. One is that revenue from hardware is proportional to manufacturing costs; that is, there is a direct relationship between the marketplace price for data-processing equipment and the manufacturing cost. The other is that the semiconductor content of mainframe computers, etc., either will remain constant or will grow over time—that it will be a growing percent of manufacturing cost.

Exactly how big, then, is that future market likely to be? Davis notes that one industry observer estimates a 19-percent growth rate for mainframes, miniprocessors, storage devices, files, and terminals. Davis then calculates that the semiconductor component of the data-processing industry is currently something like 4 percent. The result of the calculation is—if the 19-percent growth rate esti-mate is used—that this 4 percent would account for $0.4 billion in 1975 and project to $1.45 billion in 1983 if the semiconductor component remains constant.

However, Davis flatly asserts, "I frankly believe that the semiconductor portion of the computer industry hardware is going to grow." And he projects that content as going from 4 percent to 6 percent by 1983, as he shows in Fig. 5. He bases this projection on the following: mainframes will go from 4 percent to roughly 6 percent semiconductor content in the next few years; other min-icomputers will range around 9 or 10 percent; terminals (displays) could be anywhere between 6 and 12 percent; peripherals will be smallest, somewhere near 3-4 per-cent.

Davis concludes that the average growth to 6-percent content will result in a 26-percent compounded growth rate for the use of semiconductors in data-processing equipment (as contrasted with the 15 percent per year that many semiconductor industry people feel should be the minimum growth rate).

But even at this higher growth rate, Davis points out, the data-processing industry will still consume only 3 X 10 1 2 bits (or functions, in Moore's terminology); and for the lower 4-percent constant portion, the data-processing consumption would be a mere 1 × 10 1 2 bits. Even in the optimistic 6-percent model, total usage of semiconductors will be nowhere near the projected capacities of which the industry is capable. As Davis puts it, "It's a race between the productivity engine and the finding of new applica-tions."

Welty on microprocessors and automobiles "I bet," says John Welty of Motorola, "that there have

been more automobile engineers in our plant during the

I E E E spectrum OCTOBER 1977 44

past three years than there were in the first 20 years of our semiconductor business!" That is one sign of a very sig-nificant change in the conditions and outlook of the au-tomobile industry, which is now, according to Welty, moving rapidly to make microprocessors an integral part of automotive design. And, he is quick to add, the auto-mobile industry "is not looking at electronics in the iso-lated way it did in the past—radios, alternators, transistorized ignitions—but from a systems point of view."

Many factors have pushed the U.S. automobile in-dustry, notorious for its inertia regarding change, toward new philosophies and methodologies. Beset by Govern-ment standards on emissions and safety, by the increasing cost of fuel, and by competition from overseas that has achieved a significant penetration of the market (as much as 20 percent in some cases), the industry is being chal-lenged as it has not been in half a century. Automobile manufacturers are finally scrambling, Welty reports, to become thoroughly versed in the microprocessor and semiconductor technologies, as a way "to solve many of their problems." As a result, Welty projects that the au-tomotive industry will be using some four million mi-croprocessor units (MPUs) by 1981, "making it the largest single application."

But this deep involvement in microprocessors is also changing the basic relationship between the automotive and semiconductor companies. In the past, Welty points out, there was an easy relationship with automotive customers. "They told us what to make and we made it!" Today, however, the semiconductor supplier is intimately involved in helping automobile designers solve a variety of problems. Thus, not only are the automobile companies

; 1000,

[4] Projections indicate that the average semiconductor industry price per bit will continue to decline, to about 30 millicents by the mid-'80s.

[5] From $10 billion in 1975, mainframes, peripherals, and minicomputers should experience a 19-percent growth in rev-enues by 1985.

2000i

j§ 1500

s ιοοομ

500

1970

building up their internal expertise in electronics, but companies such as Motorola are adding to their design teams engineers who are firmly grounded in the auto-motive industry.

This new situation is posing a number of challenges to the semiconductor companies involved. The first Welty cites is to get the cost curve on MPUs down as fast as possible. The dynamics should be simple: The demand over the next five years will lead to a massive production volume that will reduce MPU prices dramatically, which, in turn, will provide a strong stimulus for continuing ex-pansion of MPU markets. But this need to get the costs down to the levels required in automotive applications is somewhat ironic, comments Welty, since the cost of the semiconductors in the contemplated systems are the least expensive parts of those systems. The high costs are for sensors, actuators, and interconnection wiring harness— which have also been the major cause of failures, as in, for instance, the antiskid system for trucks.

The second major challenge cited by Welty relates to the lack of efficient, low-cost sensors. Unless sensor technology grows rapidly over the next year or so, Welty predicts, the rate at which MPUs are incorporated by the auto industry will be seriously retarded. Moreover, no one now "seems to know what approach to take in sensors... this is a problem that should be addressed by the semi-conductor industry."

The third challenge, and probably the toughest one, is the efficient coupling of electronics to the actual auto engine. Welty seems to think that once the MPU gets on board the car, and the analog-digital interfaces are spe-cifically defined, many new applications will emerge in terms of engine management and control. To this end, "all the major auto companies have large and heavily funded MPU programs."

Besides challenges, Welty also identifies two major debates now going on in the auto industry vis-a-vis the use of MPUs, both of which will impact the development of the semiconductor companies. One concerns a problem very familiar to the computer community—dedicated versus general-purpose processors. Both approaches are being taken, depending upon a particular company's previous experience with LSI technology. There appears to be a concensus, Welty observes, that eventually all microprocessors in cars will be dedicated and custom-ized. However, general-purpose MPUs will likely dominate the market for some years, as the auto makers feel their way. A gradual evolution to dedi-cated systems over the next four or five years will be beneficial since it will give the semiconductor industry time to accommodate to the specialized requirements— something that didn't hap-pen easily in the early days of LSI.

The second debate is familar to computerists— centralized versus distrib-uted systems. Here, too, the car makers may track the

John Welty

Lindgren—Semiconductors face the '80s

experience of the computer field in general, advancing from one or two central processors initially toward greater redundancy and a distribution of electronic functions throughout a vehicle to enhance safety and performance. Welty's perception is that eventually cars may be equipped with numbers of small satellite semiautono-mous computers, which would receive their "exotic" capabilities from a central computer, but which would function on the local level in cases of wiring or other malfunctions. As such systems evolve, the roles of opto-couplers and fiber optics will grow in importance.

Such projections have led to a lot of "loose talk," con-cludes Welty, about the "ultimate smart car" with space-age sophisticated collision-avoidance systems. And certainly cars will be smart enough, but the ultimate in-telligent automotive vehicle "isn't likely to happen in the 1980s."

Soshea on communicat ions

Where else will some of Gordon Moore's projected functions be applied in the '80s? Richard Soshea's an-swer, based on Hewlett-Packard's perspective in op-toelectronic and microwave semiconductors, is that large numbers will be utilized in relatively new communica-tions media. The traditional forms of communications— voice and data communications via wire or cable—says Soshea, are not likely to grow or change much over the next ten or 15 years. But the newer forms—terrestrial and satellite microwave—are growing rapidly. Starting from nothing, they may, he projects, approach a billion dollars a year by 1990. It is in some of these new high-growth media that he sees specialty semiconductors playing an important "synergistic" role.

Although we have been hearing a lot about satellite communications systems, Soshea comments, they have been chang-ing the complexion of communica-tions more rapidly than is generally recognized. Geosynchronous satel-lites have supplemented terrestrial microwave and submarine cable links considerably. For instance, present transatlantic capacity is about 15 000 voice circuits—half by submarine cable and half by satel-lite. With this growing use of satel-lite communications, the rates for transatlantic telephone calls have

Richard Soshea been cut in half. And it is clear, says Soshea, that satellite voice com-munication will continue to grow.

Just getting underway are business data communica-tions, of which the most publicized example is the Sat-ellite Business Systems (SBS) venture put together by IBM, Aetna Insurance, and Comsat General. The SBS all-digital systems, operating at higher frequencies, will have different capacities and characteristics than today's satellite business communications. As their antenna sizes decrease, it will become cost-effective and convenient to mount them on plant rooftops for intracompany com-munications via satellite as an alternative to microwave links. Fiber-optics communications is another medium that is predicted will be a billion-dollar enterprise by the end of the 1990s. Probably the bulk of fiber-optics traffic will be telecommunications, but a significant portion will be digital mainframe-to-terminal connections. Because

digital communications promises to be rapid and rela-tively error-free, it will probably compete with traditional mail for business data communications.

Another form of communications that will require specialty semiconductors is television via satellite as an adjunct to the traditional line-of-sight broadcasting and cable television. Though it probably will not find wide use in the U.S., except for specialized local communications such as for law enforcement agencies, satellite television is likely to become popular in parts of Europe, in Japan, and in other insular areas. The key, says Soshea, will be smart antennas operating in the high-gigahertz range.

Still another communications medium—navigation and marine—is in the early stages now, but ship-to-ship and ship-to-shore communications via satellites in geostationary orbits will be standard by the 1980s when, Soshea projects, every significant naval vessel will have its own such communication system.

For navigation, the present Loran and OMEGA sys-tems will be supplemented within a few years by a mili-tary-funded system called NAUSTAR, or Global Posi-tioning Satellite system. By the mid-'80s, 24 of these satellites will be deployed in three rings around the planet; any vehicle will be able to determine its position through two of the satellites with an accuracy of a few meters. A "very healthy industry," comments Soshea, has already begun to exploit these satellite signals for com-mercial activities as well. Microwave landing systems and aircraft-collision-avoidance systems also should, once international questions have been settled, come into ef-fective operation by the mid-'80s.

These systems, all underway and funded, are just a few of the many that will be developed in the next decade. To make them viable, the semiconductor industry will need to develop the necessary specialty semiconductors—the higher-performance devices and circuits. There is, con-cludes Soshea, a synergistic effect between these new communications systems and semiconductors that will create new business for all segments of the semiconductor industry.

Corrigan on vert ical integration

Another response of the semiconductor companies to the potential of producing more electronic functions than the world might easily absorb is to become vertically in-tegrated. Though the usual connotation of vertical inte-gration is that a semiconductor company, for instance, would move toward making and marketing the end products of which semiconductors are a part, Wilf Cor-rigan of Fairchild Semiconductor reminds us that the term can be read in both directions—that is, a semicon-ductor company might also move down the ladder toward making its own silicon, gallium arsenide, etc.

What most people also forget, says Corrigan, is that the original semiconductor companies—General Electric, Westinghouse, Hughes, et al.—were vertically integrated "the other way." Even Motorola got into the business by starting to make germanium transistors for its automotive radios. And the Japanese, who have been somewhat of a thorn in the side of the U.S. semiconductor industry, have always been vertically integrated in their operations.

The differences in direction that vertical integration takes can be instructive. For instance, the original ger-manium transistor with which the U.S. thought it would start the electronic-data-processing industry is what the Japanese used to launch their transistor-radio business.

IEEE spectrum OCTOBER 1977 46

Wilf Corrigan

And companies like Sony made their own germanium transistors. Thus, vertical integration is not a new idea, remarks Corrigan, and the reasons why companies are moving into it today are not new either. It can serve, wisely applied, as both a defensive and offensive business strategy.

Outside the United States, whether in Europe (Philips, Siemens, et al.) or in Japan (Fujitsu et al.)9 vertical in-tegration is the general rule. Only in the U.S., observes Corrigan, has the concept prevailed of an individual semiconductor company whose sole function is to make semiconductors. However, that situation is reversing as more and more semiconductor companies have decided to integrate vertically. For example, National, Intel, and AMS are going into memory systems; Fairchild, Texas Instruments, National, and Intel have gone into electronic watches; Fairchild, National, and a few others have gone into video games; and Texas Instruments, among others, is moving toward minicomputers. "And," says Corrigan, with sarcastic amusement, "everybody makes micro-computers, even those who don't!"

The question is—why? What are the motivations and reasons this time around? One major reason may be that the semiconductor business has proved to be strongly cyclical. The trick in vertical integration "is to make certain you don't get two down-cycles coinciding... your company may not be able to handle it."

But perhaps the most important reason for semicon-ductor companies to become involved in some form of end product relates to the projections drawn by Gordon Moore. As Corrigan puts it, "When you have drawn all those lines on log paper out to 1985, and get a nice warm feeling, and then come back the other way to talk to a relatively timid customer base, you begin to worry about where all these devices are really going to go." For their part, the customers of the semiconductor houses have generally wanted the semiconductor industry to put the brakes on a bit, so that they can run through their own product life cycles, and get their return on investment before their products are obsolete. On the other hand, a semiconductor company with a new item can't afford to wait to sell it, since the competition will be moving in with the same kind of item, possibly at a lower price. Caught in these dynamics, the semiconductor producer is driven to move faster than the orderly pace of the normal end customer; it virtually "forces you," says Corrigan, "to venture out and build the product yourself."

Still another reason for these "tentative steps" toward

end products is the capital-intensive character of the semiconductor business. In terms of sales, it simply takes very high equipment assets to produce semiconductors. But the industry is in a continual mode of making its productive equipment obsolete; as, for example, in moving from 3-inch to 4-inch wafers. The only way to "lever those assets," concludes Corrigan, "is to move into end products that are not quite so capital-intensive, and yet utilize the technology."

With warnings about the dangers and risks involved for companies that undertake to make particular end products, Corrigan nevertheless concludes that the semiconductor companies must integrate vertically, even if only partially—because the "technology is going to come, and it is going to come faster and faster in the next five years!"

Clough on market ing

In addition to the technology itself, a crucial factor for the semiconductor industry in the '80s will be in the ways it handles its marketing. In addressing such questions, Charles Clough of Texas Instruments stresses the fact that although complexity of products and billings mul-tiplied steadily from 1960 to the present, marketing techniques remained relatively unchanged. Whether it was simple transistors originally, or integrated circuits more recently, the products were sold basically for OEM users by customer-oriented technical salesmen and en-gineers backed up by local distributors and factory ap-plication engineers and by extensive advertising.

But now the microprocessor is bringing about a basic change because it is creating entirely new end products, and in some cases applications entirely new to electronics itself. Thus, to meet the needs of its large customers, whether in the automotive or telecommunications fields, in television accessories or in computers, the semicon-ductor marketeers are obliged to build up their technical sales forces with people who have a thorough knowledge of the particular industry they serve. What this means, says Clough, is that the sales representative is becoming a "local systems-oriented or microprocessor-applications specialist."

Modes of distribution will also be changing to a degree. Experiments in recent years with distributor microproces-sor design centers have failed financially so that, Clough says, the "creation of user de-mand for high-technology products will remain the re-sponsibility of the semicon-ductor manufacturer." To win the significant markets that microprocessors promise, the producing companies will have to provide technical design services rather than expect their present distributors to reshape their style of service-oriented management. No doubt this means that many distributors face some critical decisions regarding their own

Charles Clough

Lindgren—Semiconductors face the '80s 47

survival in the next decade. With growing product complexity and a widening

customer base, the impact on manufacturing production cycles is also becoming more critical, so that better and more timely information is needed for the maintenance of backlogs. To meet this need, manufacturers are setting up systems that allow customers to enter orders directly to manufacturing backlogs, with all the handshaking, acknowledgements, shippings, and billings handled au-tomatically. As these systems expand and become more sophisticated, predicts Clough, sales representatives, and even customers, will be able to forecast or match usage on a real-time basis with production cycles. With such terminal systems, communications between producers and users—always important—will become more timely, and presumably more effective.

"Customer education" has always been a working slo-gan for most high-technology companies that aim to survive and grow. With microprocessors, the need for such education is more pressing than before. Accordingly, companies such as Texas Instruments have established "learning centers" in key market areas around the U.S. By 1980, says Clough, these special learning centers, aimed at creating new customers for electronics, will be located in every key market in the world. With micro-processors, the need for "software technical support is enormous."

Yet, some basic laws about marketing haven't changed despite high technologies. "The marketing organizations that will be most successful over the next ten years," concludes Clough, "will be those that can supply better and more reliable products, sell them at a lower price, and deliver them on time. And that 's really what it's all about!"

Kvamme on Japanese compet i t ion

Though his presentation was brief, Floyd Kvamme's message about a special problem faced by the U.S. semiconductor industry—"equal treatment" in the Japanese market—was pointed. Citing the beneficial aspects of open competition in the semiconductor in-dustry—its spur to innovation and cost reductions, and

its enlargement of the industry Floyd Kvamme into a worldwide electronics

community—Kvamme asserts that one country, Japan, "does not play by the same rules that everyone else does in the in-ternational market."

Tariffs are unequal, al-though this is not the major problem. In the U.S., tariffs on semiconductors are 6 percent; in Japan they are 12 percent. Kvamme says that represen-tatives of the U.S. semicon-ductor industry have urged that there be a total elimina-tion of such tariff barriers.

He goes on: "That, however, is only the beginning of re-straints on selling in Japan. In the U.S., we are open for the marketing of Japanese prod-ucts but a U.S. company selling in Japan faces increased im-

port duties, import quotas, requirements for import li-censes, red tape, and the necessity to establish a pres-ence—but a simple presence such as a sales office takes years to establish. We also face what I call the Japan Club, an unwritten policy among Japanese users to buy outside only what cannot be obtained locally in Japan." In con-trast, Japanese companies selling in the U.S., in addition to tariffs, get massive Government subsidies for devel-opment, assistance in export, insurance coverage, very-low-interest loans to subsidize sales, etc."

"To gain technology," Kvamme asserts, "the Japanese are establishing listening posts in the U.S. They have strict control on foreign investments in Japan to protect their own markets, control tied to the necessity for li-censing arrangements. They have a patent policy that effectively forbids U.S. ownership, or non-Japanese ownership, of the most basic patents in this industry, patents on devices that were invented and paid for by U.S. industries."

What must be done, asserts Kvamme, is to seek equi-table treatment in the international market for semi-conductors, where each company competing in a given market is competing using the same rules. That is, firms competing for a portion of the U.S. market must abide by U.S. requirements that their developments not be mo-nopolistic in practice; namely, that the cartel-like ac-tivities of heavily subsidized joint research between competing companies and patent-pooling from the joint research not be allowed with regard to products intended for the U.S. market, since such practices are forbidden in the United States. Until equity is reached, Kvamme urges that U.S. companies consider the source of all products that they buy so as not to encourage unfair trade practices and that U.S. industries work with the U.S. Government to establish procedures to curb the Japanese practice of subsidizing development of products intended for export to the U.S.

The U.S. semiconductor industry is not afraid of true competition, says Kvamme, and, in fact, encourages it. In its absence, however, he warns that "as the semicon-ductor technology goes, so goes the entire electronics in-dustry."

Other perspectives

These different facets of the semiconductor industry today as it faces the 1980s did not touch, except for Japan, on many other factors that could impact it from the outside. There was little guessing about unforseeables in the international arena, or the potential interplay with the Third World, except in the role of the Southeast Asian labor force in U.S. semiconductor production, and the stated recognition that the electronics industry is a worldwide industry. To that point, Wilf Corrigan did project that, given the economics and capitalization of the semiconductor industry, most of its products would be assembled with automatic equipment in the market of use, and probably in less than ten years. The economic crossover point for automatic assembly versus hand wage rates, according to Corrigan, will come in 1982.

Basically, the perspective that emerged in Palo Alto is one of an industry caught up in its own momentum, wondering how to transform itself into other kinds of industries—which seems to be a generic problem of many successful high-technology ventures. The triumphs and the troubles of the semiconductor industry are tightly interwoven. •

Lindgren—Semiconductors face the '80s 48

Wiley Highlights for Autumn 1977... SURFACE WAVE FILTERS Design, Construct ion, and Use Edited by Herbert Matthews A practical state-of-the-art text and reference that teaches the art of designing, making, and using ultrasonic surface-wave devices. Presents basic principles, design methods, fabrication, packaging, and typical applications. Provides a complete understanding of surface-wave filters proven practical and presently in use. 528 pp. (1-58030-9) 1977 $29.95

GROUNDING AND SHIELDING TECHNIQUES IN INSTRUMENTATION, 2nd Ed. Ralph Morrison New, updated edition of a popular reference for engineers, techni-cians, designers, and all who have occasion to design, specify, or apply electronic equipment. This book gives the basic principles of electrostatics in a step-by-step, easy-to-understand manner that lets the reader visualize the nature of the electrostatic enclosure and avoid common errors. New sections cover digital circuits and other recent developments. The electrostatic enclosure diagrams have been carefully redrawn for clearer understanding. 146 pp. (1-02992-0) 1977 $15.50

SEMICONDUCTOR POWER DEVICES Phys ics of Operat ion and Fabrication Technology Sorab K. Ghandhi The physics of operation and the fabrication technology of power diodes, transistors, and thyristors will interest all those involved in the field of power engineering. Emphasis is on those areas which are unique and of special importance to semiconductor power devices. This book also brings together material heretofore scattered through-out the literature. 352 pp. (1-02999-8) 1977 $19.50

AN ANATOMY OF RISK William D. Rowe Examines how society, or its agents in Congress and regulatory agencies, can set levels of acceptable risk for technological systems and programs. Provides techniques for addressing the problems and focuses on the formal analysis of risk, as well as on the more subjective problems of risk acceptance and its interaction with estimated levels of risk. 488 pp. (1-01994-1) 1977 $26.95.

ADVANCED SYSTEMS DEVELOPMENT MANAGEMENT John de S. Coutinho Describes the management of the development of large advanced systems incorporating significant advances in new technology and involving elements of uncertainty and risk. Discusses the development cycle, beginning with the recognition of the need for the new system, its technical definition, source selection, contracting, development, qualification, acceptance, operational testing, and release for pro-duction. Many of the techniques described evolved from the military and space programs that developed since World War II. 433 pp. (1-01487-7) 1977 $23.00

NUCLEAR POWER REACTOR SAFETY E.E. Lewis An integrated treatment stressing the principles of reactor safety and offering extensive discussions of the safety characteristics of the major types of fast and thermal power reactors. Covers the interaction of neutronic, thermal-hydraulic materials phenomena during reactor transients and includes materials on the modeling of reactivity feedback, of hydraulic behavior and of several other aspects of transient analysis. approx. 640 pp. (1 -53335-1) Dec. 1977 $32.00

Available at your bookstore or write to Nat Bodian, Dept: 092-8374

WILEY-INTERSCIENCE a division of John Wiley & Sons, Inc.

A USER'S HANDBOOK OF SEMICONDUCTOR MEMORIES Eugene R. Hnatek A practical how-to-do-it volume that shows the design engineer how to select from among the available semiconductor memory technologies and product offferings for his design needs; and how to use and apply semiconductor memories to a particular systems requirement. 652 pp. (1^0112-9) 1977 $29.50

COMPUTER STORAGE SYSTEMS AND TECHNOLOGY Richard Matick

Combines all related aspects of computer data storage systems and technology in one volume, and relates the various aspects of data structure to system hardware design and tradeoff considerations. Emphasis is on fundamentals. Ideal for self-study, as a text, or as a working tool. 667 pp. (1-57629-8) 1977 $29.95

SOLAR HEATING DESIGN By the f-Chart Method William A. Beckman, Sanford A. Klein, & John A. Duffie Provides a practical method for sizing a heating system that combines solar heating with a conventional auxiliary energy source (furnace) to obtain the least costly design. Considers all parts of the system: a solar collector that heats either liquid or air; an energy storage unit; and auxiliary heater. 200 pp. (1-03406-1) 1977 $14.95

THE COUPLING OF EXTERNAL ELECTROMAGNETIC FIELDS TO TRANSMISSION LINES Albert A. Smith, Jr. Unwanted electromagnetic fields are often coupled,from external sources into digital transmission lines, communications circuits and other sensitive electronic circuits. Here are effective techniques for analyzing and removing these undesirable effects, from basic trans-mission-line theory to practical design procedures and worked exam-ples. 132 pp. (1-01995-X) 1977 $14.50

NONLINEAR ELECTRONIC CIRCUITS Aldert van der Ziel A simple, comprehensive introduction to nonlinear circuits, transient phenomena, and logic circuits, including idle circuits and their transient responses; transient behavior in nonlinear bipolar transistor circuits; time bases and blocking oscillators; generators of sinusoidal waveforms; flip-flops, latches, and their application to storage and transmission of information; and more. 267 pp. (0 471 02227-6) 1977 $14.95

Tpiease send the books indicated for Ð 15-DAY FREE EXAMINATION. (Restricted to the continental U.S. and Canada.) Mail to: WILEY-INTERSCIENCE

P.O. Box 092, Somerset, N.J. 08873 • Payment enclosed, plus sales tax. Wiley pays postage/han-

dling. We normally ship within 10 days. If shipment cannot be made within 90 days, payment will be refunded.

• Bill me. • Bill firm or institution.

• SURFACE WAVE FILTERS $29.95 (1-58030-9) • GNDING. & SHIELDING TECH. IN INSTR.

2nd Ed $15.50 (1-02992-0) • SEMICONDUCTOR POWER DEVICES $19.50 (1-02999-8)

• ANATOMY OF RISK $26.95. (1 -01994-1) • ADVANCED SYSTEMS DEVELOPMENT

MANAGEMENT $23.00 (1-01487-7) • NUCLEAR POWER REACTOR SAFETY . . . . $32.00 (1-53335-1) • USER'S HNDBK. OF SEMICOND. MEM $29.50 (1-40112-9) • COMR STG. SYSTEMS & TECH $29.95 (1-57629-8) • SOLAR HEATING DESIGN $14.95 (1-03406-1) • COUPLING OF EXTERNAL ELECTMAGNTC.

FLDS $14.50 (1-01995-X) • NONLINEAR ELECTRONIC CIRCUITS $14.95 (1-02227-6) Prices subject to change without notice.

NAME

605 Third Avenue New York, N.Y. 10016

In Canada: 22 Worcester Road, Rexdale, Ontario 092 A8374-51 Circle No.

AFFILIATION^

ADDRESS

CITY _STATE/ZIP_

29

IEEE spec trum OCTOBER 1977 49