Information Processing

Intel the Microprocessor Champ Gambles on Another Leaf Forward

Business Week

April 14, 1980

Twice during the past 10 years, Intel Corp. has launched controversial drives aimed at drastically changing the way in which the manufacturers of computers and other electronic equipment design their products. And each time, forging into the future paid off dramatically -- first with the computer memory chip and then with the microprocessor, the computer on a chip. Not only did these powerful new semiconductor technologies create new industries and disrupt existing ones, but they also turned Intel into the premier technology-based growth company of the 1970s.

The Santa Clara (Calif.) semiconductor maker grew more than a hundredfold in the last decade -- from $4 million in sales in 1970 to $663 million last year -- rising from obscurity to a strong No. 4 in the integrated-circuit business.

Now, Intel is kicking off a concept that could dwarf the two earlier campaigns both in audacity and in potential impact. In a series of top-level forums with its customers and in a spate of trade-press publicity, Intel is outlining long-range plans for devices complex enough to put an entire minicomputer -- and even a large mainframe computer -- on a handful of silicon chips. The company is also proposing to produce and market as standard products most of the software -- the lists of instructions telling the computer what to do -- that would put this power to work, a job that semi- conductor makers have always left for their customers. "We've made computer users out of people who don't know what a computer is," says Intel Chairman and Chief Executive Officer Gordon E. Moore. "Now we'll do the same thing with computer systems."

The task that Intel is setting for itself is truly awesome. The integrated circuits at the heart of the family that it proposes will each contain the equivalent of more than 100,000 individual transistors. "Just to define one of the processors we're talking about takes 10 man-years of engineering time," Moore points out. So far, Intel claims, it has invested $100 million in its existing and prospective lineup of "microminis" and "micromainframe" chip computers. And that may be just the beginning.

The marketing task facing Intel is at least as formidable as that required to popularize the computer on a chip a decade ago. The company then had to persuade equipment makers to give up developing new circuit designs for every new product and learn instead to program the microprocessor to do the job. Now, by standardizing major software elements such as the operating system or a high-level language, Intel hopes to reduce much of the programming cost that it sees as the biggest barrier to widespread application of its powerful new products. "What Intel is offering is a whole new way of doing things," says Marshall C. Kidd, manager of computer technology analysis for General Electric Co., a major customer.

But many customers may resent any intrusion into this domain, and others even question Intel's competence to carry it off. One such doubter is Herbert J. Richman, senior vice-president at Data General Corp., the Massachusetts mini- computer company that is both an Intel customer and a competitor in a growing number of markets. "Intel has the potential because of its technical capabilities and financial resources," he says, "but that's not a guarantee. There is a long way from being creative in semiconductors to implementing sophisticated operating systems."

To make matters worse, Intel must pull off this transition at a time when its entire organization is being stretched thin by the unprecedented spurt of growth that has gripped the semiconductor industry in the past two years. Thanks to the seemingly insatiable demand for its leading-edge memory and microprocessor products, Intel's sales jumped an astonishing 66% last year, and most analysts following the company believe that sales will reach $900 million this year, even though some softening is now appearing in semiconductors.

Through recessions and booms, Intel has managed to keep its pretax profit margins safely above 20% -- twice the average of its major competitors, according to industry analyst Benjamin M. Rosen, of Rosen Research Inc. its return on equity has, stayed in the 25% range for several years, compared with about 16% at Texas Instruments Inc. and Hewlett-Packard Co. Nevertheless, the company is finally showing signs of stress. "When half your people have been here a year or less," admits one Intel vice-president, "the speed and quality of decision-making can suffer."

Fervent search

From its start in 1968 on a $2 million stake assembled by venture capitalist Arthur Rock, Intel has been guided by a management triumvirate that seemed perfectly balanced for success in high technology. Robert N. Noyce, co-inventor of the integrated circuit and former general manager of the pioneering Semiconductor Div. of Fairchild Camera & Instrument Corp., recruited Intel's initial team and served as its chief spokesman with the financial community. Moore, who headed Fairchild's research operation during its most innovative years, set Intel's technology focus. And Andrew S. Grove, a scientist turned management ramrod, provided day-to-day production efficiency.

But Noyce, now vice-chairman, currently spends only about half of his time on Intel affairs. "Running a large corporation by the numbers is less than satisfying to me," he explains. And Grove and Moore, too, increasingly see their roles changing as Intel expands and technology races into new frontiers. "I'm qualified to sit in a meeting of people who are working their way to a decision and guide the process," says Grove, who became president a year ago, "but I can't tell people what to develop."

To meet the challenges presented both by its growth and its move into computer systems, Intel has embarked on a fervent search for new management ideas and structures that will keep it positioned at the forefront of technology in the 1980s. Except for such essentials as finance and law, Intel has never bothered with staff positions. But after consulting with such experts as Jay R.Galbraith, a specialist in corporate structures currently on leave from the Wharton School at the University of Pennsylvania, the company recently installed Senior Vice-President Leslie L.Vadasz as director of a corporate strategic staff. This gives him oversight over a variety of planning activities that previously had been played by ear.

Intel is not the only semiconductor company planning a major commitment in computer systems. Texas Instruments, the industry leader, has been talking for years about the software problems that its customers face and has even copyrighted the term "Solid-State Software" for the application program sit sells in form for its home computer and programmable calculators. And National Semiconductor Corp. has jumped into the systems business by acquiring mainframe-computer marketing operation of troubled Intel Corp.

But so far, at least, neither TI nor National has mounted a major threat to Intel's dominance of the sophisticated microcomputer business. Says one executive of a company competing with all three: "TI is locked into an architecture [in the microminis that handle 16 bits of data at a time as do full-fledged minicomputers ] that is difficult to use and inefficient. And it's hard to see where National will get the money needed in this business with all the other things they are involved in."
 
'Never-never land'?

Intel has been stung, however, by micromini products from both Motorola Inc. and Zilog Inc., the Exxon Enterprises Inc. affiliate started five years ago by two former Intel engineers. Zilog's Z8000 chip arrived about 18 months after Intel's 8086 went on the market, but it has recently picked us some important customers. Italy's Olivetti, for example, chose the Z8000 for a new generation of small computers and electronic typewriters after deciding that it could not wait for the next generation of Intel processors. And Charles E. Thompson, director of world marketing for Motorola's Semiconductor Group, claims that his micromini "does very well whenever we get into a competition based on the architecture of announced machines." The Intel disclosures, he adds acidly, "get us into never-never land."

Not only is Intel apparently losing its early dominance of the 16-bit micromini field, but the entire move upward from the older, less powerful 8-bit chips first introduced in 1974 is being retarded by the enormous investment in programming required for such a jump. Intel shipped only 72,000 of its 8086 devices last year, according to estimates by Dataquest Inc., the Cupertino (Calif.) market research firm.

Because of its strength in 8-bit microprocessors, Intel accounted for fully 40% of the $820 million in microprocessors sold last year, according to estimates by Dean Witter Reynolds Inc. But the company has long since lost its lead in small, 4-bit microcontrollers to TI, and its own 8-bit machines are getting increased competition from "second-source" producers such as National Semiconductor, Advanced Micro Devices, and Japan's NEC, which all build a chip identical in performance. If Intel sticks to its policy of avoiding dog-eat-dog battles over commodity products, it must soon win a lot of new customers for its more sophisticated offerings.

Intel's current campaign of previewing a family of new microprocessor products that may not be complete until late 1981 is at least partly related to this need. "They are running a little scared," observes Daniel L. Klesken, a Dataquest analyst. "Customers have been waiting to look at the [Zilog] Z8000 and the [Motorola] 68000, and Intel was getting concerned that its next generation would get shut out if it didn't let customers know about it," Klesken says. Admits William H. Davidow, general manager of Intel's Microcomputer Systems Div.: "There was some effective competition that was attempting to delay decisions by raising questions about our direction."

While leaks about future products are common in the semiconductor industry, Intel decided to turn this one into a major marketing effort. Dubbed "Project Crush," the plan to reveal major products and the strategy behind them took shape late last year with the creation of a high-level task force. But Intel's Grove insists that the main reason for the disclosure has been the growing importance of microcomputer technology to Intel's customers. "The customer has to be ready to use what we introduce," he says. "When we are selling memory devices, all we have to provide is some basic specifications. But when a customer depends on you for an operating system, he has to know as much about it as you do."

Intel has not exactly given away the store. "It's interesting how much you can talk without saying anything," remarks one customer, after sitting through an Intel presentation. "They have given us almost no numbers." Still, Intel has said enough to assure its customers that some impressive new products are on the way:

* Specialized chip computers, called "coprocessors," that can boost the performance of its 8086 microprocessor tenfold in such jobs as solving big computational problems and controlling peripheral equipment.

* Two powerful new central processors, informally labeled "micromidi" and "micromaxi," which are compatible with the 8086 but apparently will provide enough power to compete effectively with the Zilog and Motorola products.

* A "micromainframe" -- probably more powerful, by far, than any chip computer marketed or announced -- that will provide many of the features of today's mainframe computers, including the ability to be linked in networks for higher performance.

Tying these new products together is Intel's strategy of providing more and more software in standardized packages. Intel will announce in April its first full-fledged operating system, the program that attends to such housekeeping chores in a computer system as coordinating processing tasks and managing peripheral devices. Major parts of the operating system will be offered on memory chips and eventually will be embedded in the central-processor chip itself.

Intel is being forced to provide more software tools by the skyrocketing cost of programming today's powerful microprocessors. Whereas a microprocessor application five years ago required only about one-half of a man-year of programming, the typical application this year will consume more than six man- years, estimates David House, general manager of Intel's microprocessor operation. "The sky's the limit for the applications of this technology," he says, "if you can get the final system done."

Intel's major competitors also are responding to the same customer problems and are contemplating major software efforts. But the cost of following Intel's lead will be high. "It will shakedown to a few guys who can afford to make that $50 million-to-$100 million investment" declares Michael L. Hackworth, senior vice-president at Signetics Corp., a California semiconductor maker that has not yet entered the 16 bit microprocessor arena. "We are seeking a strategy to put together a comparable effort," he says.

Even for an Intel, the move into software looks risky to some industry observers. "Intel's whole success [to date] is based on hardware break throughs," says one former manager currently working for a competitor. "Now they are talking about evolutionary steps, some of which are quite controversial."

In one important regard, Intel's selling task is much easier than it was a decade ago when the company was as unknown as the microprocessor and memory technologies that it was promoting. "There is no longer a credibility problem," says Intel's Moore. "If we say we're going to do something, people tend to believe us." Indeed, many customers have developed ties with Intel that are much closer than with other component makers. One leading maker of photo-typesetting equipment, for example, meets quarterly with top Intel executives not only to get information about future products but also to influence their design. "We have a much greater insight into their plans than with any other supplier," says an engineering manager with this Intel customer. But the company's rapid growth in the past few years has jeopardized its cozy customer relationships.

A shortage of production capacity last year, for example, led Intel to back away from supplying some commodity products. "That created significant aggravation with quite a few customers," says one customer, "and we had our share of trauma." After spending a total of $200million for capital additions in the last two years, Intel is finally bringing substancial new capacity into production. This should help ease the company's short range planning problems, which consultant Galbraith says center on the need to allocate capacity. "As they get more capacity," Galbraith predicts," these decisions won't have to be managed as tightly as they are now." And Intel plans to spend an additional $150million this year on property, plant, and equipment.

The management structure that Intel has evolved to make planning decisions is perhaps the ultimate version of so- called matrix management, which came out of the aerospace industry and which has been tried at several companies, most notably Dow Corning Corp. In its simplest form a matrix organization provides a two-boss system -- one for technology and the other for a market or product line. At Intel, workers may have dozens of bosses, depending on the problem at hand. Instead of staff specialists for purchasing, quality control, and the like, Intel has committees or "councils" -- there are 90 at the moment -- that make decisions and enforce standards in specialized fields. And these coordinating bodies are overlaid on a grid of 24 strategic business segments (SBS) that do product planning.
 
Strained style

Ideally, Intel's top management troika of Moore, Grove, and Noyce would like to see virtually all decisions handled at the council or SBS, level. But the flood of new employees and the proliferation of markets at Intel during the past few years have made this concept difficult to sustain. "We like this style," says Vice- President Jack C. Carsten, "but it becomes a big training burden to change people from the style that is typical inmost U.S. companies."

Intel has stepped up its training effort with a series of courses for new professionals, stressing the philosophy and culture of the company. But it has also backed away from its no-staff stance enough to create a full-fledged planning organization under Vadasz, one of its brightest line managers. "Our planning process is one reason we've been successful in the past," Vadasz says. "We have to make sure it evolves to handle increased complexity."

These management changes are coming at a time when Intel is in the midst of a massive decentralization of its physical facilities. With major operations now in Portland, Ore., and Phoenix, as well as at its home base in Santa Clara, the company is struggling to find new communications techniques that will tie the organization together.

At the same time, Intel has opened a design center in Israel and is planning another in Japan to try to tap new sources of engineering talent. And in Europe, where its sales have grown 70% a year for the past three years, Intel is reconsidering its decision to stay away from the kind of joint venture that its competitors have jumped into recently with European companies. "We're still seriously examining the possibility [of a joint venture]," says an Intel executive in Europe. "The market has not been wrapped up by any means."

What Intel hopes to do with all this tinkering is to keep its substantial lead in key areas of semiconductor -- and now computer-systems -- technology. From its beginning in 1968, Intel has stressed product breakthroughs. "By innovating," says analyst Rosen, "Intel is able to generate higher profit margins than it could by following, and this in turn allows the company to self-fund its growth."

Remarkably, Intel actually increased its cash position last year, despite its massive growth. And the company still has no long-term debt.

But competitors are putting more and more heat on the company. As Intel Senior Vice-President Edward L. Gelbach points out, an increasing number of semiconductor makers are taking aim at product areas that have been the company's strength. "People used to hesitate before copying us," he says. "Now, even if they hear we're working on something, they put a task force on it." Because of this, Gelbach estimates that Intel's lead time over its competitors on most products has been cut from as much as four years to a year or 18 months.

Amazingly, however, the company still has several "golden geese," as semi-conductor analyst James R. Berdell calls products that are confronted with little or no competition. The pretax margin "has to be in the 40% to 50% range on as many as 10 of Intel's memories and microprocessors," says Berdell, who watches the industry for San Francisco's Montgomery Securities.

Intel has yet to demonstrate that kind of magic outside its traditional market areas, however. Although it was an early entrant in telecommunications devices, that market has been slow to develop. Frederick S. Glynn III, a market researcher in the field, points out that "there are now more companies manufacturing [these chips] than there are major customers for them." Similarly, Intel's leap into magnetic bubble memories in 1977 has so far provided a disappointingly thin sales payoff. And Handel H. Jones, analyst for Gnostic Concepts Inc., of Menlo Park, Calif., questions whether the company is positioned well enough in semiconductor processes such as the so-called C/MOS, low-power technology that some experts predict will become a major product area in the 1980s. "Intel has no position in C/MOS now," Jones notes.

Now the company faces its biggest challenge yet as it moves onto the turf of established minicomputer, and even mainframe, producers. Although the company's 16-bit microprocessor rivals the power of many minicomputers on the market, Intel has been careful, so far, to avoid direct competition. Data-quest calculates, for example, that while Intel has 40% of the $80 million market for 8-bit, single-board computers, it has taken only about 5% of the equally large market for 16-bit board computers. "We're trying to establish our own base," says James P. Lally, general manager of Intel's development systems operation. "It makes no sense to try to attack the minicomputer market."

Toward $10 billion

Many observers, however, feel that the company will be unable to avoid a head-on battle as its technology spreads to more sophisticated systems. Noyce now talks confidently about changing the traditional concepts of computer architecture (page 98). And Norman S. Zimbel, an industry analyst at Arthur D. Little Inc., attaches great significance to Intel's acquisition in 1978 of MRI Systems Inc., a supplier of programs for data- base management. "Data-base software technology is a basic building block" for big computer systems, Zimbel notes.

Intel's boosters believe that the company will take the computer business by storm. "If they can pull it off," says one analyst, "this strategy will make them the $10 billion company they ought to be." By the mid-1980s, he predicts, the computer-systems field will be down to a handful of major competitors, including International Business Machines, Hewlett-Packard, Digital Equipment, TI -- and Intel. "If you look at where the talent is," this analyst says, "these guys will all be competing with each other."

Intel is making no such predictions. "We don't threaten our customers," maintains Grove. "But the boundaries between what we do and what they do are going to shift." Still, when a company with Intel's record bets so heavily on its next hand, the other players are bound to reassess theirs. Says Kidd at General Electric: "I'm glad we don't make computers."  

GRAPHIC: Cover Photo, THE MICROPROCESSOR CHAMP GAMBLES ON ANOTHER LEAP FORWARD, Illustration by Stanislaw Fernandes; Picture 1, Triumphant triumvirate: Noyce, Moore, and Grove have guide from obscurity; Graph 1, Intel's 10 years of textbook growth,Data: Intel; Montgomery Securities estimate, Michael Annibale -- BW; Graph 2, In integrated Circuit sales Intel's is only No. 4 . . ., Jerry Tortorella -- BW; Graph 3, . . . but its 10-year growth rate far outstrips the competition, Data: Dataquest, Jerry Tortorella -- BW; Picture 2, Processing silicon wafers: intel plans to put mainframes on handfuls of chips. Christopher Springman

Copyright 1980 McGraw-Hill, Inc.