Back in October 1982, T. J. Rodgers was startled by a telephone call from a venture capitalist. At the time, Rodgers was a top research and development executive at Advanced Micro Devices Inc. (AMD), a $350-million semiconductor manufacturer, and the caller -- Stanford Fingerhood -- wanted to ask Rodgers about two former subordinates who were planning to launch a new company. Rodger's response was quick: "If you're giving away money," he said, "why not give some to someone who can take it and run with it?"

Now it was Fingerhood's turn to be surprised. Was Rodgers himself thinking of starting a new company? he asked. The engineer thought for a moment. If his subordinates were branching out on their own, why couldn't he? "Well," Rodgers said, "I guess I'm in it now."

What he had jumped into was the latest wave of start-ups in the semiconductor business, a field that entrepreneurs just a few years earlier had surrendered to the American and Japanese manufacturing behemoths. But as semiconductors have been incorporated in an ever-increasing number of consumer, scientific, military, and industrial products, the emphasis has begun to shift away from mass-produced chips. Today, the market is demanding more short-run, customized, high-performance devices, providing the openings for start-ups in specialized niches that have the promise of growing into sizable industries of their own.

Entrepreneurs like Rodgers and his former subordinates, financed by an innovative mix of venture capital, bank loans, and corporate investments, have launched some 30 new semiconductor companies in the past three years. Even fast-paced Silicon Valley, which saw the birth of the semiconductor industry in the 1950s and the second wave of start-ups in the '60s, has been surprised by the pace of this third wave of semiconductor companies.

Within a week of Fingerhood's first call, he and Rodgers were talking venture strategy at Manhattan's 21 Club. The next day, the banker arranged a meeting with Benjamin Rosen, a former vice-president of Morgan Stanley & Co. and now a top venture capitalist in the electronics and computer industry. Impressed with Rodgers's marketing concepts, Rosen put Rodgers on a night plane to Dallas for a meeting with his partner, L. J. Sevin. Sevin himself is a wily veteran of the semiconductor wars of the mid-1970s, when he was chairman of Mostek Corp., one of the semiconductor industry's spectacular entrepreneurial successes.

On the plane, Rodgers, better known for engineering accomplishments at Stanford University, where he earned his doctorate degree, and at AMD than as a marketing whiz, took out his blunt #2 pencil and began to scratch out on a pad of his engineer's graph paper a rough business strategy for the company that would become Cypress Semiconductor Corp.

The next day -- only 48 hours after his first face-to-face meeting with Fingerhood -- Rodgers laid out his rudimentary plan in Sevin's North Dallas office. Sevin studied the figures, paying particular attention to the all-important return on investment. Sevin remained skeptical and told him only that his plan was "not as bad as I thought it would be," sending Rodgers home not at all convinced that he had won Sevin over.

But after a month of phone calls, questions, and relentless nit-picking, Sevin made a surprise visit to Santa Clara, Calif., called Rodgers, and told him to come to see him as soon as possible. Rodgers hustled over and Sevin made him an offer: He would pay Rodgers's salary and expenses while the entrepreneur-to-be was drawing up a formal business plan. Sevin said he would back the new venture -- if other investors could be found. If the others wouldn't come on board, Rodgers would be stuck without a job.

"What could I say?" the blond, 35-year-old Rodgers recalls in his cramped Santa Clara office. "I shook his hand and quit my job."

Within four months, Rodgers was ready to unveil an intricately detailed business plan, a market-driven, multiproduct strategy that some observers have compared to AMD's.

Not a technological superstar like Intel Corp., nor a manufacturing machine like National Semiconductor Corp., AMD, under chairman W. J. "Jerry" Sanders III, has emerged among the most profitable and successful of the major Silicon Valley companies by carefully adjusting its products and technology to market needs. This market-driven multiproduct strategy, many observers believe has allowed AMD to steer clear of ruinous head-on collisions with Japanese and U.S. juggernauts. And it is the perfect plan for surviving in the new semiconductor marketplace.

"AMD has provided the best model for marketing in our current situation," Rodgers believes. "Jerry Sanders is the supreme businessman. He isn't blurred by an engineer's vision. Everything follows the sales plan. Jerry doesn't make chips to see if they work, he does it to make money. If you keep that in mind all the time, and move fast enough, you can stay out of the way of the Japanese steamroller."

Working closely with Lowell Turriff, a former AMD marketing director and now Cypress vice-president of marketing and sales, Rodgers laid out a multiproduct strategy that will follow the growth markets and avoid key confrontations with Japanese and domestic giants.

Like most of the new chip-making startups, Cypress will focus on a segment of the market that the Japanese have yet to dominate, a promising new generation of high-speed, high-performance chips using the advanced "complementary metal-oxide semiconductor" (CMOS) process, which greatly reduces the operating power of integrated circuits. While Japanese companies have been the leaders in relatively low-speed CMOS chips, which are used extensively in their huge consumer electronics industry, Cypress's high-speed, high-performance CMOS semiconductors are targeted for the sophisticated computer and scientific instrumentation market, a field still dominated by such U.S. giants as Hewlett-Packard, Digital Equipment, and Data General.

Rodgers claims that these and other U.S. companies will provide Cypress with a lucrative customer base, pointing to estimates that show the market for Cypress's line of 20 or so different types of CMOS chips expanding from $1 billion to more than $2.6 billion between 1984 and 1987, a rate of growth better than twice the semiconductor industry average. Rodgers projects that Cypress's sales will grow to more than $3 million this year, mushrooming to more than $110 million by 1987.

These numbers, coupled with some hard-nosed salesmanship, played a crucial role in Rodgers's successful wooing of Sevin and Rosen, both of whom had openly expressed strong reluctance to investing in new semiconductor ventures.

Three years ago, in fact, Sevin swore off semiconductors. After a decade of brilliant success in integrated circuits, Sevin's Mostek Corp. was staggering against relentless Japanese competition. In September 1979, Mostek became the target of a takeover battle, and Sevin reluctantly cashed out of the business and sold his Carrolton, Tex.-based company to United Technologies Corp. for $345 million.

"I started seeing the handwriting on the wall back in 1975," the craggy-faced entrepreneur recalls. "In 1976, I started to scream about it, but by 1979 I stopped screaming and gave up. The war was over, and the Japanese had won."

Today, the 53-year-old Sevin, to his astonishment, finds himself back in the silicon trenches as chairman of Cypress.

What changed Sevin's mind -- and the minds of many entrepreneurs and venture capitalists -- has been the massive transformation of the semiconductor market. Customized and application-specific semiconductors have emerged as the industry's fastest-growing sector. Customized semiconductor chips, only 5% to 20% of the market in 1981, could constitute a sizable majority of the business by the late 1980s. Some industry people are predicting that worldwide total semiconductor consumption, which was slightly under $25 billion in 1983, will break the $100-billion mark by the 1990s. While the worldwide integrated circuit market is expected to expand about 20% annually during the next two to three years, the niches targeted by most start-ups -- ranging from linear circuits to the most advanced types of microprocessors -- could be enjoying growth rates as high as 100% a year.

Sevin and other industry experts believe that the shift toward customization and specialization currently gives such companies as Cypress the upper hand over both Japan's vertically integrated electronics combines and the dominant American suppliers.

"The game has changed suddenly," drawls the Louisiana-born Sevin. "The technologies are exploding and driving the markets here. When it was a question of selling chips by the bagful, it was playing to their strength. Now that there are so many new niches, it's become a marketing game, and that gives the smaller U.S. firms the advantage."

Sevin is not the only one who sees the potential of the niche marketing strategy. There has been a dramatic change in the availability of start-up capital, bolstering the prospects for these new companies. A flood of venture capital, augmented by substantial financial assistance from traditionally conservative major banks and corporations, is enabling new semiconductor ventures, such as Seeq Technology, Exel Microelectronics, Linear Technology, and LSI Logic, to compete with the entrenched companies both here and abroad (see"The Age of Alliances," page 68).

In launching Cypress, Sevin took Rodgers on calls to such old Sevin contacts as Capital Management's Don Valentine. Within weeks, Sevin had harvested some $7.5 million from six venture capital firms.

Perhaps even more significant, the new company boasted one of the most impressive boards of directors in the industry. It included Sevin; Capital Management's Pierre Lamond, a longtime top vice-president at National Semiconductor; and Kleiner Perkins Caufield & Byers's John Doerr, a former Intel microcomputer marketing manager. Together they supplied Cypress with nearly a half-century of semiconductor management expertise from some of the leading companies of the previous wave of semiconductor start-ups.

New companies, founded by hungry entrepreneurs, long have provided the innovative spark that has made the U.S. semiconductor industry the world's largest and most successful. At the industry's inception in the late 1950s, it was two relatively obscure companies, Texas Instruments Inc., and Fairchild Camera & Instrument Co.'s new. semiconductor division, that seized the technological initiative from more entrenched electronics companies, such as RCA Corp. and General Electric Co. A decade later, restless engineers and executives at Fairchild and their corporate peers started breaking off on their own, founding such companies as Intel and National Semiconductor. These companies, in turn, became the bastions of U.S. technological excellence and the industry's proud new leaders.

But the influx of fresh entrepreneurial blood slowed to a trickle in the late 1970s, as Japanese manufacturers began to bring their considerable manufacturing expertise to bear on the market. The Japanese changed the rules and the competition shifted from a battle for technological innovation to a scramble to become the low-cost producer.

Atsuyoshi Ouchi, senior executive vice-president of NEC Corp., has been in the forefront of Japan's aggressive and, to date, highly successful foray into the microelectronics industry, which increased its share of the world semiconductor market from 18% in 1978 to more than 26% in 1982.

As he observed the retreat of smaller U.S. producers during the successful campaigns over the 16-kilobyte memory chip (16K DRAM) and the 64K DRAM, the 42-year NEC veteran came to the conclusion that the U.S. industry could compete only by following the Japanese pattern in which five or six giant, vertically integrated electronics companies dominate the industry. Like many analysts on both sides of the Pacific, Ouchi saw semiconductors as an increasingly mature industry with little room for entrepreneurial players.

If nothing else, Ouchi predicted, the quickly rising cost of new semiconductor facilities, now upwards of $50 million for a manufacturing plant, would by itself all but eliminate the United States's many "boutique" producers, leaving only such behemoths as Motorola Inc. and Texas Instruments to slug it out with Japan by the end of the decade.

But today, relaxing in his office at NEC's squat Tokyo headquarters, the gray and balding Ouchi, 64, admits he might have been wrong about the U.S. semiconductor industry. The explosion of start-ups -- boutique producers par excellence -- has clearly upset his characteristically Japanese vision of a rigidly structured industry dominated by megacorporations.

"You cannot deny the role of small American firms because the semiconductor industry has changed," Ouchi admits, with some reluctance. "In the past, the number of products was small; now it is enormous. In that kind of market, it's very difficult for us to cover the customer's demands."

Although small Japanese high-technology companies have taken off in such areas as software and services, Ouchi believes Japan's conservative financial and business culture -- particularly the unwillingness of top engineers and managers to leave the giant Japanese companies -- will preclude the emergence of a group of "boutique" semiconductor companies in Japan. This means such corporations as NEC, with 69,000 employees worldwide and a staggering array of products, face the unenviable task of competing with dozens of small, tightly focused U.S. companies.

Some well-informed observers, such as the University of California at Berkeley's Michael Borrus, believe the structure of the Japanese semiconductor industry could ultimately undermine its determined attempts to dominate the world semiconductor industry as that industry evolves. In the past, Borrus points out, Japan's large, vertically integrated companies, such as NEC, Fujitsu, and Hitachi have simply copied carefully selected American chip designs and then applied superior Japanese manufacturing technology to them.

But in today's marketing environment, such an approach, Borrus believes, may not prove to be nearly as effective. With the ever-increasing proliferation of new technologies and markets, each served by a wide range of suppliers, Japanese planners will have a more and more difficult time deciding precisely which products and markets to target.

One of the first to spot the opportunities in the changing semiconductor market was Wilf Corrigan, president and chairman of LSI Logic Corp., a Milpitas, Calif.-based company that is betting on the rapidly expanding market for semicustom "gate array" chips, which use standard blocks of unconnected circuit elements that can be customized quickly and relatively cheaply. A 20-year industry veteran, the British-born chemical engineer worked at Motorola Semiconductor Products Division in Phoenix before moving to Fairchild Camera & Instrument. As Fairchild's chief executive officer from 1974 to 1979, he watched the company begin to lose its technical edge to such newcomers as Intel and National Semiconductor. Finally, fed up with the onerous responsibility of administering a company with about 30,000 employees, Corrigan quit to look for new horizons.

By the fall of 1980, he found his opportunity. As large-scale semiconductor chips (those capable of holding more than 10,000 transistors per chip) were being employed in process control, instrumentation, and other complex functions, manufacturers were seeking ways to customize the circuits to fill their particular needs. Yet, although the market for such customized chips was expected to expand to more than $1.4 billion by 1985, most large U.S. and Japanese semiconductor companies disdained this generally low-volume, highly design-intensive business. Large volume orders, such as those for electronic games or personal computers, seemed to promise bigger profits. Corrigan built a company whose primary business would be to reduce the time and expense of customization.

Moving with his customary combination of energy and bravado, Corrigan rushed to fill his new-found niche. Working closely with former Intel executive Rob Walker and two top managers from Synertek Inc., a division of Honeywell Information Systems Inc., Corrigan recruited some of the nation's top designers, software programmers, and marketers.

"Why should they join? The answer is simple. Equity and challenge," Corrigan explains, as he races around his modern headquarters. "The bright people in this business are always looking for a big chance, for something new. People like being superstars. IQ likes being with IQ. That gets diluted in a big company. Here we let that top 20% loose and show what they can do on their own."

Using his own prodigious contacts, Corrigan raised more than $16 million in two rounds of venture capital financing and another $11 million in equipment lease-lines. Buying state-of-the-art CAD tools, he then turned his skill-rich employees loose on special projects, helping design prototype chips for such respected companies as Burroughs, Wang Laboratories, Hughes Aircraft, and Johnson Controls.

The design-shop business helped LSI Logic secure an initial customer base in its first two years of operation, during which the company lost more than $5 million. But early last year, as the economy began improving, demand soared for LSI's services, and over the first nine months of 1983, the company netted some $8.2 million on sales of $21.7 million.

With its bottom line rapidly improving, the company unleashed a thunderously successful public offering that netted LSI more than $152 million in May 1985. Flush with capital, LSI Logic is now putting the finishing touches on elaborate new fabrication facilities in Milpitas and Santa Clara. Building on its design base, the company is poised to reach Corrigan's goal of becoming a major, full-fledged manufacturer of semicustom circuits -- and, perhaps, the first giant to emerge from the current crop of start-ups.

In the semiconductor business, even a modest niche can provide the basis for large-scale growth. The market for the latest generation of EPROMs (erasable programmable read-only memories) and other programmable memories, targeted by such start-ups as Exel Microelectronics Inc. and Seeq Technology Corp., is expected to expand by $500 million annually over the next two years. As Kleiner's John Doerr puts it: "These companies are not just going after niche markets, they're going after niche markets that are exploding."

No market shift as dramatic as the one taking place in the semiconductor business occurs without some displacement and some unhappy people. But the controversy over the semiconductor industry's third wave of start-ups is tinged with some irony, since some of the most fiery criticism has come from the late-'60s-vintage entrepreneurs.

Such industry gurus as Intel president Andrew Grove have openly attacked the start-ups for siphoning talent and resources off from the top integrated-circuit companies. Grove, for instance, has publicly labeled the new start-ups as "gold-rush outfits" that he says are severely weakening the U.S. industry's ability to withstand the competition from NEC, Hitachi Ltd., and other giant Japanese companies. Venture capitalists, crucial to the development of high-tech start-ups, have been accused of playing the role of Darth Vader, luring innocent young engineers and managers to the "dark side" of entrepreneurial capitalism.

"You know, we don't put bags on their heads and drag them out of the companies," says Valentine, who was one of the early pioneers at National Semiconductor and has since helped launch four new semiconductor ventures. "People leave for the same reasons they always did. Nothing has changed. My friends at Intel just don't want anyone to do what they did. Was it right in 1969 or 1970? Is it right now? To claim national interest around the corporate flag is bullshit."

The new entrepreneurs certainly are persistent in recruiting their management teams. Once, one of the people most sought by Cypress, a brilliant, 29-year-old Intel circuit designer named Bruce Bateman, came back from a two-hour meeting with Intel president Andrew Grove convinced he would turn Rodgers down. But Rodgers continued to hound the designer, saying he didn't want to "ruin" Bateman's career by letting him pass up his great opportunity at Cypress. Ultimately, Rodgers got his man.

"At this company, we accept only two answers," Rodgers jokes. "Yes, or we'll talk about it some more."

In any event, Valentine and his counterparts doubt that the grand old men at Intel or National Semiconductor will be able to stop the flood of new companies. As long as there are markets and commercially viable technologies left untapped, there remains a strong incentive for entrepreneurs to leave established companies for the dream of building their own Intels.

"This is an industry that lives off new ideas, and big firms simply don't have the capacity to come up with them. Do you expect them from an RCA, GM, or General Electric?" LSI Logic's Corrigan asks bluntly. "When you start a company, you're on the frontier. You look at the markets as a whole new world. You can't afford to look back."