Special Report

The Microprocessor

Chip makers grow into systems builders

Business Week

March 19, 1979

The semiconductor industry created the microprocessor, and the powerful device is rapidly turning into the tail that wags the dog. The microprocessor is beginning to change the very structure of semiconductor companies and it now is becoming clear that a manufacturer's survival in this fiercely competitive industry could depend largely on how successful it is in this new business.

In fact, as semiconductor processing, systems design, and software development now converge, a new industry is taking shape. Building on the success of the first, relatively simple microprocessors, which emerged last year as truly high-volume components, major semi-conductor makers are pouring tens of millions of dollars each into a new generation of computers-on-a-chip. Like minicomputers, the new devices are capable of handling data in 16-bit word lengths, addressing capacious stores of program memory, and supporting advanced software languages.

One by one, the semiconductor companies are discovering that these advanced microprocessors cannot be treated as just another family of components. Instead, they must be regarded as full-blown computer systems, requiring new approaches to product definition and design, new engineering and marketing skills, new business strategies, and, in many cases, separate organizations. "The user has always had the problem of figuring out what to do with our components," says Bernard H. List, vice-president of Texas Instruments Inc. "Now for the first time in the history of our industry, we have a systems problem."

At the same time, the new microprocessors must be manufactured using the most advanced semiconductor processes if they are to achieve the combination of low price and high performance that will open up major new markets. "We had to create a melting pot of technology that hadn't existed in a single company before," says Federico Faggin, who founded Zilog Inc. four years ago in hopes of creating the prototype microcomputer company.

The importance of picking the right strategy in microcomputers is borne out by the rapidly growing role of microprocessors in shaping the competitive balance of the semiconductor industry. In 1975 only eight U.S. companies were selling microcomputer components, according to Creative Strategies Inc., a market research firm, and their sales accounted for just 3% of industry revenues. Last year, Dataquest Inc., another market researcher, counted more than 20 producers of microprocessors and their associated circuits, with sales totaling $376 million, of 5% of total semiconductor sales. By 1983, according to Dataquest, microcomputer components will be a $1.5 billion business accounting for at least 11% of industry sales.

Strength in this technology already is paying off for Intel Corp. (page 42 G), which used a dominant position in medium-power, 8-bit processors to triple sales and earnings in the last three years and to push ahead of Fairchild Camera & Instrument Corp. to become the fourth largest U.S. supplier of integrated circuits. Microcomputer components and systems accounted for more than a third of Intel's $400 million in sales last year and probably generated another third in memory sales, according to Daniel L. Klesken, Dataquest analyst. With a continuing boost from such sales, Intel's total volume is likely to reach $1 billion as early as 1981 or 1982, figures Benjamin M. Rosen, industry analyst at Morgan Stanley & Co.

Big gains in microprocessor sales also helped such major companies as Texas Instruments (TI), Rockwell International, and Motorola strengthen their positions in integrated circuits, while relatively flat sales in microprocessors contributed to slower semiconductor growth for American Microsystems, Fairchild, Signetics, and several others.

A computer for $1

Signs are multiplying now that success in microprocessors will soon be crucial to survival in the semiconductor industry. "That's where the technology is going," suggests Glenn E. Penisten, president of American Microsystems Inc. "As we move toward [chips containing more than 100,000 functions], we quickly get to a single-chip solution to our customers' problems."

In terms of pure volume, most of the action last year was in the lower-powered, 4-bit microprocessors, used mostly in relatively simple control functions. As prices dipped close to $1 for very large quantities -- and to an average of no more than $2 -- the 4-bit devices turned up in toys, games, appliances, and a host of other applications once the province of springs and gears. "At $1 and $2, we're fully competitive with mechanical solutions," says Thomas P. Harper, a marketing manager at National Semiconductor Corp.

Shipments of 4-bit units quadrupled last year to more than 16 million, or nearly two-thirds of all microprocessors sold, Dataquest estimates. Texas Instruments alone sold 9.4 million 4-bit chips, 70% of them for use in toys and games. But prices fell so low, and sales of associated peripheral and memory devices were so meager, that industry revenues in this market came to only $36 million.

Of much more interest to semiconductor manufacturers is the more powerful 8-bit device, which accounts for two-thirds of microprocessor-related revenues. With the glaring exception of TI, which dominates the 4-bit arena but missed the 8-bit opportunity almost completely, every major producer has taken a serious run at this market. Not only are prices higher than in 4-bit processors but the 8-bit product also generates substantial additional "drag" business in memories and peripheral chips.

The 8-bit microprocessor, with its enormous requirement for customer support and software development, has determined the ground rules for the coming battle at the 16-bit level. "It boils down to being a resource game," says William H. Davidow, general manager of Intel's Microcomputer Systems Div. "A lot of our competitors have been ineffective because they didn't have the dollars to do the total package."

In Davidow's view, Intel leads in the 8-bit race because it built strength early in all the key areas, ranging from a large family of components to software and program-development systems. With all these elements available, customers could start designing Intel's 8080 processor into their products immediately. And later rivals, many boasting superior performance, never overcame that advantage. "We had a better part," claims an executive of Signetics Corp., which brought out a microprocessor of its own a year after the Intel chip was in production, "but the 8080 already was designed into the end products."

Because the 8-bit market is becoming so large, there is room to correct mistakes and to try new strategies. Motorola Inc., for example, was not only a full year late with its 8-bit 6800 processor but suffered a massive production foul-up caused by its decision to move microprocessor manufacturing to Austin, Tex., from Phoenix, where its semiconductor group is headquartered. "For a year and a half," says Alfred J. Stein, Motorola vice-president, "we brought out no new peripheral circuits to support the product." Now, with the Austin facility running smoothly and a broad range of supporting devices available, the 6800 is taking off. Motorola's shipments doubled last year to an estimated 850,000 units.

Another turnaround is under way at Rockwell International Corp., which entered the 8-bit market in 1975 without the programming support to back up its product. "Our processor was highly efficient but too difficult for customers to program," admits Robert E. Anslow, director of business development for microelectronic devices. Two years later, Rockwell joined forces with MOS Technology Inc., a small company that was pushing a new 8-bit processor. Rockwell quickly introduced a development system and a high-level language software package for the processor, and sales of the device tripled last year. "Before, we had been mostly concerned with the efficiency of the device," says Anslow. "Now we are addressing the needs of the microprocessor user."

The effort to address customer needs also has required basic changes in marketing strategies, which for most semiconductor makers involve a heavy reliance on distributors. "Most distributors can service a market," says one Intel executive, "but they need a lot of help from the factory to develop a market." So most chip vendors have had to create their own sophisticated field organizations to deal directly with big customers. And the problem is compounded by the advent of powerful, 16-bit microcomputers, with their far bigger requirements for software support. "It takes a radically different sell," says Wilfred J. Corrigan, Fairchild's chairman.

The semiconductor makers also have started taking on the task of assembling processor chips, supporting circuits, and memories into complete microcomputers. For the most part, assembly stops at the circuit-board level, but some companies are taking a further step by packaging the boards as finished computer systems. "We are driven into the systems market by the need to understand the application of our products," says Faggin of Zilog.

If the early generations of microprocessor products have forced the semiconductor companies to look more like their customers, the next round promises to push many of them into direct competition with makers of microcomputers, electronic instruments, and a host of other products that now can be reduced to a handful of components.

So far there have been more failures than successes in the semiconductor industry's attempts at vertical integration into consumer products. But the successes have been big enough -- Texas Instruments' calculators and low-cost watches are the outstanding examples -- that new efforts are inevitable. "As each new industry is approached by large-scale integration," notes Edgar A. Sack, senior vice-president of General Instrument Corp., "they ask themselves, 'How long will it be before Texas Instruments puts me out of business?'"

Legal battles already have flared up between the minicomputer and semiconductor industries. Fairchild has produced a microcomputer able to execute the instructions of the Nova minicomputer made by Data General Corp. "It's not important to have access to Data General's software," claims Vice-President Thomas A. Longo of Fairchild, "but there's a body of capability out there that's extremely important." Data General has sued, charging misappropriation of trade secrets, and Fairchild has countersued, trying to invalidate the computer maker's program license agreements. Other suits surround National Semiconductor's development of a minicomputer that emulates the PDP-11 made by Digital Equipment Corp.

Many semiconductor industry executives are puzzled by such battles. By their standards, the current minicomputer market is not big enough to be worth fighting for.

"We're not interested in knocking heads with DEC," says Robert N. Noyce, Intel's chairman. "If they sell 50,000 units a year that's a minor market for us." The microprocessor has been successful, Noyce points out, because its low cost has opened up entirely new applications for computing power. "Our high-margin position depends on continuously introducing new products in new markets," he adds.

The problem of finding new markets is complicated by the very power and complexity of the latest microcomputer generation. Even deciding what features to incorporate in the new products has become a major headache, notes List of Texas Instruments. "You don't have to be too bright to figure out what the next new transistor or memory will be," he says. "It's denser and faster. But there's no simplistic guide for the next new microprocessor."

This new challenge is typified by the "analog microprocessor" that Intel expects to market this year. This remarkable device is customized to handle signal processing and must spark new product designs in that field if it is to succeed. A similar problem faces the entire industry as its powerful new families of microprocessors come to market. "The 8086 is the largest single project this company has ever undertaken," says Intel's Noyce. "it will have a niche in all of our traditional markets, but it won't justify our development expense if that's all there will be. We want to see the applications explode."

The leading makers
of microprocessors
1978 shipments
(thousands
Company of units)
4-bit microprocessors
Texas Instruments 9,400
National Semiconductor 2,535
Rockwell International 2,475
Nippon Electric 1,600
8-bit microprocessors
Intel 1,661
Fairchild Camera & Instrument 951
Motorola 950
National Semiconductor 750
Synertek 710
Mostek 670
Rockwell International 660
Zilog 540
Advanced Micro Devices 445
General Instrument 425
16-bit microprocessors
Texas Instruments 182
National Semiconductor 91
General Instrument 60
Intel 31

Data: Dataquest Inc.

GRAPHIC: Picture, Zilog's Faggin: His company had to create "a new melting pot of technology." Liane Enkelis; Graph, The zooming growth of microcomputers, Gerard Kunkel -- BW

Copyright 1979 McGraw-Hill, Inc