Special Report

Computers: The New Look

The Mainframe Era Is Fading -- And The Micro Is Taking Command

John W. Verity and Geoff Lewis in New York, with bureau reports
Business Week

November 30, 1987

For 30 years the computer industry has been the ultimate growth business. Every year produced some breakthrough that created new jobs, spawned new companies, and changed the way the world did business. A parade of computer stocks in the go-go years of the 1960s produced billions of dollars in shareholder wealth. In the 1980s startups launched by entrepreneurs who were just kids in the 1960s floated to market on oceans of venture capital. And why not? It seemed to be a law of nature that computer sales had to grow 15% or more annually. In a good year that figure could double. And when the economy sagged, computer makers seemed recession-proof. ''We thought this industry was always in its infancy,'' says William J. Caffery, an analyst at Gartner Group Inc. in Stamford, Conn.

Not any more. The worst slump in the industry's history--an unexpected and severe slowdown in sales that began in late 1984 and continued until early this year--changed all that. Today the slump is over, and most computer makers are enjoying a healthy rebound. With fewer employees and better cost controls, many are posting record earnings. And until the October market collapse, computer shares once again commanded dizzying multiples in the stock market. Computer industry revenues will be up 17% this year over 1986, says InfoCorp.


But the long-term picture is not quite so bright. The prospect is that in the future, computer makers may be hard pressed to match pre-slump growth rates. This year's percentage gain is deceptively bright, coming as it does after the depressed results of 1985 and 1986--and it will soon give way to annual sales increases of about 8% in 1988, according to executives at International Business Machines Corp. Such lower-than-usual gains could continue through the early 1990s, industry-watchers predict, and become the norm for a more mature computer industry.

The slump itself now appears to have been a turning point--a sign that a profound change is taking place in the computer industry. Analysts say that the business stalled in 1985 not so much because of a drop in real demand for computers but because a fundamental shift was taking place in the way businesses buy and use them. ''A lot of the reasons given for the slump have been wrong,'' maintains George F. Colony, president of Forrester Research Inc., a consulting firm in Cambridge, Mass., that tracks computer use in large corporations. ''I've said all along that what was really happening is that the industry is in the midst of a major transition.''

That gut-wrenching change is a shift away from the monolithic back-office mainframe to inexpensive small computers scattered throughout an organization and providing computing power to individuals and small groups. While the move to such ''distributed processing'' has been under way for years, in the mid-1980s it achieved critical mass--largely because of the flood of 15 million personal computers into U. S. offices. In 1984 revenues from sales of microcomputers surpassed those of mainframes for the first time. There's no doubt that mainframe sales will continue to grow from the current level of $22 billion a year, producing big profits (page 121). By 1991, however, the smallest systems will be outselling the largest by 65% a year. And their sales will continue to grow faster.


With the balance of power moving to small systems that didn't even exist 10 years ago, the mainframe's 30-year reign as the dominant force in computing is drawing to a close. ''Our policy is to minimize the number of large, centralized computers,'' reports Lorn Buxton, director of operations research for Kendall Co., a Boston-based subsidiary of Colgate-Palmolive Co. At Westinghouse Electric Corp., the biggest chunk of this year's estimated $70 million computer budget has been spent on networks of departmental minicomputers and microcomputers. ''We're not abandoning mainframes,'' says David M. Edison, director of corporate information and communication systems. But he adds: ''We're seeing a surge in networking PCs and engineering workstations and a consolidation of mainframes.''

That, in turn, is altering the economics of competition in the computer industry. Even to achieve an 8% annual increase in sales, the industry will have to crank out significantly more computers each year as the average selling price per system declines. According to market researcher InfoCorp, in 1982 computer makers around the world shipped 5.8 million machines with an average price of $8,250. By 1992, when they will ship a total of 38 million units, the average price will be $4,264. Worse, the small systems deliver slim gross margins in relative terms--40% to 50%, vs. 70% and more on the largest mainframes. So, no matter how strong overall demand becomes, ''we're not going back to the glory days,'' says Kidder, Peabody & Co. computer analyst William D. Easterbrook.

In the post-mainframe era, hardware itself will tend to become a commodity that customers can buy in inexpensive chunks and install as needed. As a result, profits from hardware sales are 'The people who can really manage the networks will be the most successful companies' likely to shrink as a share of the total. But for the computer makers that come up with the software to tie small computers into a coherent information system, the profits should be substantial. These ''dramatic shifts in the business,'' in the view of International Data Corp. Vice-President William F. Zachmann, may spell opportunity for the nimblest young companies and danger for some of the industry's entrenched leaders.

Zachmann's alarming conclusion is that before IBM can adjust fully to the new conditions, it will suffer at least one money-losing quarter. That's likely, he says, because IBM, with its tradition of not laying off employees, can't reduce its overhead quickly enough to break even in the face of slower growth and lower margins. IBM Chairman John F. Akers disagrees. His company has already trimmed down to the point where it can ''turn on a dime,'' he says.

Zachmann's gloomy views may be extreme, but the varying records of computer makers during the slump and the recent recovery may at least bear out the basis for his theory. IBM, which counts on mainframe systems for about 55% of its revenues and almost 40% of its earnings, has been among the hardest hit, and it continues to suffer. Two years of sluggish mainframe and mini sales led to a 27% earnings slide in 1986. Until this year's September quarter the $51 billion company saw its earnings fall for five straight quarters--its weakest showing since the Depression. By contrast, Digital Equipment Corp., whose business is concentrated on networks of small and mid-range computers, has prospered as never before. In the fiscal year ended June 30 its earnings climbed 84%, to $1.1 billion, on $9.3 billion in sales.


But the real stars of this year's turnaround have been personal computer makers. For its fiscal year ended Sept. 25, Apple Computer Inc. gained 40% in both sales and earnings. Compaq Computer Corp., IBM's chief rival in office PCs, tripled earnings in the first nine months of 1987, as sales leaped 80%, to $792 million. And even though they have not been enough to substantially boost IBM's earnings, Big Blue's personal computer sales have been setting records all year, according to Entry Systems Div. President William C. Lowe. Other standouts in the current market include companies such as Sun Microsystems Inc., which makes workstations--high-powered personal computers for engineers. What's behind the upheaval in the industry? The answer is technology itself: Microprocessor chips used in personal computers have become vastly more powerful. Microchips have always been cheap compared with the more complex electronic circuits used in larger computers. But they have also been underpowered, limiting microcomputers to fairly mundane tasks such as word processing. Now a new generation of microchips such as Intel Corp.'s 80386 and Motorola Inc.'s 68020 can crunch information in 32-bit chunks--twice as much as the fastest PCs consumed just a year ago and the same amount as DEC minicomputers costing $50,000 and up.

Machines using the new Intel and Motorola chips started hitting the market in 1986 and since then have shown astounding performance--compelling enough to catch the eye of even the most technology-jaded customers. Indeed, it's hard to ignore a $5,000 desktop machine based on Intel's 80386 that runs as fast as many mainframes costing $500,000. Even IBM, which might be expected to think twice about broadcasting such a comparison, has boasted that its fastest 80386-based Personal System/2 has the raw processing speed of a 1975 370/168 mainframe costing $3.5 million.

Putting yesterday's mainframe on today's desktop turns the fundamental economic principle of the mainframe era on its head. The basic rule has been that processing power is dear and, therefore, must be shared by as many terminals as possible. A few years ago, when minicomputers and PCs began distributing relatively inexpensive computing power beyond the mainframe, IBM and other mainframe makers still argued that their high-powered behemoths could handle most tasks more efficiently. No longer. The new chips will make it cheaper to do all but the most complex jobs on microcomputers.

The new generation of PCs using the advanced chips is catching on faster than most experts predicted (page 118). Approximately 325,000 computers using the Intel 80386 will be shipped in 1987, and company officials have indicated that they expect to sell 2 million additional 80386s next year, mostly for use in personal computers. Combined with the sale of 2.2 million equally powerful Motorola chips, that adds up to enough raw computing power to handle something like 20 trillion instructions per second. In comparison, Computer Intelligence Corp. estimates that all the IBM and IBM-compatible mainframes in operation in the U. S. today have enough combined processing power to perform about 145 billion instructions per second.


Initially, many customers are using the power of the new micros to speed up the work they have always done on PCs. But as better software becomes available, customers are finding that they can do a minicomputer or mainframe job on a micro.

The American Savings & Loan Assn. in Stockton, Calif., didn't have much choice. When its parent, Financial Corp. of America, lost $591 million in 1984, the company had to cut expenses in a hurry and chose to eliminate an entire mainframe data center. That forced Greg Salcedo, assistant vice-president for financial systems development, to look for ways to do more work on PCs. One project is a system to manage the trading of mortgage-backed securities. Consultants told Salcedo he would have to spend $500,000 to $3 million for a mainframe system to do this work. Instead, Salcedo does most of it with a network of PCs that costs less than $100,000. And, he claims, the desktop computers make it possible for the association to manage its $18 billion portfolio with one-third to one-half the traders typically needed for the job.

But the new shape of the computer industry reflects more than the better economics of micro-hardware. Companies are finding that it's much faster to program a new function on a micro than on a larger machine. In big corporations, it is not uncommon for programmers to have a two-year backlog of requests from various departments to add new programs to the mainframe system. And even if mainframe programmers can get to a request immediately, it can take months to complete a major new application. At American Savings, Salcedo says, data processing management said it would take up to 18 months to create a program that keeps track of loan-reserve requirements. Instead, Salcedo's developers did the work with Paradox, a PC data-base program. ''We started with about a day's worth of work and a single microcomputer,'' he boasts. Adding new functions to the program can take from a few hours to a few days.

That flexibility has created a special relationship between PCs and the people who use them. ''People used to talk of 'the system,' '' says Charles G. McCaig, senior vice-president for information services at Mutual Benefit Life Insurance Co. ''Now they say 'my system.' '' It wasn't always love at first sight, but by now ''PCs have become very comfortable,'' observes Jeffrey L. Ehrlich, manager of product technology at General Electric Co. in Bridgeport, Conn. By comparison, he adds, ''larger computers are still pretty strange.'' The PC generation has been weaned on a style of computing that is difficult to duplicate on mainframes and minicomputers. When a PC operator punches in a command on the keyboard, there's an instantaneous response on the screen because the PC's power isn't shared with other workers. As anyone who has waited for an answer from an automated teller machine knows, a mainframe slows down when too many people try to share it. Moreover, ''we realize that our employees like working with PC programs better than mainframe programs,'' says Michael Ippolito, manager of corporate financial systems at CPC International in Englewood Ciffs, N. J. To please CPC accountants--and to save money--the company now consolidates financial results from its affiliates and divisions on PCs instead of on a mainframe. After five years in Corporate America, the PC has changed the habits of computer buyers, too. Just as office workers have grown accustomed to the autonomy of a personal computer, the PC has liberated data processing managers. The hugely popular IBM PC and its horde of clones have created the closest thing to a generic computer the industry has ever seen. The same basic PC software will run on anything from a $600 Taiwanese clone to the latest $12,000 PC based on an 80386. For the first time consumers can make direct comparisons among different brands.

That puts unprecedented power in the hands of buyers. And in the long run this may mean more to computer makers than even the most powerful microchips. Before the proliferation of mass-produced micros, each manufacturer had its own proprietary hardware and software designs. Once a customer bought a particular computer, he was effectively ''locked in'' because the software and peripheral equipment he added to that system could not be transferred to another computer brand. It was a profitable marriage: Software and peripherals on large IBM machines have consistently delivered pretax margins of 70%.


While that will continue to be the case with mainframes, buyers can try to avoid getting trapped in the future by choosing small systems. A market dominated by microcomputers seems destined to operate with lower margins because the machines are built from nearly identical sets of chips and components. The differences among brands tend to blur, putting greater emphasis on price competition. Even Apple, whose proprietary software has prevented clones from driving down the prices of its machines, has gross margins of only 52%. That's better than the 40% that other microcomputer makers live with, but it still reflects the effects of commodity components on microcomputer margins. IBM's experience with the original PC shows that once a commodity mentality takes hold, even the most powerful brand name in the industry loses its clout. As clones of its PC flooded the market, IBM slashed prices--and still its share of the PC market fell, from 71% in 1984 to 34% last year, according to Dataquest. Its pretax margins on the personal computer line were only 43% in 1986, estimates Mark D. Stahlman of Sanford C. Bernstein & Co.

As a result, IBM Chairman Akers declared a year ago that his company would withdraw from such a ''commodity business,'' and with the introduction of the Personal System/2 line in April, he tried to make his PC more proprietary. That's been done in the more powerful PS/2 models with a unique socket for plugging in the circuit cards that customers add for communications, extra memory, and other functions. IBM is encouraging outsiders to design boards to fit the socket. But to stop them from cloning the PS/2 computer itself, IBM claims the socket, which it calls ''microchannel,'' is proprietary and has refused to license it.

Now there are signs that the market is resisting IBM's move. Many customers are reluctant to commit themselves to a computer that might leave them at the mercy of a single supplier. Despite IBM's claims about the performance benefits of its microchannel, the majority of the 1 million PS/2 computers sold since April have been less powerful models that don't use it. Richard A. Shaffer, editor of Computer Letter, says that as few as 250,000 microchannel computers have been installed by customers, though Lowe claims that as of September the majority of PS/2s leaving IBM plants are microchannel machines. Robert Pickering, executive vice-president for technology at First City Bancorporation in Houston, sums up the sentiment of many computer buyers: ''If we start to experience problems with IBM's proprietary attitudes, I'll buy more from Compaq.''


While it slows down the clones, IBM also wants to speed up its own product cycle. By the time the cloners come up with a response to the PS/2, probably in mid-1988, IBM will be moving ahead with PS/2 refinements. That way the PS/2 won't be the sitting target that the PC became. ''We plan to push the technology as hard and as fast as we can,'' Akers recently told securities analysts. ''We will improve the PS/2 every year in function and cost.'' If those improvements don't offer substantial benefits over this year's model, however, the plan could backfire. ''All of us are more skeptical,'' says Pickering. ''If they were selling cars, you'd be replacing your car every year and a half.''

The enormous benefit demonstrated by the standardized PC design has inspired computer buyers to push for standards throughout the industry. With the backing of U. S. government agencies and European computer buyers, American Telephone & Telegraph Co.'s Unix operating system software--the program that rules the basic functions of a computer--is gaining acceptance as a standard for mid-range computers. Unix has been adapted to run on dozens of brands of computers, making it relatively easy to transfer software written for one machine to another. This not only gives buyers of mid-range computers increasing leverage over their hardware suppliers but it may also help smooth communications among different brands.

Communications standards themselves are a top priority for customers. General Motors Corp., with help from numerous computer suppliers, is working out standard ways to interconnect machines in the factory. Several industry and government groups are trying to define data networks that could accommodate all types and brands of computers. And a European-U. S. coalition of manufacturers called X/Open is promoting an elaborate set of software standards designed to make customers' programs easily portable between many machines. ''Suddenly the entire industry has gotten religion about standards,'' notes Norman Weizer, senior consultant with Arthur D. Little Inc.

Computer makers, of course, have mixed emotions about standards. If the PC experience is a guide, standards broaden the market but reduce margins. Because they have invested billions of dollars to develop proprietary hardware and software, both IBM and DEC resist standards that would open their products to more direct competition--and eliminate the inherent profits.

Standards don't always mean low profits, however. Some companies have been nimble enough to make standards work to their advantage. In the PC business, Compaq has stayed ahead by sticking with the standard and adding a few technological twists such as faster memory chips and better graphics. Sun Microsystems has enthusiastically promoted Unix and networking standards for its workstations and has managed 49% gross margins. Bernard J. Lacroute, Sun's executive vice-president, says ''there is plenty of room to innovate cleverly within the standard.'' He likens it to the automobile business: All cars offer the same foot-pedals, but companies such as BMW and Mercedes have succeeded with innovations in the machinery that those pedals control. ''We think Sun is like that,'' he says.


Still, even for a fast-moving young company such as Sun, keeping ahead in hardware technology is a struggle. Right now Sun is gambling that it can distinguish itself by using a proprietary microchip. But, notes Lacroute, in time the basis for competition will shift away from basic hardware. ''The people who can really manage the network will be the most successful companies,'' he says. ''I'm not sure we've reached that point yet, but it's certainly coming.'' Indeed, with hardware becoming less lucrative, the next great battle in computers may be over networks--and the extremely complicated software links that make them function. Customers increasingly see better networks as strategic tools to gain competitive advantages. As a result, ''the thing that sits on the desk is unimportant compared to the data and the network behind it,'' says Claude A. Thomas, DEC vice-president for financial industry marketing. DEC has succeeded, he maintains, because its networks and computers extend information to more workers, rather than simply speeding up work such as accounting. ''The back-office stuff was done long ago,'' says Thomas. ''It's not a growth industry.''

But there is growth in selling networks--particularly in service industries such as finance and insurance. There, delivering information to salespeople, analysts, and others who deal with customers is vital. Manufacturing companies have similar needs. At Westinghouse, for example, Edison says that his networks of micros have helped speed the response to customer requests for products and services. That's a key advantage for the $10.7 billion company, which has recently shed marginal product lines to become more competitive.


DEC and other network suppliers have prospered as customers have grown weary of waiting for IBM to deliver effective ways for its various incompatible computers to share information. Arthur Simonian, director of information systems at United Technologies Corp.'s Pratt & Whitney division, connected several early PCs to his IBM mainframes but says, ''We didn't get much out of them.'' Missing was an easy way to get large chunks of data from the mainframe's central files to a micro for analysis.

IBM has ''made lots of promises of better communications with the PS/2,'' Simonian adds, but he won't buy many until he sees how IBM solves the communications problem--and at what price. ''I'm not going to get thousands of PS/2s out there and then be offered better communications for $1,500 each. We're not in business to transfer all our money to IBM,'' he says. ''We're trying to hold open the option to use another vendor if we have to.''

Ironically, solving the networking problem may also solve IBM's mainframe problem. The challenge for IBM is to transform the mainframe from a back-office number-cruncher into the hub of corporate networks. Once that happens, all the PCs in an organization can start calling on the mainframe for the latest information. The result, says William J. Filip, IBM's director of business plans, will be a ''turnpike effect.'' Just as car sales multiplied once interstate highways were built, computer use will grow when networking capabilities improve.

If the turnpike theory is correct, computers will finally begin to live up to their promise of becoming strategic tools for businesses. The power of microcomputers will make individual computing jobs inexpensive and convenient. The mainframes will survive--and perhaps even thrive--as the nerve centers of information systems.

Graph: A SLUMP NO MORE DATA: COMMERCE DEPT. LAUREL DAUNIS MUTUAL BENEFIT LIFE'S ACCOUNTING DEPARTMENT: THERE ARE NOW 15 MILLION PCs IN U.S. OFFICES KEN KERBS Photograph: BERNARD LACROUTE Executive Vice-President Sun Microsystems 'We're not abandoning mainframes,' but 'we see a surge in networking PCs' PHOTOGRAPH BY ROBERT HOLMGREN Photograph: DAVID M. EDISON Director, Corporate Information & Communications Westinghouse Electric PHOTOGRAPH BY JIM JUDKIS Photograph: ARTHUR SIMONIAN Director, Information Services Pratt & Whitney 'We're not in business to transfer all our money to IBM. We're trying to hold open the option to use another vendor if we have to' PHOTOGRAPH BY GALE ZUCKER Graph: WITH MICROCOMPUTERS OUTRUNNING ALL OTHERS. . . DATA: IDC CHART BY LAUREL DAUNIS Graph:. . . THE INDUSTRY'S MARGINS ARE SHRINKING DATA: GARTNER GROUP CHART BY LAUREL DAUNIS Graph: THE MAINFRAME'S PROBLEM DATA: TECHNOLOGY NEWS OF AMERICA PAUL GRANGE

Copyright 1987 McGraw Hill, Inc.