Deconstruction the Computer Industry

As the monolithic mainframe gives way, the industry breaks into leaner, faster, smaller parts

By John W. Verity, with bureau reports
Business Week

November 23, 1992

It sure looks like an industry on the skids. The signs are everywhere and grow more painful every day: Worldwide leader IBM Corp. is shedding 40,000 workers this year, for a total of 100,000 since 1985. No. 2 Digital Equipment Corp. ousts its founder, after taking $3.1 billion in charges over two years to cut 18,000 jobs and vacate 165 facilities. Wang Laboratories Inc. files for Chapter 11 protection. France's Groupe Bull lays off 8,000 workers and closes 8 of 13 factories; Italy's Olivetti downsizes by 20%; Siemens Nixdorf plans to lose 6,000 workers. And the list goes on.

''Think,'' an industry byword in IBM's heyday, is giving way to ''Shrink.'' It's not just computers that are getting smaller, it's most of the companies that make them, too. They're deconstructing--shutting factories, cutting jobs, spinning off subsidiaries, farming out work, and slashing managements.


Even the younger generation is feeling the pinch. Relative kids such as Apple, Sun Microsystems, and Compaq are attacking costs, too. Apple Computer Inc. and Compaq Computer Corp. have laid off thousands to cut overhead and stay a step ahead of the hundreds of wannabes nipping at their market shares. Only in Japan, it seems, have companies avoided downsizing. But they may be next. Sales are slowing, and Fujitsu Ltd. reported a $160 million loss for the first half of fiscal 1992.

What's going on? Like all manufacturers, computer makers are under pressure to slash costs. But they also face a unique challenge. In their business, every 18 months, advances in microprocessor technology double the amount of computing power a dollar will buy. These powerful chips and standardized software have already turned PCs into a low-margin commodity, and PCs have become the industry's biggest part (chart, page 93). Now that computer makers are building large-scale systems--including machines that will surpass IBM's biggest mainframes--from the same cheap chips, the harsh economics of the PC are sweeping computerdom. ''The computer-systems business is rapidly evolving to become similar in structure to the PC business,'' says John Levinson, computer analyst at Goldman, Sachs & Co.

Already, the industry's net income has plunged from 6.5% of revenues in 1986 to 0.12% last year, excluding special charges, according to Standard & Poor's Compustat Services Inc. That's billions in earnings up in smoke--a figure reflected in the dismal performance of computer stocks (chart, page 93). And the forecast is for more of the same, as powerful microcomputers gobble up more of the market. ''We never plan to see margins come up again,'' says Peter Bonfield, chairman of Britain's ICL PLC.


So the business is a disaster, right? Not at all. Although falling prices have slowed revenue growth, from nearly 20% annually five years ago to the low single digits, demand for computer gear remains strong. Lower prices are inviting more customers. And there's plenty of new technology on the drawing boards--pen-based PCs, handheld ''digital assistants,'' and massively parallel supercomputers--to create demand well into the new century.

But before then, the old industry has to deconstruct. With gross margins of 40% and falling, the vertically integrated companies that thrived on margins as high as 70% must do some painful soul-searching. They now have to choose at which points along the ''value chain'' they can most profitably apply their skills and resources. They may write software, build parts, or make complete systems. They may sell computers built by others. They may team up with partners. Or they may just help customers choose and install computers to solve specific problems. But even the mightiest, it seems, can no longer do it all.

The pattern for the industry's future structure has already been set--again, by the microprocessor revolution. In the past decade, the industry has splintered into an array of specialty companies. Each focuses on a different part of the value chain: chips, disks, distribution, data-base software, customer service, and so on. And, as a whole, they are proving more efficient than old-line, integrated makers. Again, there's an analogy to what's happening in hardware itself: Computing power has fragmented, breaking out of the monolithic mainframe and spreading to flexible networks of powerful, desktop machines.

Big Blue and other giants grew up as vertically integrated organizations, a la General Motors, because that's how companies worked then--and because there was little choice. They had to build their own chips, circuit boards, disk drives, terminals, printers, tape drives, and even the boxes the stuff went into. They wrote software, too, and fielded teams of salespeople, consultants, and technicians. Tens of thousands of employees, dozens of plants, and huge investments in research and development were involved.

But the PC changed all that. In 1981, when IBM chose an Intel Corp. microchip and Microsoft Corp. software for the IBM PC, it inadvertently sowed the seeds of its own deconstruction. Because anybody could buy the same parts, everybody got into the business. From Taipei to Tampa, from Delhi to Dublin, an infrastructure of suppliers sprang up to feed the right bits and pieces to thousands of assemblers.

With the PC market saturated with players, gross margins hurtled below 30%. As long as the big companies had a healthy business in ''big iron''--minis and mainframes--this didn't hurt bottom lines much. But those core businesses have slowed. And now, sooner than DEC or IBM expected, big machines are threatened with extinction by low-cost, micro-based alternatives. ''It has unfolded beyond our wildest imagination,'' says Gilbert P. Williamson, president of NCR Corp., acquired by American Telephone & Telegraph Co. in 1991. NCR's big thrust these days is building machines using multiple microchips that can do transaction-processing work for a retailer--but for a fraction of what it would cost with a mainframe.


So the falling tide is lowering all boats. McKinsey estimates the companies that in 1986 built and sold finished computer systems were capturing about 80% of the total profits being generated by computer sales. The reason: Older, high-margin systems from the big computer makers still dominated. These computers all had proprietary software that kept the customers locked in--and paying high prices.

By 1991, however, systems makers were getting just 20%. Why? Because the PC had cut out the fat--and not just by lowering costs. The PC and other ''open'' systems such as minis and workstations using Unix software made it possible for customers to choose from a wide range of machines that all ran the same programs. They turned the industry into a free market. And the market soon reallocated profits. The biggest chunk, 49%, simply stayed in the pockets of customers in the form of lower prices--an economic surplus that ''customers are unlikely to give back,'' notes Michael Nevens, a McKinsey principal.

That's not all. Computer makers also gave up profits to suppliers of components, software, and services, whose share of the pie rose from 20% in 1986 to 31% in 1991, according to McKinsey. In effect, the market said that chipmakers and software writers added more value than the folks cobbling those parts into systems and should be rewarded accordingly. That's why, as computer makers scramble, companies such as Intel and Microsoft, suppliers of the main chips and software, respectively, for the IBM PC market, are lapping up the gravy.

This explains the frenzy of deconstruction within the old vertical empires. Without fat profit margins on complete systems to mask inefficiencies, big companies can no longer afford in-house divisions--unless they're really competitive. And in many cases, the best way to improve performance of a division or factory--indeed, just to measure it--is to expose it more fully to market forces.

Clearly, that's the motive behind IBM Chairman John F. Akers' massive deconstruction project announced last December. He broke the sprawling company into 13 semiautonomous units. This past September, he created yet another, just for making and selling PCs. Each of those units is likely to subdivide until there's ''a whole host of little companies under the IBM umbrella,'' says President Jack D. Kuehler. The free-standing units will keep their own books, and before long, it will be apparent where IBM can be truly competitive--and where it might cut its losses. ''There will be some fallout and dislocations,'' says Kuehler. ''Some won't make it.'' But if all these Baby Blues are operating at peak performance, the sum of the parts will be greater than today's whole--to shareholders and customers.


The risk, of course, is that computer makers whose divisions can't cut it will quickly hollow out, shutting more plants and laying off more workers. That could boost short-term profits. But they would soon lose manufacturing and design knowhow, making them less able to bring innovations to market. And that would be the beginning of the end for a computer maker. ''You have to have something where you're the best,'' says Charles E. Exley Jr., retired chairman and CEO of NCR.

The companies that dig out ''best of breed'' capabilities in their organizations and free them to compete can hope to rebuild profits. Of course, most of the industry's best gigs are already taken: Even if you're great at operating software, Microsoft has that about sewn up. Ditto Intel in microprocessors; Conner, Quantum, and Seagate in disk drives; Andersen Consulting and Electronic Data Systems in systems integration; and so on.

The deconstructionists can take heart, though, from some examples of surprising innovation by their compatriots. Consider Hewlett-Packard Co.'s printer division, in Boise, Idaho: A decade ago, the minicomputer maker saw a chance to carve out a new niche in laser printers. It freed the division to pursue the new market--regardless of whether the printers would help sales of other HP machines. And now, HP has 43% of the $5.4 billion U.S. market.


There are already some promising possibilities among IBM's deconstructed units. Adstar, which used to design disk and tape drives primarily for use in IBM computers, is now mounting a massive effort to sell its drives on the open market, even for use in computers that will compete with IBM's. Another unit is beginning to sell memory chips and other components outside, so Big Blue will have a chance to prove what it has frequently asserted--that it has world-class capabilities in chipmaking. In fact, IBM may come to resemble its Japanese rivals. Those electronics giants have always built components for the open market as well as for internal use.

Success in chips and other components may be essential if big players want to avoid even more downsizing--to the hollow extreme of deconstruction epitomized by Dell Computer Corp. As Chairman Michael S. Dell acknowledges, his fast-growing company's value added--its ''core competency,'' as the management gurus would have it--is not computer technology at all, but distribution and marketing. Dell is set up to excel in two areas: handling orders and queries from 30,000 customers a day and providing an endless supply of new models and options. Dell does no real manufacturing--only final assembly and testing. And in many cases, it doesn't even design the machines it sells: Dell engineers, using input from a vast customer base, create specifications for new computers and hire subcontractors to build them.

The message is focus, as Sun Microsystems Inc.'s huge success attests. The leader of the technologically demanding workstation business builds no components itself; it relies on outside subcontractors to build the guts of its systems (box). That leaves it money to spend where it can really add value: the Solaris operating software and Sparc, its microprocessor design.

Of course, IBM and Digital can't just turn into Dell or Sun overnight--nor should they. Despite the battering they have sustained, the big players have enormous skill and unmatched assets, not least their tight relations with blue-chip customers and decades of experience creating comprehensive information systems for specific industries.

Solving complex information-handling problems for corporations is an obvious way of adding value--and replenishing profits--for old-line makers. IBM, for instance, is pursuing a variety of professional services, including writing custom software, outsourcing, and most recently, management consulting.

Before they can reap profits in new businesses, however, computer makers need to trim more costs in the old ones. That would have been a lot easier if they had started sooner. But in many cases, management assumed the slowdown in hardware profits was caused by the recession, not by fundamental changes in the industry. ''A lot of people in our industry, including myself early on, didn't see that clearly,'' says James A. Unruh, CEO of Unisys Corp.


One company that caught on early was Britain's ICL. Having survived a brush with extinction in 1981, ICL pared product lines and limited its marketing to a few key types of customers. More important, management punctured old assumptions. For example, when it bought a Finnish PC maker, Nokia Data Systems, instead of remaking the company in the ICL image, Bonfield encouraged his managers to pick up some fresh pointers from the younger organization (page 98). ''If you've got the same structure you've got now in two years,'' he says, ''you'll be out of business.''

Digital Equipment, on the other hand, has only just grasped the need for a massive overhaul. Under former CEO and founder Kenneth H. Olsen, the company split into 150 business units--none with bottom-line accountability. The new CEO, Robert B. Palmer, has moved to consolidate those, to just 10 units, each of which will have profit-and-loss responsibility. ''Clearly, we have some business units that can be more independent,'' Palmer says. One early candidate: the $775 million disk-drive business. But he has no plans to spin out subsidiaries as IBM and some other big players have done, he says.

Once regarded as a basket case, Unisys now looks like a winner in the deconstruction game. Pressed by the enormous debt that former Chairman W. Michael Blumenthal took on in the 1986 merger of Sperry and Burroughs that created the company, Unisys had to come to grips with reality early. Unruh says he saw that ''the economic model had changed and the cost structures of the past were obsolete.''

The resulting makeover has put Unisys well ahead of IBM and others in many respects. Unisys began selling computers that were designed and produced by other companies. It has pared its work force by 54% since 1986, in part through divestitures. It stopped making some PCs, began phasing out two entire lines of mainframes, shut several factories, and narrowed its marketing to focus almost entirely to four industries: government, financial services, telecommunications, and airlines. The idea, says Unruh, is to understand those industries extremely well and help customers there install complex information systems, which may not be 100% Unisys-built.


Unruh has had the company back in the black for four quarters, after three years of losses, and has pared debt to a more manageable 59% of capital--down from 65.7% at the peak. But it took measures that were painful in human terms (page 100). And, Unruh says, there can be no end to honing operations. ''Our phrase is 'a little restructuring every day.' ''

Every day and every way. As deconstruction ramps up at IBM, it's clear there can be no artificial boundaries--no sacred cows. That means the mainframe division no longer can squelch projects just because they might compete with its machines. Indeed, with the future of traditional mainframe technology dimming, IBM has several projects under way to create chip-based alternatives. Adstar, for instance, is teaming up with a small Silicon Valley company to produce a ''file server'' that could replace a mainframe as the hub in data networks. ''We want to jump in and use the new technologies,'' says Kuehler. ''I'm not trying to protect our old ways.''

One new way IBM is catching on is partnering. A major fact of life in the deconstructed computer industry is that more products are of ambiguous parentage--a blend of hardware and software from many sources. Indeed, even as it fragments, the industry is developing a thick web of strategic alliances, joint ventures, technology-licensing deals, and consortiums aimed at divvying up development costs, getting products to market quicker, and pooling technologies and skills. So IBM is working with Apple and Motorola Co. on a new generation of chips and software. Meanwhile, Apple works with Japan's Sharp Electronics Corp. on handheld computers, and IBM teams up with Toshiba Corp. on screens and memory.


Such globe-spanning alliances may be the only way for some computer makers to stay in the game. Take Groupe Bull. It sells NEC Corp. mainframes, IBM workstations, and personal computers from Zenith Data Systems, which it acquired in 1990. Bull has sold 4.7% of its equity to NEC and 5.7% to IBM and is developing new computers that IBM will sell, too. Also, it has joined Olivetti and Siemens Nixdorf in Trans European Information Systems, a venture that is bidding on pan-Europeanprojects.

''Do I worry about being hollowed out?'' asks Michel Bloch, president of Bull Systems Products, the company's main product-development group. ''Yes, it's a permanent concern. But everything in life is a trade-off. We had to think in terms of cost and time to market.''

Indeed, next to cost, the biggest motivation for deconstruction is speed. The vertically integrated companies often have trouble keeping up with the product cycles being set by tightly focused component makers. In PCs, product cycles have telescoped from two or three years in the late 1980s to as little as six months now. Miss a beat--as Compaq did in 1990 when other PC makers used Intel's i486 first--and profits vanish.


Compaq rebounded this year by slashing prices and overhead, which has put its stock back on Wall Street's buy lists--for the moment. But the outlook for computer stocks remains murky. Some analysts say it's no longer possible to forecast the long-term profitability of computer companies. Over the past several years, investors have been trampled after rosy turnaround plans failed, to be followed by more draconian cuts. Who's to say that more unpleasant surprises aren't in store? ''I don't detect any management team that has a real good plan that is several years forward-looking,'' says Barry Bosak of Smith Barney, Harris Upham & Co.

Steven M. Milunovich, computer analyst at Morgan Stanley & Co., says he has begun looking at companies altogether differently: Forget technology and focus on marketing clout. He likes Sun, because its identity as workstation leader is planted so firmly in the minds of computer buyers that it doesn't matter if Sun's technology isn't always ahead. Still, Milunovich warns, ''nothing's safe as a predictor for the long term. You constantly have to, more than in other industries, reassess the situation. The pieces are constantly moving.''

And what does the industry's fragmentation mean for overall U.S. competitiveness in computers? Judging from the PC industry, the U.S. can field a lineup of aggressive specialty firms that is unmatched anywhere. Unlocking the resources that have been hidden inside the vertically integrated monoliths can only make those companies better at what they do. Competition will goad them to innovate at a ferocious pace. New products may surface that might not have in earlier times.

That's the gleam of a bright future that computer makers can focus on. But in the here-and-now, there is still a painful transformation to struggle through. A massive industry in the midst of deconstructing itself is not a pretty sight.

Sun is Shining--With a Lot of Help from its Friends

Robert D. Hof in Mountain View, Calif.
Business Week

November 23, 1992

For years, Sun Microsystems Inc. had big problems shipping its workstations on time. Customers could wait weeks for orders. Things got so bad finally that Sun just gave up. It shut down its 18 distribution centers around the world and turned the work over to Federal Express Corp. and others. ''A few people thought we were insane,'' admits Robert J. Graham, vice-president for worldwide operations. But Sun subsequently set shipment records.

Even as the company, based in Mountain View, Calif., has blossomed into a $3.6 billion operation, its management has heeded a simple corporate commandment: Thou shalt not do thyself what others can do better.

By hiring others to do everything from circuit-board assembly to customer support, Sun can focus on the things it does best: designing microprocessors, writing software, and marketing workstations. And to help keep those core functions efficient, Sun has split them among several independent subsidiaries, each chartered to make a profit. ''That's the only way to run the computer business,'' says Chief Executive Scott G. McNealy. Indeed, it seems rivals such as IBM and Digital Equipment Corp. are beginning to agree.

Perhaps the best symbol of Sun's approach is how it handles its crown jewel, the Sparc microprocessor. When Sun created Sparc in 1987, it couldn't afford a chip factory, so it licensed Texas Instruments, Fujitsu, Cypress Semiconductor, and others to make them in competition with each other. Today, Sun could afford to make its own chips, but it still doesn't. Why? Because the $500 million that would require buys lots of engineering and marketing. Likewise, Sun avoids spending millions building and staffing factories by contracting work out to manufacturers such as Solectron Corp. While doubling shipments since 1990, Sun has cut its manufacturing work force by 10%, to just 2,000 people. And it has Eastman Kodak Co., Bell Atlantic Business Systems Services Inc., and others do the repairs on its machines in the field.


The results are dazzling. Sun's 12,800 employees generate $280,000 each in annual sales, topping all but Silicon Graphics Inc. in the workstation market and putting Sun in a different league from, say, IBM, which gets $188,000 per employee. Marvels Goldman, Sachs & Co. analyst John C. Levinson: ''They blow everybody away on almost every measure.''

Sun's don't-do-it-yourself approach has drawbacks. The SuperSparc microprocessor it developed with Texas Instruments Inc. was a year late, depressing profits. But later, it convened a task force of engineers, purchasing people, and others from Sun and TI to speed things up and get the new MicroSparc chip ready months ahead of schedule.

Still, it's unclear how far Sun can take its lean-and-mean model, especially as it moves into more complex computers. Recently, for example, Sun brought some circuit-board assembly in-house because outside contractors couldn't do it the way Sun needed. Sure, it cost more to do it in-house. But, McNealy says, it had to be done to get a new model out faster. Even in a deconstructed computer business, sometimes another rule applies: If you want something done right, do it yourself.

ICL is Still Standing--And Still Refocusing

Jonathan B. Levine in London
Business Week

November 23, 1992

Peter Bonfield is one of a rare breed. His peers at Siemens, Groupe Bull, Ing C. Olivetti, and Nixdorf Computer lost their jobs amid three years of savage industry restructuring. Yet after eight years on the job, the chairman of Britain's International Computers Ltd. PLC seems more secure than ever. Rarer still, ICL is one of the few mainframe makers outside Japan to remain consistently profitable in the past decade. ''I feel like the last of the Mohicans,'' Bonfield says.

What's his secret? Bad luck, of sorts. Facing bankruptcy in 1981, ICL had to turn to Japan's Fujitsu Ltd. for some mainframe processors. That led to selling an 80% stake to the Japanese powerhouse in 1990. But the crisis helped Bonfield refocus ICL's business substantially. Long before rivals felt the need for such moves, he had killed a whole line of mainframes, closed five factories, slashed employment by 32%, and revamped ICL's strategy.


The payoff: This year, more than 50% of ICL's $4 billion in sales will come from software and services, up from 30% in 1987. IBM reached only 36% in 1991. ICL also has targeted just a few industries--retailing, government, and financial services--has specialized in ''open'' networks using standardized software, and has built a successful desktop business. Says analyst Amit Chaudhuri at market researcher International Data Corp.: ''Management has turned the company on its head.''

Still, Bonfield expects the current shift to desktop computers to push gross margins down as much as 10 points by 1995. To stay ahead, he has been seeking ''catalysts for change.'' In the past 18 months, ICL has bought PC maker Nokia Data Communications Corp., a PC distributor, a facilities-management firm, and stakes worth $60 million in 12 software developers. It's also in a joint venture with Bell Atlantic Corp.'s Sorbus unit, which fixes computers.

Last year's $355 million purchase of Nokia not only boosted European market share but it also delivered shock treatment to ICL's struggling PC arm. ''We bent over backwards to make sure Nokia changed ICL, not the other way around,'' says PC Div. Director David A. Mills. With control over PCs moved to Nokia, ICL engineers were pressured to shave expenses and speed up development. Within three months, the team combined their two PC lines into one. This year, ICL says, shipments should grow 47%, with sales reaching about $620 million. If so, ICL will rank as Europe's No. 5 producer, according to market researcher Dataquest, up from No. 9 last year.

Likewise, Bell Atlantic's Sorbus is teaching ICL's service group how to win contracts on IBM and Digital machines. Says Bonfield: ''By integrating ourselves into these other companies, we're changing old mind-sets.''


No doubt the biggest influence will come from Fujitsu. The Japanese company bought control of ICL from STC PLC when the British phone-equipment maker decided it could no longer afford to support ICL. While Fujitsu manages ICL at arm's length, the companies work more closely than ever. Fujitsu sells ICL PCs and workstations in Asia while ICL pushes Fujitsu automatic-teller machines in the U.S. and supercomputers in Europe. ICL also is developing the optical-communications channels for Fujitsu's next generation of mainframes.

Eventually, subcontractors will probably handle manufacturing and even some engineering for ICL, says Jules Goddard, a visiting professor at London Business School and a consultant to ICL. That would leave ICL to focus on managing big software projects. Says Goddard: ''In 5 to 10 years, ICL will be a much simpler, focused business.'' And it won't be the only one.

When Benefits--and Jobs--are Deconstructed

Joseph Weber in Philadelphia
Business Week

November 23, 1992

A few years ago, when Walter Bucior needed cancer treatment, Unisys Corp. stood by him. As an assembly-line worker in the company's Flemington (N.J.) personal-computer plant, his benefit plan paid nearly all his medical bills. It also paid when his wife, Linda, needed back surgery, and her bills alone topped $100,000. No wonder Bucior says he never balked at going the extra mile for Unisys.

Now, Bucior, 40, is a victim of Unisys' struggle to slash costs. The company abruptly decided last December that it couldn't keep up as a manufacturer in the cutthroat PC market. So it closed down Flemington in May, leaving almost 700 workers out of a job.


Bucior is one of about 400 still jobless. Federal grants are paying his tuition for refrigeration-repair school, but he fears matching his $25,000 salary will be tough. His severance--a week's pay for each year of work--is long gone, and though his $308 weekly unemployment check helps pay the mortgage, he can't afford $600 a month for health insurance.

It's true that Unisys is winning praise from investors for its financial recovery--four profitable quarters after seven losing ones--but the pain is being borne by thousands. Since early 1991, Unisys has closed 7 of 15 factories and trimmed nearly 25,000 jobs.

For the jobless, the knowledge that Unisys is part of an industrywide struggle to deconstruct costly old organizations offers cold comfort. So far, new jobs have eluded most of the Flemington workers. Because they were highly trained and highly paid, many may end up working for far less, if they find work at all.

To be sure, Unisys did make some extra efforts. It helped workers qualify for federal trade-adjustment assistance, designed for industries that can show they were hurt by imports. The program pays for up to two years of training and supplements jobless benefits. But many of the workers are bitter that Unisys didn't find another manufacturer for the site, though the company says it tried. ''It's a marketable plant and a marketable work force,'' says Mary Frances Regan, an ex-Flemington worker who is leading efforts to find new employers.

So far, manufacturers have found the 205,000-square-foot site too big. Regan's group has hired a consulting firm to lure employers. Says Regan: ''I call it a human recycling project.''

Copyright 1992, by The McGraw-Hill Companies Inc. All rights reserved.