THE NEW COMPUTER REVOLUTION The successes and failures that have shaped this important industry hold lessons for every manager. Its future course will transform the way all business works.
By Stratford Sherman REPORTER ASSOCIATES Jane Furth, Alicia Hills Moore, and Joshua Mendes

(FORTUNE Magazine) – TWENTY YEARS after its invention, the microprocessor -- the computer-on-a- chip, a sliver of silicon not much bigger than your thumbnail, like the one on FORTUNE's cover -- has suddenly brought forth PCs as powerful as mainframe computers. In the process it has destroyed hundreds of thousands of jobs, unseated CEOs at world-renowned companies, reshuffled tens of billions of dollars of shareholder value, and set off convulsions that will wrack the 50,000 companies in computing -- and all their customers -- through the end of this century. As technological developments transformed the industry's economics, time- tested precepts became as obsolete as vacuum tubes. The changes overwhelmed such leading companies as IBM, Digital Equipment, and Compaq Computer. Apple Chairman John Sculley denies having any sense that his company is similarly threatened. Nevertheless, he says that while discussing his candidacy for CEO of IBM earlier this year, he put to rival Big Blue a proposition never publicly disclosed till now: To strengthen both companies, he said, Apple should merge with what he calls ''the best parts'' of IBM. But nothing came of the idea. Many other managers, being human, were simply too slow to respond. Don't sneer; it could happen to you. ''This is the information age,'' says Bill Gates, the billionaire Harvard dropout whose ever tightening vise-grip on PC software has made him the richest 37-year-old in America and the supreme rajah of computing. ''If your business has anything to do with information, you're in deep trouble.'' If he's right, that means almost everyone. Get ready for tougher competition than you've ever imagined. The pioneering experiences of the computing industry provide striking insights into the challenges managers in every business eventually must face. Computing is the bellwether, but it isn't unique: Capable global competitors are assaulting high profit margins, like those once possessed by IBM, wherever they're found. The way to win is to reexamine every cherished assumption and redesign entire organizations around customer needs. Anyone who hopes to escape the woozy inertia that disabled so many erstwhile giants of computing can benefit from the lessons -- as well as the products -- this industry provides. Five principles -- one familiar, the other four less so -- emerge from the startling economic changes in computing:

-- You can't say it often enough: Don't lose touch with the customer.

-- Even in a high-tech industry, management skills are more important than technology.

-- Today's successes often obscure the first signs of tomorrow's failure.

-- The company with the highest unit volume almost always wins.

-- The place to find unit volume is the bottom of the market, where low prices create new customers.

Just as Toyota moved up from cheap Tercels to outsell Mercedes in the U.S. luxury-car market, so the makers of commodity chips, PCs, and shrink-wrapped software have become today's computing giants. Marc Schulman, president of Technology Strategies, a consulting group, concludes: ''Every time there has been a battle between an inexpensive, low-end technology and a high-end technology, the low-end technology has won.'' More than any other agent of change, information technology is transforming the way business works. It is helping companies get leaner, smarter, closer to the customer. Those who seize the opportunities inherent in this revolution are capturing important competitive advantages. Those who lag behind are forced to scramble breathlessly in a race to catch up, or die. As advantage passes from hoary monoliths like IBM to relative youngsters like Intel and Microsoft, customers are in a quandary about where technology is leading: Is my $100 million investment in mainframes obsolete? Must all my software be rewritten to run on Unix? Which suppliers can I count on to survive the shakeout and light my path into the future? This story cuts through the confusion and draws out the lessons from the halting, often painful journey into the jungles of the information age. Start with a point Compaq Chairman Benjamin Rosen made back in the 1970s, when he was still a security analyst: He wrote that the best way to understand the future of computer technology is to imagine infinite increases in performance, for free. The computing industry will never quite achieve that magical feat, but it is finally close enough to have arrived at a historic pivot point: The cost and performance of hardware, as well as the power and ease of use of software, have all come together in a way that will fundamentally alter the way most people and businesses work. That change has swept through the industry in the past year or so, and it will soon transform virtually every other business known to man. Not long ago, the information that corporate executives needed for decision- making was locked up in the ''glass house,'' a roomful of mainframe computers where only a professional computer programmer could get it. Today low-priced innovations such as PCs, spreadsheet programs, Microsoft's relatively user-friendly Windows software, E-mail, and local area networks linking groups of desktop computers bring ever more of this information out where people need it. That enables corporations to flatten organizations, yoke together teams across the barriers of specialty, rank, and geography, and forge closer strategic relationships with customers and suppliers. Computers, particularly PCs linked into networks, play a dual part in business's search for faster, more flexible responses to the shifting marketplace. They have helped unleash that quest, but at the same time they have made the goal more attainable. ''Until recently, computers were simply automating the old manual processes we'd had before,'' says Les Alberthal, CEO of Electronic Data Systems, the $8.2-billion-a-year information technology services business that Ross Perot sold to GM in 1984. Computing is not yet the largest industry in the world -- at $360 billion a year, it runs well behind autos and oil -- but it has become the most important because of its power to transform the way people work. This polyglot collection of interdependent businesses includes software, semiconductor chips, components such as printers and disk drives, hardware from $1,000 desktops to $30 million supercomputers, and services -- notably systems ^ integration, the critical, ever-more-daunting task of making these complicated products work together to help people make money. (For a sampler of industry innovation, see ''Products That Make Markets.'') The business no longer depends much on funky Silicon Valley startups. Fifty- five companies with over $1 billion of sales each -- all but 13 of them American -- dominate the industry, collecting two-thirds of total revenues. Esther Dyson, whose Release 1.0 newsletter has long tracked industry trends, says wryly, ''All the guys I knew and loved when they were so young and cute have become the Establishment. Especially Mr. Gates.'' Indeed, the most visible result so far of the upheavals in computing has been the transfer of enormous wealth from one near monopolist, IBM, to two others: Microsoft, the software maker, and Intel, which produces most of the microprocessor chips that do the thinking in PCs. IBM's collapse chillingly demonstrates the high price of failure to change with the times: Though still the world's No. 1 producer of mainframes, minicomputers, and PCs, Big Blue has lost two-thirds of its market value -- over $70 billion -- since 1987. IBM has an excellent chance of rebounding, but it may never fully regain that lost value. WHILE IBM was declining, the combined market capitalization of Microsoft and Intel swelled by more than $35 billion. ''With all due respect to Microsoft and Intel, there's no substitute for being at the right place at the right time,'' says Intel's impish CEO Andrew Grove, who fled Hungary for the U.S. in 1956. ''That's not the whole story, but it's a lot of the story. I'm very grateful to IBM for making us that gift.'' When Big Blue introduced its personal computer in 1981 -- an epochal event, in retrospect -- IBM relied on Intel's microprocessor and Microsoft's DOS operating system, the fundamental software that manages traffic between the computer and application programs such as word processors or spreadsheets. As it turned out, chips and software today account for most of the value in computers.

Every step IBM, Microsoft, and Intel take rattles the teeth of all the other companies in computing. The fastest-changing, most brutally competitive business in the world, computing has fragmented into a dozen or so mini- industries. All are experiencing some sort of shakeout. As commodity pricing of hardware and, to a lesser degree, software dissolves fat profit margins, financial and management skills are replacing technology as the key determinant of success. Dell Computer CEO Michael Dell puts it briskly: ''Ideas are a commodity. Execution of them is not.'' Several major shifts are happening at once. Until the Eighties, the business was dominated by vertically integrated companies, particularly IBM and Digital, that made every element of their products, from silicon to software. Since their custom-designed hardware and software did not work with products made by anyone else, companies that bought these products became captives. For decades, these practices contributed to double-digit annual revenue growth and gross margins -- the revenue you get for a product less the cost of manufacturing it -- as high as 70% of sales. Microprocessors, always a more affordable source of processing power than mainframes, have finally caught up to the larger machines in speed. That is why they are making obsolete the technology used in most earlier mainframes and minicomputers -- and with it the old business model. Most new hardware and software is now designed to work with other computer companies' products. As a result, scores of suppliers can leap into the business to do battle over each segment, from chips to operating system software, from disk drives to networking programs. Since everyone from the PC buyer to the mainframe maker is increasingly free to shop for value, the competition in even the smallest niche has taken on such ferocity that only the toughest, most innovative competitors can survive. Gross margins of roughly 20% to 30% are becoming the norm, forcing companies to eliminate every penny of unnecessary cost. At the same time, unit volumes have soared: In mainframes, IBM's worldwide installed base amounts to 50,000 machines -- a lot, considering that one of these hunks of big iron can cost over $1 million, but tiny compared with the 135 million PCs whirring away on desktops and laptops around the world. Since 1988, PCs have brought the industry more revenue than mainframes. Standardization is transferring the profits in computing products from hardware to software and services (see ''Profits Shift'' chart). According to Peter Schavoir, IBM's head of strategic planning, hardware produced just under 50% of the industry's worldwide revenues last year, down from 65% ten years ago, and the percentage will continue to drop. The most valuable products are those -- like Intel's microprocessors and Microsoft's operating systems -- that so dominate their markets that other companies are forced to make their products work with them. Here's a report from the battlefront in each of the industry's main segments:

-- PERSONAL COMPUTERS. As powerful by some measures as the mainframes of just a few years ago, with plentiful software and those 135 million machines worldwide, PCs have become the industry incumbent. The most popular type, with three-quarters of the world market, are the IBM-compatibles, which rely on Intel's microprocessor design. Ferocious price wars have taken the fun -- and most of the profit -- out of making PCs. The response to this challenge has been heroic: For example, in under eight months starting late in 1991, Compaq Computer -- No. 2 behind IBM in IBM-compatible PCs -- transformed itself by cutting costs more than 30% and basically rethinking its entire business. Unit sales tripled. ''Corporate change on this scale is not hard once you decide the direction you want to take,'' says Daryl White, Compaq's chief financial officer. ''That's why it baffles me that some companies take so long to change. I guess they're not really committed.'' IBM has also dramatically reshaped its PC business and, along with Compaq and Dell, is gaining share. (The three now account for about 20% of the world market.) Second-tier clonemakers such as Tandy are abandoning PC making. Apple Computer, whose machines are not IBM-compatible and whose gross margins, at 40%, seem impressive but unsustainably high, is working hard to expand beyond its extremely profitable niche. One cruel lesson from the experience of IBM and Digital is that during times of transition, corporations can produce terrific numbers even as they are being hollowed out from inside. Suggestive of some anxiety about Apple's prospects as an independent company is CEO John Sculley's quiet attempt to merge Apple with IBM this spring. ''We had all the pieces to put together a very strong business,'' he says. ''The problem was that IBM stock continued to weaken, so to acquire Apple would have meant too much dilution. In my view they had a much greater chance of being successful with Apple on the IBM team.''

-- WORKSTATIONS. These high-powered desktop computers, sold mainly to engineers and scientists, have become a $9-billion-a-year business. Instead of running Intel microprocessors, most workstations use a newer, more powerful type called reduced instruction set computer (RISC) chips. IBM, Digital, Motorola, and other producers of these chips are betting that RISC-based computers will enable them to grab back some of the profits they have lost to Intel. If they fail, workstations could lose much of their appeal. Either way, terror will reign as the workstation business consolidates.

-- MINICOMPUTERS. Now that workstations, PCs, and laptops are so powerful, these machines -- in effect, smaller mainframes -- are hard-pressed to offer competitive performance at an attractive price. IBM and the widely admired Hewlett-Packard are still prospering in this arena, but Wang, Prime, Control Data, and many others have left the business or died altogether. Last year, amid $2.8 billion of losses, Digital abruptly fired its burly, independent- minded founder, Kenneth Olsen, now 67, and replaced him with a man who could hardly be more different: Robert Palmer has a diamond ring, a sleek haircut like Michael Douglas's in Wall Street -- and an uncommon willingness to face reality. He has cut $1 billion out of Digital's bloated manufacturing costs and nearly completed a sweeping reorganization. Says Palmer: ''We're reengineering our own enterprise to meet customer needs. We're not just talking about theory. We're doing it.''

-- MAINFRAMES. ''Profits are worse than awful,'' says E. Joseph Zemke, 52, CEO of Amdahl, one of many mainframe makers that are losing money. ''Computing is a utility now, and there's just too much capacity in the industry.'' IBM has 75% of the market. By the first quarter of 1993, IBM's mainframe prices were 30% lower than a year before, according to one estimate. The once-fearsome Japanese -- Fujitsu, NEC, and Hitachi -- are suffering too. Unisys has staged an impressive turnaround. NCR, owned by AT&T for the past two years, looks stronger than most, owing to its early decision to offer computers from mainframes on down based on microprocessor technology.

-- SOFTWARE. A battle among competing operating systems will decide the shape of this market for years to come. Microsoft's strategy is to extend its dominance in operating system software for PCs to the ''enterprise systems'' that connect whole businesses -- the powerful information-swapping networks that will redefine the way people work in the Nineties even more dramatically than the PC itself did in the Eighties. Microsoft's weapon: Windows NT, a cousin of its well-established Windows software, due by summer and touted as capable enough to drive the powerful computers at the hubs of networks. Microsoft's main opponent here is Novell, the leader in software that runs networks of PCs and workstations. Novell, taking on the world from Provo, Utah, is touting a powerful, 25-year-old operating system called Unix, variants of which run on most workstations. Sun, Hewlett-Packard, and most other producers of work-stations have allied themselves with Novell. In applications such as spreadsheets, word processors, and databases -- where most of the money in software is made -- Microsoft, the market leader in PC software, is taking advantage of unit volumes that dwarf its competitors' to cut prices brutally. It's a classic case of the guy with the most volume winning. Among the wounded: Lotus Development (of 1-2-3 spreadsheet fame) and Borland (Paradox database and Quattro Pro spreadsheet).

-- MICROPROCESSORS. This is where the second great battle to control the future of computing will be staged. Intel, which turned out 32 million IBM- type chips last year, seems almost unassailable. But that hasn't squelched the assailants: Producers of competing RISC microprocessors hope to win converts beyond their traditional workstation market. Digital Equipment is pushing its version of RISC, the Alpha chip, technologically the hottest performer. More formidable commercially may be the PowerPC, which IBM is hawking in partnership with Apple and chipmaker Motorola. This triad benefits from market clout and solid technology, but faces an uphill fight: Sales of all RISC computers last year amounted to just 500,000, less than 2% of Intel's volume. A similar gap exists between the amount of software written for RISC machines vs. Intel-based machines. It's hard to believe that RISC can ever catch up. The shift to microprocessor-based computing challenges customers every bit as much as producers. The buyers of computing products and services are increasingly anxious and frustrated as they search for ways to take advantage of low-cost but still relatively crude desktop equipment -- without at the same time forsaking the hard-won reliability of the mainframes and minis that still run the most important business software. Corporate information officers, CIOs for short, expect the transition to the desktop to proceed with deliberation over the next several years. But they're feeling intense pressure to make demonstrable strategic contributions by improving the way their companies work. An outstanding example: Wal-Mart's inventory-management / system, which integrates information from stores, warehouses, and suppliers. The dynamic exchange of data allows each of them to respond immediately to the buying behavior of customers. The system also helps the company tailor shipments to the particular needs of each of its 2,000 stores. One near the sea will require slightly different stock than another by an inland golf course. Says Wal-Mart systems expert Bob Martin: ''The idea is to run each store as if it's the only one you have.'' In The Business Value of Computers, Paul Strassmann, former vice president for strategic planning at Xerox and later the Pentagon's chief information officer, argues that corporate investment in computers -- which can range from 1% to 8% of annual revenues -- often adds to overhead without producing a return. ''If your investment isn't related to your mission, there's no return at all,'' he says crisply. The big payoffs come not from management systems but from new services that add value for customers. Computers create business opportunities in many ways. ''If you can track it, you can market it,'' says Charles Wang, CEO of Computer Associates, at $1.8 billion a year the largest independent supplier of application software for IBM mainframes. Consider MCI's Friends & Family program, which builds loyalty by offering customers lower rates on calls to specified phone numbers that MCI's computers automatically identify the moment the calls are placed. There's no human intermediary. Says David Lockhart, director of systems integration at Toys ''R'' Us: ''Companies like ours have to integrate new technologies and applications to be competitive.'' TO EXTRACT the lessons from the industry's experience, we need to look quickly at how computers work and how the market for them evolved. They're fundamentally stupid machines: Their main aptitude is for adding and subtracting ones and zeros. Software translates myriad minuscule tasks -- displaying the letter s on a screen, for example -- into the crude binary code that computers understand. The wonder of computer design is the speed with which processors perform simple calculations. A computer's processor works rather like a pump; the more powerful the pump, the more calculations it can handle. Perhaps the most common measure of this pumping capacity is millions of instructions per second, or MIPS. (An instruction is one line of software code, the amount a processor can handle at one time.) A typical modern mainframe carries a rating of 100 MIPS. By contrast, the 1981-era PC, based on the Intel 8086 microprocessor, clocked in at 0.3 MIPS. The vastness of that gulf helps explain why mainframe makers long dismissed PCs as trivial instead of recognizing and responding to the threat from microprocessors. At its simplest, any processor is a series of interconnected on-off switches. More switches mean more MIPS. During the 1960s, when IBM engineers created modern mainframes, no single chip could contain enough switches to handle all of a computer's needs. So the designers had to use lots of processor chips to get the job done: as many as six to run the central processing unit, perhaps 20 more to handle specialized tasks such as routing data to and from the magnetic-tape drives where the computer stores information. A microprocessor consolidates most of those functions onto a single chip, using a newer technology called CMOS, for complimentary metal oxide semiconductor. Until very recently, mainframe processors outperformed CMOS microprocessors. IBM and its competitors have been able to hang on to a 30-year-old mainframe design partly because their customers were stuck. Until the advent of the microprocessor, standardization was unthinkable. Since no parts were shared, the machines were incompatible -- that is, unable to run each others' software. This incompatibility gave rise to powerful brand loyalties, since customers often spend far more on software than on the computers themselves, and switching hardware would make entire libraries of expensive software obsolete. The cost of rewriting software in order to switch from mainframes to smaller machines can be almost inconceivably high. Says Max Hopper, American Airlines' computing czar: ''If we had to rush out and rewrite everything in one fell swoop, it could take hundreds of our programmers three to five years and cost hundreds of millions of dollars.'' Like most sensible CIOs, Hopper is planning to move off mainframes gradually instead of trying to do it all at once. These economics make corporate computer buyers a conservative breed. Steve Jobs, who made history as co-founder of Apple and now runs an exciting but chancy Unix software venture called Next, scoffs at the idea that change comes quickly in the computer business. ''I work in one of the slowest changing industries in the world,'' he says. ''The installed base of applications software is what governs the rate of change. Unless you can give customers a good reason to re-create a whole universe of software, the rate is glacial.'' What has helped keep customers happy for decades is the computing industry's unmatched ability to improve the performance of its products. ''The power curve'' is industry shorthand for what Compaq's Ben Rosen was talking about back in the 1970s -- the remarkable advances in design and manufacturing that have enabled computer companies to roughly double every 18 months the amount of computing power available for a given price. The power curve applies not only to processors. Memory chips, disk drives, laser printers, video displays, and other key hardware components also continue to improve price-performance ratios at stunning rates. Hard drives for personal computers, for instance, can now store 50 times the data that more costly models could a decade ago. The application of the power curve to microprocessors destroyed the status quo in computing. When they first appeared in the early 1970s the chips seemed little more than quaint toys, but Intel's just-introduced Pentium processor, successor to the ubiquitous 486, is rated at a walloping 100 MIPS, the same as a typical mainframe, on a chip that sells for less than $1,000. While mainframes still outperform even the best PCs in other ways -- the rate you can feed data into them, for instance -- the price-performance advantage of microprocessor-based computers is too overwhelming to ignore. So will microprocessors topple the great god Mainframe? Gordon Bell was the chief designer of Digital's hugely successful VAX minicomputers. A gruff and respected industry graybeard, he contends: ''The mainframe is a poorly designed dinosaur. IBM should have thrown it out long ago.'' Counters Nicholas Donofrio, who led the development of IBM's RS/6000 workstation and now runs its ailing mainframe unit: ''We'll be selling hardware based on this architecture through the end of the century.'' Donofrio is probably right. New and projected applications, such as huge digitized film libraries that will enable cable companies to offer movies on demand, will require very large computers of some sort. Eventually most of those large computers may be based on so-called massively parallel designs, in which arrays of hundreds of inexpensive microprocessors work together within a single machine in order to handle larger volumes of data than PCs can. Until then, IBM plans to extend the commercial life of its old-fashioned + mainframes by making them more affordable. Processors account for more than three-quarters of the machines' cost. IBM is belatedly designing CMOS microprocessors that can power its mainframes without requiring any software changes; by one authoritative estimate, that could drop the cost of a mainframe's processors as much as two-thirds, permitting substantial price cuts.

Here the iron law of unit volume exerts its grip once again: Low volumes in the mainframe business robbed manufacturers of the extraordinary benefits of the learning curve in semiconductors. Gordon Bell says that every doubling of volume in a semiconductor plant should lower the unit cost 10%. Because microprocessors and the PCs that use them have been inexpensive enough to attract large numbers of buyers from the start, that high volume of sales produced rapid increases in performance per dollar that brought in still more customers. In its rush to get a personal computer to market, IBM decided to rely on Intel's microprocessors and the DOS operating system, which Microsoft didn't create from scratch but bought from another Seattle company for $50,000. IBM made both Microsoft and Intel sign elaborately detailed contracts that, quite amazingly, left them completely free to sell to IBM's competitors. Those rivals quickly exploited the opening. In 1982, Compaq designed the first computer that achieved full compatibility with IBM PC software without infringing on any of IBM's intellectual property. That created the clone market at a stroke, transforming Intel's microprocessors and Microsoft's operating system software from niche products into standards that every American PC company except Apple would design its products to work with. That, in turn, knocked down the proprietary barriers to entry that had always protected entrenched computer makers. Zeos, Dell, Gateway 2000, and many others barged into the marketplace with little more than an 800 number and a few thousand dollars' worth of standard parts. IBM, accustomed to the fat margins of its mainframe business, maintained high prices that created an umbrella over the whole PC industry. Compaq undercut IBM just enough to get attention; the original clonemaker had 40% gross margins for years. Compaq's prices sheltered the like of Zeos and Dell, which sheltered Gateway, and so on. The newer entrants, living on much thinner margins, learned to operate far more efficiently than the like of IBM, Compaq, and Apple. Their low prices further expanded the market and spurred their own growth. Unit volume was at the bottom of the market. As the second-tier companies got richer, they kept improving their offerings until their quality and performance was virtually indistinguishable from that of computers priced 50% to 100% higher. The great economic shift that is now convulsing the entire computing industry began in the second quarter of 1991. That is when, as Ben Rosen puts it, ''we got hit in the head by a two-by-four.'' A large number of Compaq's customers suddenly rebelled, switching to less expensive brands and depressing Compaq's profits and market share. Rosen got the message. ''The old margins were artificially high,'' he says. ''That was a mistake, in retrospect. We allowed other competitors to come in with impunity.'' Rosen felt Compaq had to lower its prices -- and therefore its costs -- in a hurry. Many of the company's top executives, including CEO Rod Canion, disagreed. The board replaced Canion with Eckhard Pfeiffer, a German who had run the company's European operations. Recalls Pfeiffer: ''We had to overcome the denial before anything constructive could start.'' Once begun, the transformation of Compaq proceeded quickly. Cutting one- fifth of the work force was agonizing but clearly more desirable than the alternative -- no jobs at all. The biggest savings were in materials, which accounted for 80% of the company's manufacturing costs. Compaq's demanding specifications had forced suppliers to produce components that, while superb, no one but Compaq was willing to pay for. The simple decision to settle for more popular components resulted in dramatic cost cuts. TO CHANGE the culture, Pfeiffer began basing executives' pay on such criteria as increased product profitability and reduced cycle times, particularly in product development. In mid-1992, Compaq unveiled the result of these efforts: the new low-price ProLinea line of PCs. Backed by three-year warranties that attest to Compaq's confidence in their quality -- one year is typical in the industry -- the ProLineas are priced astonishingly close to Gateway's bargain machines. The competition had to respond. IBM, burned by its experience with mainframes, got the message first, quickly introducing its own lines of PCs offering more choices and increased performance at lower prices. The transition wasn't painless: IBM lost an estimated $1 billion on PCs last year, but the company says the business achieved profitability in the most recent quarter. Even Dell, always known for low prices, has been forced to trim its margins. Apple has stayed in the game by cutting prices on its most popular Macs, but its add-on components and more powerful machines are still overpriced by the new industry standards. Compaq CFO White sums up the lesson his company learned: ''There's no doubt that the best entry point for your competitors is the low end, so active cannibalization of your own product line is the watchword.'' The attempt of IBM's new CEO, Louis Gerstner, to lead what was once the world's preeminent company back to greatness promises to be the business- transformation story of our time. The challenge of realigning this company with the needs of its customers is enormous. At a cost of over $15 billion, IBM has shut down a quarter of its manufacturing plants and let go 100,000 employees, 25% of the peak total. The company is still losing money, and some knowledgeable IBM watchers suggest that Gerstner probably must cut another 100,000 people and close still more factories to get costs in line with the competition. Gerstner should be able to restore profitability more quickly and easily than most people expect. Cutting the 300,000-person work force by one-third, for instance, could add perhaps $4.5 billion to after-tax profits. To get a sense of what's possible, consider Lexmark, the $1.8-billion-a-year desktop printer and typewriter business that IBM spun off in 1991. Chairman Marvin Mann eliminated 40% of the work force, organized the business into four autonomous product units, and ensured employee accountability -- for instance, by compensating salespeople according to the profitability of the products they sell. In two years the business has generated enough cash to pay off $250 million of debt, and gross margins are up five percentage points. It doesn't pay to underestimate the capacity of a corporation for change. IBM, along with Digital, still has great opportunties as a service provider, helping companies stitch together complex systems. If it can solve some of its own problems, it could make big money applying its newfound wisdom to other companies' troubles. THE COMMON ailment of the corporations that got hit on the head by two-by- fours -- particularly IBM, Digital, and Compaq -- was distance from customers. They ended up making products customers didn't want. The lesson here, which applies to every industry, is that successful companies depend ! utterly on customer feedback. At winning outfits from GE to Wal-Mart, a primary goal is to create structures -- from flattened management to E-mail systems linking employees with customers and suppliers -- that increase that closeness. In the past year or so the general shape of computing's future has emerged from the mists. Neither the centralized mainframe nor the radically decentralized PC will provide tomorrow's model. Instead, we will all be linked into increasingly coherent networks that connect all varieties of computers, from big iron to laptops to the pocket devices of tomorrow. The buzzword is client server. The ''servers'' -- whether PCs or mainframes -- will act as efficient central repositories of data, holding anything from corporate files to videos and voice-mail messages. Our personal machines -- the ''clients'' -- will tap into these databases, grab the information we need, and then (we hope) help us use that information to make better, faster decisions. The changeover will probably be slower, more costly, and much more painful than optimists predict. Like the increases in their power, the capacity of computers to disappoint users is nearly infinite. One measure of the high cost of downsizing: Strassmann, the former Pentagon CIO, says that although plummeting equipment prices have shrunk the hardware share of some companies' information technology budgets to just 15%, the total budgets keep rising as companies move to client-server networks. Warns Amdahl CEO Zemke: ''Client server is an idea whose rhetoric is way ahead of the reality. Downsizing costs more than people think, and there are a lot of problems on networks, such as security, that we don't have on mainframes.'' But even Zemke, an IBM alumnus who sells traditional IBM- compatible mainframes, agrees that by the turn of the century client-server computing networks are likely to connect -- or ensnare -- us all (see ''Want a Network? Read This,'' page 76).

As networks become the norm, computers will change the way we live and work far more than they have to date. The computer illiterate will be at a severe disadvantage as companies increasingly organize into teams of people who work together electronically. The democratic spread of information will force corporations, sometimes against their will, to depend less on command-and- control hierarchies and more on values that guide independent contributors toward furthering the corporate good. That much sounds swell. The downside, as ; Intel's Andy Grove sees it, is that ''we'll all be able to work ourselves to death'' -- because ubiquitous computers mean that our work will always be with us. And our competitors will always be working too. For many corporations, client-server computing still seems more challenge than opportunity. Spurred by the need to stay ahead of competitors, companies sooner or later will rewrite virtually all their old mainframe programs to run on newer machines, despite total costs of countless billions. Smart outfits such as Corning, John Hancock, and Wackenhut Corp., the security company, have already begun. Having finally got past the idea of using computers simply to speed up old-fashioned ways of doing business, they are treating this transition as a welcome -- if often agonizing -- opportunity to reexamine the way they operate, designing new business systems that make the most of advances in both management thinking and computer design. The client-server breakthrough became possible only last year, when PCs based on Intel's 486 microprocessors started selling for less than $2,000. Even these computers work just well enough to make client-server networks feasible. Easy-to-use software often requires a lot of computer power. That's because software tricks such as the richly graphic menu displays in Apple PCs and Microsoft Windows depend on extra lines of program code that eat up MIPS. To excel in business -- no, even just to remain marginally competitive -- every corporation will have to find its path through the emerging client- server technology. The signposts are unclear. Almost every supplier claims to have the solution customers need. One of the remarkable aspects of the business is the way it first causes you horrific problems, and then takes your money to solve them. For CIOs generally, the most pressing question concerns software: Which operating system and networking program will provide the best platform for their client-server needs? The decision is crucial, since this software -- particularly the operating system -- determines the specifications to which existing applications must be rewritten. Most corporate buyers will seriously consider only three choices: Windows NT, Unix, and IBM's OS/2. All three are designed to do the heavy lifting required of network servers. Although several mainframe operating systems are actually better qualified, they aren't in the running. After decades of paying suckers' prices for proprietary systems, executives want so-called open systems based on standard designs. Says Barry Sullivan, marketing vice president at EDS: ''The natives are restless. To some degree client-server systems are an emotional buy -- managers want to be free of the hegemony of that glass house.'' Windows NT, whose much delayed release was due in May, is by far the leading candidate. Over the past nine years Bill Gates has focused relentlessly on expanding Microsoft's control of the software market through its dominance of operating system software; NT is his latest move. ''It's all about scale economics and market share,'' says Gates. ''When you're shipping a million units of Windows software a month, you can afford to spend $300 million a year improving it and still sell it at a low price.'' Recently engaged to be married, Gates has shed the odd mannerisms, like his constant rocking back and forth, that used to mark him as one of America's odder CEOs. Encountered recently over pepperoni pizza at a Seattle restaurant, he seemed relaxed, intellectually curious, and remarkably normal for a self-made billionaire. Microsoft introduced Windows, an extra layer of software between DOS and application programs, in 1985, but it suffered from severe defects and required more power than most PCs then provided. Instead of acknowledging defeat, Gates listened to customers' criticisms and learned from them. Windows 3.0, an improved version released five years after the original, has become a de facto standard despite its many remaining flaws. More than half of all PCs sold worldwide now come with Windows already installed by the manufacturer, and the percentage is growing. Windows NT is designed to capitalize on the huge popularity of Windows. Although the guts of the operating system are brand new, to users and to application programs Windows NT will look just like familiar old Windows. Novell, the $933-million-a-year producer of NetWare networking software, is the only competitor with the clout to challenge Microsoft. While Microsoft has incomparable grassroots reach to end users, Novell, whose software runs on 80% of network servers, has established strong relationships with the corporate information officers who will be plotting client-server strategy. Its networking products are widely regarded as superior to Microsoft's. Raymond Noorda, Novell's folksy CEO, can take advantage of the widespread enmity that Gates inspires within the computing industry. As Noorda puts it, ''We look at it like the Second World War: The good ole US of A tells the + dictators, 'You can't do it that way.' We're the good ole US of A.'' With workstation producers as its allies, Novell hopes to position a single, unified version of Unix, the operating system of choice for workstations, as the all-American alternative to evil NT. Created by AT&T, Unix has had a long time to work out its kinks and demonstrate its power. While customers who fear too much dependence on Microsoft may want to keep a second supplier of systems software in business, most buyers probably won't care whose software they're buying, as long as it works. And as Steve Jobs points out with characteristic hyperbole, there is a major flaw in Novell's Unix strategy: ''No one wants to run Unix,'' he says. It's hard to use. Indeed, for all the years Unix has had to win converts, all versions of Unix together sell only 1.3 million copies per year -- barely more than a tenth the volume of Windows. That may not be enough volume to finance an ongoing development program equal to Microsoft's. IBM, the also-ran in operating systems, recently introduced a new version of OS/2. Unless IBM can quickly increase the software's popularity, OS/2 may soon be forgotten. The poor sales performance of this excellent product, which has had no worthy direct competition for two years, is a testament to the business-destroying capabilities of IBM's monolithic sales force. Until recently, OS/2 had no dedicated sales team; IBM salesmen who hawked heavy iron hardly got excited about moving software priced under $100 a copy. THE OTHER looming battle affecting customer choices in the client-server market concerns microprocessors. Software must be adapted to each type of microprocessor it runs on; since adaptations are expensive, software companies and corporate CIOs have an incentive to stick with the most popular microprocessors. Intel's products are not the performance leaders -- RISC chips beat them. But they seem as safe a bet in this realm as Windows NT does in operating systems. Once again, volume is the key. Intel estimates that it will produce 40 million 486 chips this year, plus a relatively small but growing number of Pentiums. That volume makes it possible for Intel to invest $2.5 billion annually in R&D and new plant capacity while still earning enviable profits. For almost everyone else, the economics of chipmaking seem daunting. It costs nearly $1 billion to build a state-of-the-art microprocessor plant, and with each new generation of technology, the cost doubles. How then can Digital Equipment, which has produced just a few tens of thousands of its technologically impressive Alpha chips, possibly hope to make money at this game? Striking as he is as a corporate revolutionary, Palmer doesn't have a convincing answer. Since Alpha chips can be plugged into newer VAX minicomputers, they surely will help Digital milk revenues from its old customers for a few more years. But Alpha seems unlikely to win millions of converts in the outside world. Sun, although the volume leader in RISC workstations, is often tagged as a likely victim of the industry shakeout. Its Sparc microprocessors, which Sun designs but does not produce, are price-performance laggards. Sun CEO Scott McNealy, complains: ''Intel's and Microsoft's proprietary control has been so strong that they have slowed innovation. That's the price of having two companies own control of the computer language.'' Intel does face two challenges to its hegemony in microprocessors. The first comes from Advanced Micro Devices, which began shipping clones of Intel's lucrative 486 chips in April. At present AMD lacks the plant capacity to steal more than 5% of the IBM-compatible chip market from Intel -- a fleabite. But over time, competition from AMD will cut into Intel's profit margins. The second, more serious, challenge takes the form of that PowerPC RISC microprocessor. Apple's backing of this IBM-Motorola venture, moving to production, virtually guarantees annual unit sales of one million to two million PowerPC chips, enough to justify huge investments in product development and factories. IBM hopes to amortize that investment further by using the chip in a variety of its own products, from PCs on up. Although it is not the hottest RISC chip of all -- Digital's Alpha is -- PowerPC is more powerful than Intel's Pentium. Properly priced, and produced in high volume, it might gain a substantial second-place share of the microprocessor market. YOU'VE JUST BEEN in the trenches, looking through the eyes of the soldiers, sweaty and bloody, and listening to the cannon roar. Now imagine yourself soaring high above the battlefield. The war is over. Quiet descends. New boundaries have been drawn. Within a few years, customers may not perceive much difference between the rivals now so strenuously vying with one another for their favor. Buyer pressure is forcing suppliers to build bridges that connect competing operating systems and chips. Today Windows runs only on Intel-type chips, but the introductory version of Windows NT will work with two RISC microprocessors, Digital's Alpha and Silicon Graphics' MIPS. The great question is whether Andy Grove's good friend, Bill Gates, will someday sell a version of Windows NT for the PowerPC chip, which would increase the Power PC's ability to compete with Intel microprocessors. Odds are that Microsoft will feel compelled to make software that works with any very popular chip. Phillip Hester, the IBM executive responsible for PowerPC, says Big Blue is working on a whole range of so-called emulation software that will allow one operating system or chip to run programs written for another. A few emulation programs exist already; Sun and Apple are readying products similar to IBM's. This trickery uses up vast amounts of computer power, but the new microprocessors, both Intel's and RISC, offer power to spare. ''With that kind of product,'' says Hester, ''I could run standard Windows applications on a PowerPC as fast as on an Intel chip, and I could do the same thing with Mac software.'' If operating systems such as Apple's are to avoid becoming orphans, they must run popular Windows applications. Software developers want to piggyback on the success of Microsoft and Intel, and in practice that means conforming to Windows specifications. The inevitable arrival of ever cheaper, ever more powerful computers and easier, more useful software will continue to break down technology's once forbidding barriers. Before much longer, perhaps by the turn of the century, voice-recognition software will let people operate computers without ever touching a keyboard. When that day comes, the computing industry may finally expand deep into the home market. Half a century after the computer was invented, less than 3% of the 5.5 billion people on this planet own one of these annoying but marvelous machines. The enormous opportunities to enrich companies and to change our lives are just beginning to be tapped.



CHART: NOT AVAILABLE CREDIT: FORTUNE CHART/SOURCE: GARTNER GROUP YARDSTICK CAPTION: PROFITS SHIFT AWAY FROM HARDWARE As standardization and cutthroat competition transform high-tech computers into commodities, operating profits are migrating away from hardware -- especially money-losing mainframes -- toward software and services, where companies can add value with ideas.