(FORTUNE Magazine) – DESPITE his widely noted disavowal last summer of the need for a vision for IBM, Lou Gerstner has one. It's clear and specific. The company he sees will know how to seize more opportunities than it fumbles away. Though monolithic, it will be decentralized and far more efficient than the IBM of old. It will suck in cash as a service provider, helping companies stitch together complex systems. It will dominate dozens of market segments, from chips to disk drives, hand-held computers to mainframes, as a high-volume, low-cost manufacturer. And it should wrest leadership of the personal computer industry -- computerdom's hotbed of innovation -- from the upstarts that nearly monopolize it, Intel and Microsoft. Gerstner's most decisive step toward those goals has been to scrap the plan by departed CEO John Akers to split IBM into as many as 13 parts. In just seven months on the job, Gerstner, 51, has discovered compelling reasons, outside and inside IBM, to keep the business whole. The outside reason is the advice he's hearing from customers: ''Of the thousands of customers I've talked to, not one has argued that they want IBM to be broken up into little pieces.'' Instead, says Gerstner, they want an IBM that can dispel the confusion created by myriad hardware and software vendors and deliver overarching technology proposals that are grounded in an understanding of the customer's business needs. The inside reason involves IBM's products, a topic Gerstner is reluctant to discuss in detail. Yet interviews with 15 other IBM executives make clear that the company is embarked on its most ambitious technology strategy in decades. Without abandoning its existing lines of PCs and mainframes, IBM is launching a vast new line of hardware based on technology known, aptly, as the Power architecture. It will span the entire range of modern computing, from the most compact hand-held communicators to massively parallel supercomputers, or, as IBM says in computerese, ''from palmtops to teraflops'' (teraflops are a measure of computing speed). Gerstner figures that if many different IBM businesses are to share the same computer architecture, they ought to be pretty tightly linked. If Power succeeds, within five years most of the computers IBM makes will have this technology at their core. The company will have drastically lowered its manufacturing costs without any reduction in its ability to produce state- of-the-art machines. Customers will enjoy a hardware and software setup that is easily expandable and accommodates virtually all the important and popular software in the industry. IBM will have returned to a path of steady revenue and earnings growth. To realize his vision, Gerstner knows he has to work fast. Says a longtime senior executive: ''Now it's kick ass. Let's get it done yesterday. The company's so friggin' big that if you don't have somebody driving the horse 100 miles per hour, it will give up.'' IBM's fortunes have never been at a lower ebb. By Wall Street estimates, sales will shrink this year to just about $61 billion, down from $69 billion in 1990. Restructuring charges have mounted to $20.5 billion since the beginning of last year; IBM had a $325 million operating loss in the first & half of 1993. Since Gerstner took over on April 1, the stock price has dropped about 15%. In the short run, IBMers don't expect much of a rebound. But if worldwide economic conditions do not worsen unexpectedly, IBM should stop shrinking in 1994. Revenues from fast-growing segments of the business will offset the continuing decline of mainframe sales, which in the past four years have fallen over 50%, to $5.8 billion in 1993, as estimated by Daniel Mandresh, a security analyst at Merrill Lynch. Chief among IBM's growth businesses: services such as technology consulting and ''outsourcing,'' in which IBM takes over a customer's computers and runs them itself. In the past three years, IBM's annual service revenues have more than doubled, to nearly $10 billion. For Jerome York, 55, a soft-spoken West Point graduate whom Gerstner recently recruited from Chrysler Corp. to be IBM's chief financial officer, the immediate concern is carving some $7 billion from annual costs. That Herculean task, he says, won't be finished until mid-1996. Some $5.25 billion will disappear as a result of restructuring for which IBM has already set up reserves. York says the company can eliminate all $7 billion without further write-offs.

The CEO has zeroed in on what he calls ''a classic Pogo-type problem: We have met the enemy and it is us.'' IBM still maintains squads of salespeople and technicians at customers' offices, yet has trouble translating its knowledge of the marketplace into results. At an employee gathering in September, Gerstner seethed about the rivalry among IBM's divisions: ''We have people who will not respond to a customer because one unit is debating with another how they're going to share the revenue. That's ludicrous!'' The emphasis on responding to customers isn't just sales talk. Increasingly, IBM will offer on the open market the same chips and components its computer divisions use. So if there's nothing proprietary about the hardware, what will make customers buy from IBM? Gerstner believes the answer lies in IBM's knowledge of what customers in different industries need. Says Bill Filip, 48, who runs IBM's hugely successful workstation division: ''We will build our systems out of parts that will be available to our competitors. Our differentiation will be a matter of targeting specific customer segments, knowing what combination of characteristics are necessary for an airplane designer or schoolteacher or molecular chemist. There are opportunities for us to do that better than anybody else.'' Gerstner thinks he can leverage IBM's knowledge of customers by getting all parts of the organization talking to one another. He tells employees he wants them to think about what's good for IBM before they think of what will benefit their division. (Invariably he adds that they should think first about the customer.) In September, he created an executive committee that includes the six people to whom the divisions report and four other senior executives. He wants the group to spend most of its time talking about ways to improve coordination: ''The main task of the corporate office is to bring the advantages of size to the operating divisions.'' Nine task forces, each led by a member of the group, will find ways to further cut costs and reengineer companywide practices such as hardware development and inventory management. The committee's first official act: to ban profit taking between divisions on ideas, copyrights, names, and other intellectual property. Gerstner was outraged to learn that Eduquest, the IBM group that sells PCs to schools, had been told it would have to pay royalties to the IBM PC unit if it labeled its computers ''IBM.'' In an E-mail message to employees he chided, ''Henceforth, it will be our policy to share with our IBM colleagues, enthusiastically and without cost, whatever we develop.'' THE POWER PROGRAM is remaking IBM at its foundations. Instead of mainframes, the company's central technology will be a family of lightning-fast microprocessors that can work in all sizes of computer. Not since the development of the System/360 mainframe line in the early 1960s has a single technology been so important to IBM's future. System/360 cemented the company's near-monopoly on corporate computing. IBM wants to reassert the dominance it forfeited when it introduced its first PCs in 1981. By giving Intel and Microsoft control of, respectively, the processor and operating system software in PCs, IBM passed up billions in profits. With the mainframe business shrinking, IBM needs another horse to ride if it wants to remain Big Blue. The concept behind the Power architecture, reduced instruction set computing (RISC), was hatched in IBM's research labs in the 1970s. But for years, IBM failed to exploit it commercially -- an example of the dithering that burns up the new CEO. (Power stands for Performance Optimization With Enhanced RISC.) RISC systems are generally much faster than traditional designs such as Intel's microprocessors. While IBM delayed, Sun Microsystems and Hewlett- Packard, among others, brought out successful RISC workstations. That fact makes Gerstner fume: ''We should have dominated. Where have we been?'' Power made its debut in 1990, embodied in a set of chips in IBM's RS/6000 workstations. The machines were an instant hit. They ran Unix, the industry- standard operating system popular among scientists and engineers, and offered far more computing power per dollar than previous IBM products. From virtually zero in 1989, IBM quickly carved out a 17% share of the workstation market, according to Silicon Valley market research firm Dataquest. That was worth more than $2 billion in hardware and software sales last year. By the time Gerstner arrived, IBM had dramatically raised its sights for Power. Though no one in IBM is willing to say so, strategists clearly had concluded that the architecture offered a chance to remedy a potentially crippling weakness: IBM's dependence on Intel for its microprocessor designs. Vice chairman Jack Kuehler, the company's chief technologist under Akers, enunciated a principle: To control its destiny, IBM must make its own microprocessors. Gerstner agrees. ''We believe that the microprocessor is the fundamental engine of system development,'' he says, ''and we will be the creator of our microprocessors. That will give us substantial competitive advantage over those who buy their microprocessors from others.'' In 1991, IBM recruited Motorola and Apple Computer -- the No. 1 producer and user, respectively, of non-Intel microprocessors -- to help shrink the Power architecture onto a single chip. The partners call the new line of microprocessors PowerPC. Motorola and IBM each make the chips, an arrangement that lets IBM bring to bear a great hidden strength: its huge, world-class semiconductor division. IBM Microelectronics, little known because it only recently began selling chips on the open market, will have revenues of an estimated $5.5 billion in 1993, almost all of which consist of sales within IBM. Michael Attardo, 52, the division's chief, boasts about its capabilities: ''We believe we have a worldwide lead in basic silicon manufacturing processes of six to 12 months.'' In the fast-forward realm of semiconductors, that puts IBM nearly a full generation ahead. To back his claim, Attardo cites IBM's ability to manufacture silicon circuits only 0.35 micron wide, about one three-hundredth the thickness of a human hair. The division's latest memory chip, built in a joint venture with Germany's Siemens and being delivered to manufacturers in sample quantities, stores 64 megabits, the data equivalent of five copies of Anna Karenina. Most PCs today use four-megabit chips. IBM and Motorola started selling the first PowerPC chip, the 601, in September. It is a marvel of compact design. Executed in CMOS (for complementary metal oxide semiconductor), the type of chip experts agree has the greatest potential for rapid improvement, the 601 packs 2.8 million transistors in four layers on a silicon square less than a half-inch on a side. Intel's top-of-the-line Pentium, by contrast, requires twice as much silicon for its 3.1 million transistors. By most measures of performance PowerPC is as fast or faster, and it is less likely to overheat in the close confines of a computer case. Because the 601's small size makes it cheaper to manufacture, it beats Pentium hands down in price: $455 vs. $898 in quantities of 1,000. Pentium's main advantage: some 100 million people worldwide are accustomed to software written for Intel-based machines. PowerPC is getting raves from techies. Decision Resources, a Waltham, Massachusetts, research firm, issued a report in September entitled Chip Wars: Handicapping the Six Leading Microprocessors. It compared Pentium, PowerPC, and RISC processor chips from Digital Equipment, Hewlett-Packard, Sun Microsystems, and Silicon Graphics. The conclusion: ''PowerPC appears to be the only architecture that can compete with Pentium.'' The author, Gordon Bell, is no anonymous nerd: He is the architect of Digital Equipment's VAX system, which dominated the minicomputer industry for more than a decade. He's also on the technology advisory board of Microsoft. IBM has a complex but promising strategy to make PowerPC computers attractive to all those users of Intel-based PCs. Up to now, they had little choice if they wanted to run application programs on Microsoft's Windows or MS-DOS operating systems, which were written specifically to operate on Intel chips. On the same day early next year that IBM introduces its first PCs built around PowerPC chips, it also plans to start selling a new operating system, tentatively named Workplace OS. Calling the product an operating system is misleading, though, since Workplace will let users run applications written for several such systems, including Windows, MS-DOS, and IBM's OS/2. Later, | IBM will give Workplace the ability to run applications written for Unix and for the operating system being developed by Taligent, an IBM-Apple joint venture. At Workplace OS's core is a compact bit of computer code known as a microkernel, which works as a sort of universal connector that can link a variety of hardware architectures to a variety of operating systems. Programmers can readily plug into it all those other operating systems. At the same time, with little modification, the microkernel can be adapted to run on different microprocessors. Implication: If Workplace OS comes into wide use, its effect could be truly revolutionary. By undermining technical barriers that keep invaders off Intel's and Microsoft's turf, it would open the hardware and software industries to broader competition. IBM has already unveiled its first PowerPC product: a low-end RS/6000 workstation (price: around $4,000). A laptop RS/6000 with PowerPC inside is scheduled to go on sale by January. Soon after that, IBM Power Personal Systems, a startup division, will begin its assault on the $63-billion-a-year personal computer market. To run the unit, IBM has chosen a star: Nobuo Mii, 62, former head of the company's PC technology labs. Without his work, IBM could not have come up with the ThinkPad, its most successful laptop. TO SUCCEED, Power Personal will have to win customers from some of the toughest marketers in computing: Compaq, Dell, AST -- and the biggest Intel- based computer maker, IBM itself. Big Blue has set Mii's unit directly in competition with its fast-growing $10-billion-a-year personal computer unit, the IBM PC Co. Bob Corrigan, 53, a professorial IBM veteran who is president of the PC Co., is skeptical that PowerPC computers will gain much ground in the Intel world: ''I've got a great partner in Intel, and the competition is probably good for them. But if the Power architecture is very much of a threat to Intel and the IBM PC Co., Intel will make substantial improvements in its own chips. For us, it would be crazy to abandon the Intel architecture when all these customers have so many resources invested in it.'' Possibly remembering Gerstner's directive to put the interests of IBM first, he adds, ''We want IBM to be on the winning side no matter what happens.'' The rival divisions report to Jim Cannavino, 49, an intense, bushy-eyebrowed mainframe veteran who supervises PCs and workstations. Says Cannavino about PowerPC: ''I don't know where this ends up. It's a large set of markets evolving very quickly. I'm not trying to segment the market or make decisions for customers. This doesn't mean the Intel architecture will go away. It's just got a competitor. A substantial one.'' Cannavino maintains that the industry has room for more than one major architecture. But he can't talk about the PowerPC without his enthusiasm becoming apparent. Describing how an upcoming version of the chip will be the ''hottest thing in the market,'' he cracks his knuckles with glee. ''When you design an architecture as modern as Power, we know it will be the platform of choice for some customers.'' IBM is finding lots of places to embed PowerPC chips. Its AS/400 minicomputer business, which generated $4.5 billion in revenue in 1992, will convert its entire line to PowerPC within two years. IBM will start incorporating the chips in its data network switching products within the next year or so. To apply the architecture to supercomputing tasks that are beyond the reach of even the fastest microprocessors, IBM has set up a joint venture among its workstation, mainframe, and research units, called Power Parallel Systems. The group has developed machines that incorporate as many as 128 Power processors working in sync. Already 22 such supercomputers are installed at customer sites, mostly research institutions. Starting at $350,000 each, the Power Parallel systems handle computing jobs that used to be reserved for supercomputers costing $10 million and more. The group will sell hundreds of machines next year, says Irving Wladawsky-Berger, 48, the group's general manager.

In a future dominated by PowerPC, what will become of mainframes? In Gerstner's vision, IBM's big iron will be ''one more part of the product line'' -- and a source of substantial revenues and profits for years to come. The Power architecture, potent and flexible as it is, can't be adapted to run software written for IBM's current 390 series of mainframes. Internal IBM estimates of the total investment by customers around the world in programs that operate on these machines run as high as $1 trillion. To keep such customers satisfied -- and wedded to IBM's proprietary technology -- the company will soon begin to deliver a generation of mainframes designed to be sold profitably at greatly reduced prices. They'll supersede complex water-cooled computers that derive their processing power from densely packed modules of exotic high-speed chips. IBM has virtually abandoned development on machines of this type and has scaled down costly plants that built them. Instead, the new mainframe generation will be parallel machines built from microprocessors specially designed to run 390 programs. IBM will make the 390 chips using the same CMOS techniques and factories that turn out PowerPCs and memory chips. That will radically bring down the manufacturing cost for powerful computers, and make it far simpler for IBM to expand or shrink capacity as the market dictates. ''So we're driving toward a uniform technology in terms of both architecture and manufacturing process,'' says Attardo of IBM Microelectronics. He expects CMOS microprocessors to keep doubling their power every 12 to 18 months, a pace the old-style mainframe processors could not match. To make the most of the new technology, IBM will at first tailor the new mainframes to business tasks for which parallel machines are especially suited, such as transaction processing or the searching of databases. For months, Gerstner and other top executives have been briefing corporate customers around the world on elements of the product strategy. Partly as a result, the view of the company among information systems executives is more optimistic than that of investors. A September survey of senior technology officers at major companies by Computer Economics, a Carlsbad, California, research firm, found that 62% think Gerstner will be able to turn IBM into a more competitive company. Only 29% thought he could not; the remainder were unsure. Bill Anderson, the widely respected chief information officer at Prudential Securities, will spend about $80 million on Big Blue products and services this year. He says, ''IBM has as good a product line as it's had in its history. Nobody's making better hardware. I believe it is going to do better each succeeding quarter.'' Daniel Mandresh of Merrill Lynch is bullish on IBM's stock: ''I'm amazed at how much more respect IBM has in the technology world than in the investing world.'' Even some who find IBM's performance in the marketplace maddeningly erratic have high hopes. Jim Lawson is a vice president of Auspex Systems, a $100- million-a-year Silicon Valley maker of servers that help manage the operation of large numbers of PCs in a network. IBM's Storage Systems division distributes the servers under IBM's logo, and Lawson is frustrated by the contention among divisions. ''We can't get anybody in IBM to make a goddamn decision,'' he grouses. He says salespeople working with other IBM divisions confuse customers by trying to talk them out of buying Auspex machines in favor of IBM's own. ''We complain to Storage Systems, and they say they can't control the sales organization,'' says Lawson. Yet he bought IBM stock recently. He sees Gerstner moving to correct problems, and he is dazzled by many of IBM's newest products. ''IBM is going to come back,'' he says simply. ''It will be a $100 stock by next year.'' Gerstner still has a long way to go, but the world is slowly starting to focus on his vision.

BOX: KEY ELEMENTS OF GERSTNER'S VISION

-- He won't split up the company. IBM's parts will be even more closely linked.

-- IBM will reassert its identity as customers' primary computing resource.

-- The company will be the dominant supplier of technology in the industry.

-- PowerPC, a new microprocessor design, will be IBM's centerpiece. Built into many future computers, it will run a wide range of standard industry software. And it will steeply cut manufacturing costs.

-- Mainframes are no longer central to the strategy, but IBM will still make them, now with microprocessors.

- IBM is its own worst enemy. Employees must waste fewer opportunities, minimize bureaucracy, and put the good of the company before their division's.

CHART: NOT AVAILABLE CREDIT: FORTUNE CHART/SOURCES: COMPANY REPORTS, WORLDSCOPE CAPTION: IBM Armonk, New York