A Bold Move In Mainframes

Cover Story

IBM Plans To Make Them Key To Networking--And So Restore Its Growth

John W. Verity in New York
Business Week

May 29, 1989

It's the big question in computer circles and on Wall Street, too: Are IBM's glory days gone forever? Will the company ever get back to the growth it enjoyed in the early 1980s, when both revenues and earnings nearly doubled in five years? Back then, IBM's strength was in mainframes, which produced about 50% of its revenues and nearly 70% of its profits. With 60% of the world mainframe market, IBM looked impregnable. Personal computers were still toys, engineering workstations were a lab curiosity, and neither of them was a threat in corporate offices where IBM was king.

What a difference a few years make. Personal computers are everyday work tools now, and workstations are out of the lab. Both can be linked in networks that do much of what a mainframe can--for 1/100 the cost. As a result, world sales growth for mainframes has dropped to 4% a year, one-third the rate in 1982. No longer the largest computer market, mainframes are third, behind PCs and minicomputers. Only one thing hasn't changed: Despite its efforts to diversify--into PCs, workstations, even phone equipment with the failed acquisition of Rolm Corp.--IBM relies as heavily as ever on its mainframe business. And because it does, it has become a laggard.

NEW STRATEGY

IBM's revenues have grown at an average 6.1% a year since 1985 vs. 9.2% for the overall computer industry. Despite a reorganization last year and extensive cost-cutting, the company will struggle this year to match 1984's record $10.77 a share in profits--on about $18 billion more in revenues. Its stock is trading at 11.9 times earnings, 4% below the average for the Standard & Poor's 500-stock index. Even John F. Akers, IBM's determined chairman and chief executive, expects his company to grow more slowly than the overall industry for two more years, perhaps--and that may be optimistic. ''I don't see any scenario in which IBM can grow faster than the industry,'' says Rick Martin, an analyst at Prudential-Bache Securities Inc. ''It's too dependent on mainframes.''

But now, IBM is trying to turn its albatross back into an eagle. It has a new strategy that it hopes will rekindle the mainframe's growth and help sell smaller computers, too. IBM will continue to hedge its bets by investing in every other promising technology. But even if it dominated the rapidly growing markets for workstations or computers used for image processing or desktop publishing, these businesses are so small that the effect on IBM's $60 billion in revenues would hardly show up. So in a way, IBM has no choice but to spend untold billions to save the mainframe. The plan is to transform the 19-year-old System/370, which already is the central repository of data for most large U. S. companies and many world governments, into an entirely new computer. A remodeled 370 would store more data, including electronic images, voice messages, and even video.

But IBM wants to do ''more than simply turn the crank,'' says Akers. It wants to make the mainframe into a hub that will anchor vast information networks in the 1990s. Digital Equipment Corp., most notably, is already headed in this direction, selling networks based on minicomputers. But the networks IBM plans would be much more powerful. With its system, the command ''Tell me last week's sales by product line'' would produce numbers automatically calculated from data bases located around the world. Currently, such a request might mean querying 100 different business units and tabulating the answers.

In short, IBM is pushing mainframes as the solution to the No. 1 computer problem of corporations--managing an overwhelming surplus of data. Powerful PCs may be useful. But how is a company to make sense of the expanding pools of information collected by all of those, each independent of the rest? ''The fundamental problem is getting the whole organization to use its data consistently,'' says J. Bruce Harreld, chief information officer at Kraft General Foods. IBM's answer: Make the mainframe lord of data bases. Connect it to all the smaller machines, and have it hand them any piece of information they demand, even if it's in another small machine half a globe away. Revamping the aging System/370, says Brian Jeffrey, managing director of International Technology Group in Los Altos, Calif., is ''IBM's most significant project since the System/360 introduction in 1964.''

CRUCIAL EDGE

IBM took a big step toward its goal on May 16, when it announced software to tie together its own mainframes, AS/400 minicomputers, and PS/2 personal computers for what it calls ''cooperative processing.'' The programs, called OfficeVision, fit into a grand software scheme IBM calls Systems Application Architecture. SAA is an evolving set of strictly defined rules for the connections, or interfaces, between programs, data bases, and complete computer systems in the IBM product line. In theory, machines running software with SAA interfaces will more easily tap into the huge amounts of data stored in the 370.

SAA was conceived just for IBM customers. But now the company is encouraging all software makers to use it by giving them the specifications for the interfaces. ''We would like to see SAA accepted as an official standard,'' says Earl F. Wheeler, head of IBM's Programming Systems Group. If this happens, IBM's mainframe sales should boom. SAA ''will require more mainframe power and more storage,'' Wheeler adds. Beyond that, the company would begin to leverage its strength in large systems into sales of computers that would be attached to its mainframes. For instance, since it would know the most about SAA's future direction, IBM could be the first to market with SAA-compatible products.

That may be the catch, however. Indeed, the biggest threat to SAA and the new mainframe strategy may be that IBM has sole control of them. Over the past five years, customers increasingly have demanded that computer makers dispense with proprietary products--those that work only with one company's systems--and move toward open standards that let customers mix brands of machines. That forces suppliers to compete more on price and performance. SAA isn't completely proprietary, since it's being shared. But it still gives IBM a unique edge, and it can be construed as an effort to subvert the mounting open-systems movement to IBM's advantage. That may be enough to scare away buyers who don't want to depend on just one supplier.

Another problem for IBM is timing. It isn't the only company aiming to be the king of what it calls enterprise networks. DEC, Tandem Computers, and Unisys, among others, hope to beat IBM to the punch with less costly, more flexible solutions, albeit ones that aren't as sweeping as IBM's. Other computer makers, meanwhile, including Hewlett-Packard Co. and American Telephone & Telegraph Co., are hammering out nonproprietary software standards for managing large networks. And suppliers of data base software, such as Oracle Corp. and Relational Technology Inc., are tackling the problem of distributing data across many machines. The complete version of SAA may not be ready for years--after several competitive products are out. If it is too late to market, IBM's grand scheme could fall flat.

Even if IBM is first, there's the question of how wide a following it can build. James L. Cassell, a former IBM mainframe marketing executive who now tracks large systems for consultant Gartner Group Inc. in Stamford, Conn., says that DEC's inroads and rapid advances in desktop machines have loosened IBM's hold on the 1,000 or so big corporate customers that he estimates account for more than half its profits.

FUTURE-MINDED

These customers are bound to IBM by the billions of dollars they have spent on its mainframes and the software to run them. But there's a heightened risk, as open systems abound, that true Blue customers might drift away from IBM for future purchases. The No. 1 goal of IBM's new strategy, Cassell says, is to ''recapture control'' of its largest customers--partly by giving them a system so integrated it's hard to clone. ''IBM really cares about that piece of the business,'' he says.

The implication is that many smaller buyers--who in any case might be turned off by the steep price of SAA (below)--will remain up for grabs. ''There's not much there for smaller customers,'' says Steve J. Lair, an analyst with Dataquest Inc. Although not publicly, IBM has been mulling over a more versatile, network-oriented mainframe for years--at least since the early 1970s. That's when it quietly began work on a computer called Future System. The idea was to replace the System/370, which was merely an expanded version of the earlier System/360 and was severely limited in handling the kinds of terminal-based, transaction-type applications customers were beginning to plan. These ranged from airline-reservation systems to banking terminal networks, where many employees needed access to a central store of constantly updated information. FS was supposed to be much more powerful and more functional in those kinds of jobs yet much easier to program and operate. It would handle huge data bases but shield customers from their incredible complexities: Many tedious housekeeping tasks normally handled by programmers would be eliminated in the system's completely redesigned, radically new architecture.

FS never saw the light of day, though. In 1975, after spending $3 billion or more, insiders say, the company dropped the project without even acknowledging its existence. The main reason: IBM was afraid that customers wouldn't go for the machine if it meant they would have to rewrite their 370 software. That's the problem SAA aims to avoid by creating smooth connections between old 370 software and a modified mainframe.

So IBM stuck with the System/370, gradually extending its capabilities every few years, even though by 1974 it was, by IBM's own admission, outdated. It was hard to argue with the results. From 1975 to 1985, world mainframe sales and IBM's revenues more than tripled. Ironically, the 370's inherent inefficiencies--it was designed essentially to handle batches of punch cards--may have forced customers to buy more machines than they would have needed with an updated computer.

All the while, FS's spirit lived on within IBM. Its basic architecture showed up in the System/38 minicomputer in 1980, recently replaced by the AS/400. Most important, veterans of the FS project moved into top executive slots, including Vice-Chairman and Chief Technologist Jack D. Kuehler. They are slowly molding the System/370 into the shape of an updated FS. And by now, even Chairman Akers is talking like an old FS hand.

When William C. Lowe quit earlier this year as head of the Entry-Level Systems business, responsible for personal computers, Akers replaced him with the former president of the mainframe division, James A. Cannavino. ''It is very useful as we try to . . . combine intelligent workstations and large computers to have a guy with his background in large computers working from the other end,'' says Akers. He adds: ''The biggest opportunity for this industry is to make computers easier to use and manage.''

'BABY STEPS'

That task falls to Carl J. Conti, a former FSer who now runs mainframe development. His goal: to multiply many times over the 370's ability to handle huge amounts of data, a must if it is to run big networks. Last year, IBM greatly increased the machine's main memory. As early as next year, the rate at which information and programs flow into that memory will be quadrupled when optical fiber cables are added to connect disk drives and other storage devices. IBM also plans much faster, higher-capacity disk drives and connections to make several mainframes share immense banks of disk and tape drives instead of one set per machine. Meanwhile, the now laborious task of tracking data files and making backup copies is being automated to increase productivity in large data centers.

By 1991, when a new family of processors called Summit is due, the System/370 will have enough oomph to execute close to 400 million instructions per second (MIPS), nearly four times what's possible now. IBM will further boost operating power by switching to super-cooled gallium arsenide microchips--sometime in the mid-1990s, analysts think. The company also is backing Supercomputer Systems Inc., headed by former Cray Research Inc. designer Steve Chen. His scientific processor, slated for production in the early 1990s, would give the System/370 scientific number-crunching power 100 times better than today's best supercomputer, IBM claims.

Each of these changes is making the System/370 look more like what FS was supposed to be, says Pru-Bache's Martin, a former IBMer. But this time, adds Cassell, IBM is reengineering the mainframe in ''baby steps,'' not all at once. That will ease the transition for customers. It may also be unavoidable. ''There's such a vast amount of software to be written that even IBM can't do it in one fell swoop,'' he says.

The ultimate aim, says Conti, is a mainframe that can process thousands of queries a second from all kinds of computers: IBM, DEC, Apple Computer, and even Sun Microsystems workstations. That should keep IBM's mainframes at the center of its best customers' expanding networks no matter what brand of smaller computer they buy. And that would help drive the demand for mainframe equipment, with its 65% pretax margins. The synergies of connecting desktop and midrange computers to the mainframe, says Martin C. Clague, Conti's assistant general manager for marketing, will soon ''cause an inflection point'' in mainframe sales. That's tech talk for turning from a slowing growth curve to one that's re-accelerating.

RIVALRY

But how soon? And how much? Those are the $64 questions. Gartner's Cassell says demand for mainframe power and storage capacity has been growing annually at 30% to 35% in recent years. It has been satisfied largely by upgrading existing machines. But by 1992, he estimates, growth could increase to 40% to 45%, if users begin connecting their other computers to mainframes. That should raise IBM's annual growth in mainframe revenues from a projected 8% this year to 10% to 13% a year, Cassell reckons. Says Pru-Bache's Martin, using a popular measure of computer performance: ''If cooperative processing works, use of mainframe MIPS should go through the roof.''

Independent software companies might then give a growing share of their attention to IBM computers--and less to IBM's rivals. That includes DEC plus two groups of companies, one led by AT&T and the other by IBM, that are trying to establish a set of open software interfaces centered on Unix, an operating system that has been adapted to virtually all brands of computers. At least, that's what IBM anticipates.

Still, a lot could go wrong. There's the matter of delivering all the new software. IBM's first round of SAA programs, just introduced, will contain some 5 million lines of new code, the company says. The 370's 15 million-plus lines of core programming also must be updated. And Wheeler's staff must keep a tight rein on IBM's dozens of software groups to ensure that they adhere strictly to the spirit and letter of SAA.

Beyond that, there's more to conquering the network market than coming up with a bigger idea. DEC, having missed out on the seven-year-old PC boom, has developed software to connect those small machines to its networks. And DEC has been beefing up its VAX minicomputers' power to handle the kinds of data base transactions such networks require. DEC even boasts that it's way ahead of SAA on one score: Its networking software already interacts equally well with IBM PCs and DEC's own VAX workstations, Apple's Macintoshes, and the bulk of Unix-based computers.

While customers wait for SAA, moreover, PCs will make further inroads into mainframe territory. A new $10,000 DEC workstation, for instance, offers the same number-crunching power as mainframes that cost 50 times more. And for many purposes, the DEC machine is much easier to program. Such price and operating differentials are beginning to relegate mainframes by and large to the area of big, central applications. Lockheed Corp., for instance, recently came out with a workstation version of its CADAM mechanical design program, which traditionally ran on IBM 370s.Instead of keeping pace in desktop computers, moreover, IBM has fallen behind. Although it has sold more than 3 million PS/2 personal computers, its PC market share is at a six-year low of 12.5%, the result of competition from the likes of Compaq Computer Corp. and Apple. Meanwhile, IBM's workstation, the three-year-old RT-PC, has been a highly visible failure, though an improved version is due out in the fall.

Those are the types of problems that were supposed to be solved by IBM's 1988 reorganization of its U. S. operations. Akers trimmed corporate staff, thinned middle management, and made U. S. marketing and product development report to a single executive, Terry Lautenbach--all to make the lumbering giant more nimble in fast-changing markets. Akers says decisions are now being made lower in the organization and therefore faster than before. He adds: ''Some of the fruits have been picked, but I feel we have much more to do in quickening the pace'' of developing new products. Wall Street thinks he also should do more to cut costs. In the U. S., IBM's sales expenses have risen 60% in the past three years, even as slow demand has caused its revenues to decline by 20%.

PACKAGES

Partly, the higher costs reflect the huge holding action IBM has mounted until its new mainframe strategy kicks in. For instance, much of IBM's domestic cost problem stems from what U. S. marketing chief George H. Conrades calls the company's campaign to win the ''hearts, minds, and budgets'' of customers.

That started two years ago, with a redeployment of thousands of people from staff jobs to the field, where they aren't yet fully productive. Once just a seller of machines, Conrades says, IBM now wants to sell ''total solutions.'' These range from PC packages, bundled with software for specific jobs, to corporate ''systems integration'' contracts comprising everything from planning and designing an entire network, to building, maintaining, and even running it. The change in attitude shows up in the new compensation structure for IBM salespeople. In the past, commissions were calculated according to an arbitrary points system that mainly encouraged hardware sales. Now, commissions are set according to total sales, whether of hardware, software, or services. Conrades credits the new plan with putting ''emphasis on the customers and a premium on value-added selling.''

To help stimulate its customers' imaginations and purchases, IBM also has set up a $100 million ''market development fund'' that this year, Conrades says, will identify some 15,000 new ways to use computers. At New York City's Human Resources Administration, for instance, an IBM team has just spent six weeks dissecting and analyzing office operations in excruciating detail. The outcome: no immediate sales, but lots of goodwill and market intelligence that will be useful there and in other cities. Says Kraft General Foods' Harreld: ''In the past, IBM told people what to do and how to do it. Now, they come in and say: 'Tell us how we can help you.' It's a much calmer, relaxed style.''

TESTIMONIES

To back up the big marketing push, IBM also has signed up legions of small software and hardware companies to fill gaps in its product line. And it treats them more kindly. In the past, for instance, IBM might let such companies sell its machines with their customized software, but it would compete against them for the same customers. Now, it has stopped that and instead assists such companies. Financial software supplier Walker Interactive Systems in San Francisco proposed a joint series of sales seminars across the country this spring, and in less than an hour, IBM agreed to participate. It even kicked in money for the road tour and is paying Walker a commission on sales of any extra mainframe gear that its software may stimulate, says David G. Fisher, Walker's vice-president for marketing.

IBM will need every one of these allies as it executes its new mainframe strategy. They are key players in the holding action. And by their frequent testimonies on how much IBM has changed for the better in recent years, they also tend to prove IBM's assertion--crucial if it is to sell the idea of mainframe-based networks--that it knows what the customer wants and can deliver.

Once this message sinks in, maybe IBM will stop hearing the old joke about the three women who are discussing their husbands over a cup of coffee. One describes her man as just perfect--a professional wrestler who's not afraid to play rough. The second says her husband is a poet--gentle and sensitive in all the right ways. But the third woman is just plain frustrated: ''My husband's an IBM salesman,'' she says, ''and all he ever does is sit at the end of the bed and tell me how great it's going to be.''

MAINFRAMES REMAIN IBM'S BREAD AND BUTTER Revenues Gross profits * Estimates for 1989, millions of dollars Mainframe hardware, software, peripherals, and maintenance $28,262 $18,083 Midrange hardware 4,891 3,004 Nonmainframe peripherals 3,900 2,379 Personal computers, terminals, workstations and typewriters 12,252 6,249 Nonmainframe software and maintenance 7,824 4,964 Other 6,950 2,102 *Profits before selling, general, and administrative expenses, research and development, and taxes DATA: SANFORD C. BERNSTEIN & CO. ESTIMATES

Photograph: 'The biggest opportunity for this industry is to make computers easier to use and manage' JOHN F. AKERS, CEO PHOTOGRAPH BY LOUIS PSIHOYOS Photograph: CONRADES: NOT JUST MACHINES--BUT ''TOTAL SOLUTIONS'' PHOTOGRAPH BY LOUIS PSIHOYOS Photograph: IBM'S WHEELER: OFFICEVISION WILL ALLOW IBM PCs, MINICOMPUTERS, AND MAINFRAMES TO WORK TOGETHER FOR THE FIRST TIME PHOTOGRAPH BY LOUIS PSIHOYOS Photograph: CANNAVINO: A MAINFRAME MAN TAKES CHARGE OF THE PC DIVISION PHOTOGRAPH BY LOUIS PSIHOYOS