In an Age When Tiny Is All, Big Computers Are Hurting

By John Markoff
The New York Times

April 4, 1989

The mainframe computer is rapidly being turned into a technological dinosaur by the tiny ''computer on a chip,'' which is growing smaller, cheaper and dramatically faster.

For years, computers so large they had to be housed in their own air-conditioned rooms have powered American business and produced handsome profits for their makers. But reports in recent weeks of new financial difficulties at the nation's three largest computer makers reflect a fundamental shift away from the giant machines.

The inexpensive chips, called microprocessors, which are at the heart of the personal computer, represent the most serious challenge yet to the manufacturers that for two decades have dominated the industry and been a powerful force in the development of American business.

The new chips are being employed first in powerful personal computers known as work stations, which most industry experts agree will provide the growth and high profits of the next decade.

Market Is Barely Growing

A significant portion of the profits of I.B.M., Digital Equipment and Unisys has come from mainframes and slightly smaller minicomputers. Because the big companies dominated the mainframe market, they could keep prices high and realize large profits. Even though that market is barely growing, the largest customers have billions invested in the big machines and are unlikely to abandon them soon. Thus, the industry giants must keep turning out new equipment and software for large computers.

When the giant computer makers are active in the personal computer and work station market, that arena is much more competitive, with profit margins narrower.

I.B.M., Digital and Unisys have said that they expect their first-quarter earnings to be disappointing and that it will be a challenge to meet Wall Street's expectations for the year.

Compounding their problem is the speed at which microprocessors are evolving into ever-more-powerful chips.

Executives at the industry giants dispute this. ''I'm not very comfortable with these people who say that microprocessors are going to eat up midrange and large computers,'' said John F. Akers, chairman and chief executive of the International Business Machines Corporation.

But others point to the microprocessor's speed.

''We're at the threshold where low-cost microprocessors actually outstrip multimillion-dollar mainframes in absolute performance,'' said John Moussouris, a computer designer in Palo Alto, Calif., who co-founded the Mips Computer Corporation.

The new generation of low-cost computers presents users with an increasingly stark choice: spend thousands of dollars for cheap desktop machines or millions for mainframes to provide essentially the same speed.

The Computer and Business Equipment Manufacturers Association, a trade group, estimates that mainframe sales in the United States will climb just 2.3 percent by 1998.

Moving to the Desktop

While the mainframe will continue to be sold, the hot spots of activity in finding new uses for computers and writing the necessary programs will be on the smaller machines.

For example, MPR Associates, a policy analysis concern in Berkeley, Calif., that works for state and local governments, recently moved its statistical analysis software from an I.B.M. mainframe to small desktop computers based on the Intel Corporation's 80386 microprocessor.

And last year, when the American Express Company, one of I.B.M.'s largest mainframe customers, moved to offer a popular new service, sending its customers digitized images of billing slips rather than the slips themselves, it eschewed its corporate mainframes for the task. Instead it turned to several powerful desktop computers, or work stations, linked together by an office network.

Similarly, the LSI Logic Corporation, of Milpitas, Calif., recently replaced an I.B.M. mainframe used by design engineers with a network of work stations made by Sun Microsystems Inc., of Mountain View, Calif., a leading maker of smaller machines.

''All the hot new software is appearing on small machines,'' Mr. Moussouris said. ''Who's doing a hot new application for a mainframe these days? The vast majority of manpower is being devoted to the small machines and being withdrawn from the dinosaurs.''

'It's Amazing'

Will Zachmann, an industry analyst at Canopus Research, in Duxbury, Mass., said, ''It's amazing how these big companies are moving to microprocessor-based systems.''

Such trends contain the greatest risk, he argues, for the large computer makers.

While I.B.M., Digital and Unisys have pointed to specific short-term problems to explain their current earnings lag, their explanations mask an underlying dilemma that faces all three: each has a large stake in markets for the older mainframes and minicomputers. They face a wrenching transition as they try to retool for the new age.

I.B.M., which traditionally obtained 50 percent of its revenues and 70 percent of its profits from large systems, has been grappling with the slowing growth of its mainframe market since the mid-1980's. It is now struggling to rebuild its product line around powerful work stations that will store and retrieve files from a large, centralized computer.

The Digital Equipment Corporation gained ground rapidly on I.B.M. in the first half of the decade by relying on a single design and advanced networks to tie all its computers together. But the design is aging, and to catch up, Digital has introduced its own microprocessor-based work station.

Unisys Shifting Fast

The Unisys Corporation, created in 1986 by the merger of Sperry and Burroughs, is attempting to support both Sperry and Burroughs mainframes. But the company is shifting as rapidly as possible to microprocessor-based systems that run the popular Unix operating system, which can run on many computers.

Microprocessors are silicon chips that integrate all the features of traditional central data processors of larger mainframes. By squeezing millions of transistors onto a silicon chip, designers get a phenomenal increase in speed and a drastic drop in price. The newest generation of microprocessors - which will begin to show up in desktop computers next year - far outdistance the processing speed of today's fastest mainframes.

These newest microprocessors exploit a simple approach, called reduced instruction set computing, or RISC, that drastically simplifies computer design.

Gordon Bell, formerly the principal computer designer at Digital and now head of research and development at the Ardent Computer Corporation in Sunnyvale, Calif., believes the impact of RISC is only beginning to be felt. While mainframes are gaining in speed at only 14 percent each year, RISC chips are increasing in speed as much as 70 percent annually, he said. Last year, Dr. Bell said, came the crossover point - when RISC chips surpassed the speed of their mainframe ancestors.

Measuring Power

However, processing speed, determined by how quickly a computer can perform basic calculations, is only one measure of power. Until now mainframes have remained supreme in their ability to shuttle vast quantities of data instantly.

Now for the first time designers are taking aim at this remaining mainframe advantage. At least five start-up companies are designing specialized machines that will handle data with the speed of a mainframe but at a fraction of the cost. These computers will be linked by networks to desktop work stations.

''The traditional data processing center is being teased apart with a computer network inserted in the middle,'' said John Shoch, a computer scientist and a partner at Asset Management, a Palo Alto venture capital firm.

Not everyone in the industry believes the mainframe is withering. Executives at mainframe makers point to the tremendous software investment that companies have made and say it will place a big obstacle in the way of a shift to smaller systems.

Interviews with I.B.M. executives reveal that the company believes that the vast rise in desktop processing power will be soaked up by increasingly sophisticated features like voice recognition, high-resolution graphics and animation. They contend that the advent of networks will lead to a renaissance for mainframes in which they are used in new and specialized ways. Dilemma for I.B.M. But Mr. Akers acknowledges that inroads have been made by desktop systems. ''Clearly the utility and flexibility and independence of microprocessing has satisfied tremendous demand that before it came along would have had to be satisfied by I.B.M. or somebody else's midrange or large computers,'' he said.

Nowhere is the dilemma clearer than in the choices facing I.B.M., the industry leader. The company annually gathers more than $16 billion in revenues from its mainframe hardware - a profitable market that it virtually dominates.

But while it initially experienced great success in the market for personal computers - which did not compete directly with the larger machines - it has so far been unable to make significant inroads into the market for work stations.

I.B.M. executives insist that the company has not given up in the battle for the ''intelligent work station'' market. The company said it will have several big hardware and software announcements this year that will begin a new strategy for gaining a foothold in a market expected to continue to grow by 64 percent a year through 1993.