On a morning in early July, Louis V. Gerstner Jr. took the stage at the New York Hilton Hotel to tell 1,000 customers a thing or two about big computers. Declaring reports of the mainframe's demise premature, he chided his listeners: "IBM and you all deserve a shot in the head for allowing this mythology that mainframes are dead." Easing into his I-am-one-of-you role, IBM's CEO declared: "You take my mainframe away, and my business is dead." The mostly hard-core mainframe customers ate it up. At the end of his pep talk, Gerstner was given a T-shirt emblazoned with a Tyrannosaurus rex and the words "Mainframe Rex." Beaming, he raised the shirt and shouted: "Dinosaurs are back."

Not so fast. A new Ice Age may be coming--as soon as next year. To be sure, Big Blue is having a respectable year selling big iron. Analysts say it is sold out through the current quarter and perhaps the remainder of the year. Still, IBM mainframe revenues will be down for the year--although less sharply than last year's 40% drop. And the slowdown could reaccelerate in 1995 as the likes of Hewlett-Packard Co. and Sequent Computer Systems Inc. keep snagging bread-and-butter corporate customers with large-scale "server" computers running the Unix operating system. These machines anchor the local-area networks (LANS) that businesses now favor (chart).

CHIP SHOT. IBM has an answer: a new series of mainframes based on new technology. Although the machines aren't yet a replacement for conventional mainframes--they only do specific tasks, such as transaction processing and database query, and work as adjuncts to IBM mainframes--Big Blue has portrayed them as the start of a new era for big iron. Based on the semiconductor technology found in personal computers, the new models cost at least 25% less than traditional mainframes to produce, IBM says.

But interest in the new PTS and PQS has been limited--largely because the cost savings aren't reflected in the prices, which are only a bit less than those of conventional IBM mainframes. So older mainframes are selling fast. But with prices 30% higher than those of Unix servers, the new machines aren't.

Why not simply cut prices? In a word: profits. Gerstner has promised investors a substantial gain in the bottom line this year, and he needs a good year in big iron for that. "What they want to do is maximize their profit," says Gartner Group Inc. analyst Charles Burns. Mainframes and related software and services contribute some 30% of IBM's overall revenue and over half its profits. Still, if IBM loses more customers, says Burns, "this may be a win-the-battle, lose-the-war scenario."

To win the war against the upstarts, IBM must quickly move customers to newer and cheaper technology--even if it means knocking off its own products. If not, rivals will be happy to do the job. "HP eats IBM's children, and they do it well," says Meta Group Inc. analyst Carl Greiner. "For some reason, IBM just doesn't want to get real [on pricing]. I can give them a jillion [customers] they've lost on price."

IBM says prices will reflect the new technologies--eventually. The PTS and PQS models use PC-type chips, and future models, using dozens of processors, should provide a further boost in bang for the buck. "This is really PC technology being applied to your mainframe," says Philip G. Heasley, vice-chairman of First Bank System Inc. and an early PQS customer. But according to Andrew J. Hurter, director of U.S. marketing for IBM's large-scale computing business, it may take IBM up to 30 months to work such technology completely into its product line--an eternity for some customers.

In the meantime, the cost per million instructions per second (MIPS) of computing power on IBM machines remains high. According to Gartner's Burns, IBM's ES/9000 machines sell for $27,000 to $29,000 per MIPS, and the new models are at $25,000 to $27,000. Meanwhile, large Unix servers sell for only $18,000 to $20,000.

MIPS GYPS? That doesn't sit too well with customers. "We suggested that $10,000 to $12,000 [per MIPS] would be a number that would get us interested," says Boyd Hopkins, a vice-president at CIGNA Systems, the data-processing arm of insurer CIGNA Corp. Price isn't Hopkins' only worry. Key software--such as a version of IBM's DB2 database program--won't be available for the new line until next year. IBM's Hurter says lower prices and general-purpose machines are coming. And if customers won't wait? "You always worry that your market is going to run away from you before you get there," he says.

For now, Gerstner is right when he says the mainframe isn't really extinct. And he can even boast of a lineup so popular that IBM can't keep up with demand. But IBMers concede that the shortage is not entirely due to demand: Hit by a huge falloff in mainframe sales --from $10.7 billion worldwide in 1992 to a $6.4 billion last year, according to SoundView Financial Group--IBM simply underestimated demand. In other words, Gerstner should remember that some myths are based on fact.

Before it's here, it's on the Bloomberg Terminal. LEARN MORE