Chapter 5 - Transient Technologies

To create sustainable strategic advantage, a new-product wave should represent a radical shift in both technological assumptions and business model. The PC would fill the bill; the minicomputer did not. The mini makers’ processor technology and business models were simply lower- cost versions of the mainframe scheme. The mini and the mainframe were alike in most other respects, too. They were vertically integrated products with customized circuit designs, proprietary operating systems, and wholly owned distribution channels and field-service operations. The PC would challenge each of those layers and would win on every count. The minicomputer wave would soon run its course, despite the best efforts of its executive cadre.

The fifteen-year minicomputer cycle began around 1968, and nearly a hundred companies joined the half dozen established firms over the next four years—but only half actually developed a salable product and just ten succeeded. Prime Computer and Data General led the newcomers, followed at diminished growth rates by MAI Basic Four, Four-Phase Systems, Computer Automation Systems, General Automation, Macrodata, Microdata, and Modular Computer Systems (Modcomp). Creative naming could hardly account for their early successes.

Ho-hum naming aside, the biggest winners were older, more experienced companies, notably DEC, along with Datapoint, Wang, HP, and eventually IBM, all of which entered the minicomputer space through a side door. Datapoint had been languishing as a maker of mainframe monitors when it invented the programmable terminal in 1972. That same year, Wang Labs made the death-defying leap from calculators to office computing and word processing, while HP, almost simultaneously, brought in new leadership to transform its withering minicomputer business into an industry leader.

Mainframes still offered more total power, but minicomputers were less expensive to build, didn’t need special air-conditioning or costly glass houses, and had more flexible operating systems. The mainframes incorporated the clumsy and time-consuming “batch processing” methodology, which required specially trained machine operators to periodically load punch cards or tapes into the machine, which would then spew out the encoded information as updated records and printed reports. In contrast, the interactive minicomputers allowed regular office personnel to access the systems’ internally stored information as needed. Gone were the ugly, space-consuming stacks of printed outputs. Better yet, the minicomputers’ leaner cost model translated into cheaper products. Whereas IBM needed an 85 percent gross margin on its mainframes to cover everything from operating- system development to field operations to profits, Digital Equipment prospered on margins of just 60 percent. By 1983, IBM’s price per MIPS (millions of instructions per second) on its high-end 3084 mainframe was about $285,000, compared to $190,000 on its 4300 minicomputer, $170,000 on HP’s 3000, and $150,000 on DEC’s VAX 780. That implied a 50 percent premium on big-iron MIPS—a clear repudiation of Grosch’s law implying that MIPS costs should instead decline with scale. A combination of base technology and a new business model upended Grosch’s neat symmetry.

Minicomputers flowered at corporations large and small, propelled by their price and software advantages over mainframes and their ability to function reliably at typical office electrical power and temperature levels. Telecommunications costs and unreliability were the third factor in their popularity. Work sites served by local minicomputers didn’t need continuous access to telephone lines, unlike mainframes connected to “dumb” monitors. In the 1970s, telecommunications tariffs set by monopolist AT&T were projected to drop by only 30 percent over the next decade. Minicomputer price/performance would gain at least 500 percent in that same period. So local computing offered a better alternative than central mainframes accessed over expensive and spottily performing telephone lines.