Almost as complex was an airline reservation system, called SABRE, that IBM created for American Airlines in the late 1950s and early 1960s. Using a million lines of program code and two big computers, it linked agents in 50 cities and could handle millions of transactions a year, processing them at the rate of one every 3 seconds. But writing the software for SAGE or SABRE was child's play compared to what IBM went through in the 1960s when it decided to overhaul its increasingly fragmented product line and make future machines compatible—alike in how they read programs, processed data, and dealt with input and output devices. Compatibility required an all-purpose operating system, the software that manages a computer's basic procedures, and it had to be written from scratch. That job took about 5,000 person-years of work and roughly half a billion dollars, but the money was well spent. The new product line, known as System/360, was a smash hit, in good part because it gave customers unprecedented freedom in mixing and matching equipment.

By the early 1970s technology was racing to keep up with the thirst for electronic brainpower in corporations, universities, government agencies, and other such big traffickers in data. Vacuum-tube switches had given way a decade earlier to smaller, cooler, less power-hungry transistors, and now the transistors, along with other electronic components, were being packed together in ever-increasing numbers on silicon chips. In addition to their processing roles, these chips were becoming the technology of choice for memory, the staging area where data and instructions are shuttled in and out of the computer—a job long done by arrays of tiny ferrite doughnuts that registered data magnetically. Storage—the part of a computing system where programs and data are kept in readiness-had gone through punched card, magnetic tape, and magnetic drum phases; now high-speed magnetic disks ruled. High-level programming languages such as FORTRAN (for science applications), COBOL (for business), and BASIC (for beginners) allowed software to be written in English-like commands rather than the abstruse codes of the early days.

Some computer makers specialized in selling prodigiously powerful machines to such customers as nuclear research facilities or aerospace manufacturers. A category called supercomputers was pioneered in the mid-1960s by Control Data Corporation, whose chief engineer, Seymour Cray, designed the CDC 6600, a 350,000-transistor machine that could execute 3 million instructions per second. The price: $6 million. At the opposite end of the scale, below big mainframe machines like those made by IBM, were minicomputers, swift enough for many scientific or engineering applications but at a cost of tens of thousands rather than hundreds of thousands of dollars. Their development was spearheaded by Kenneth Olsen, an electrical engineer who cofounded Digital Equipment Corporation and had close ties to MIT.