Herman Hollerith

Herman Hollerith, an engineer who had worked on the 1880 United States Census, watched that count take nearly eight years to tabulate by hand and realized the next census would not survive the same approach. The 1880 count ran from 1880 until roughly 1888 before the Census Bureau finished processing it, clerks working through tally sheets by eye and by pen, and with the US population growing fast, the 1890 census was projected to take even longer manually, possibly overrunning the full decade before the 1900 count would need to start, which would have made the constitutionally mandated decennial census functionally impossible to finish on schedule. Hollerith's answer was a punched-card tabulating system built on the same basic idea Joseph-Marie Jacquard had used to control a loom nearly a century earlier: encode information as a pattern of holes punched into a card, and let a machine read that pattern rather than a person. Each census respondent's data, age, occupation, marital status, and more, was encoded as specific hole positions on a standardized card, one card per person, so a stack of cards became a machine-readable encoding of the entire population being counted.

The genuine technical leap in Hollerith's design was electrical, not mechanical: where Jacquard's loom and Babbage's analytical engine designs used the punched card to physically block or permit a mechanical linkage, Hollerith's reading mechanism used the card to complete or interrupt an electrical circuit, which is the first point in this series where electricity itself enters the hardware. A bank of spring-loaded pins pressed down onto each card at a fixed grid of positions; wherever a hole had been punched, a pin passed straight through and dipped into a small cup of mercury sitting beneath the card, closing a circuit, and wherever the card was solid, the pin simply rested on the surface and no circuit formed. Every completed circuit sent a pulse to a corresponding mechanical counter dial on the operator's console, advancing it by one, so a tally that once required a clerk to read a tally sheet and make a pencil mark now happened automatically at the moment a hole passed under a pin. A second machine, the sorting box, read the same hole pattern and used it to trigger the opening of a specific bin lid, letting an operator physically sort a stack of cards into groups, by state, by occupation, by any encoded category, so the data could be cross-tabulated by running already-sorted subsets back through the tabulator for finer counts.

The result was not a modest improvement: the 1890 census, with roughly 63 million people to count, was completed in about one year using Hollerith's tabulating machines, versus the nearly eight years the 1880 census had needed by hand, a roughly eightfold speedup that let the Census Bureau finish comfortably within the decade and kept the whole decennial counting system from breaking under its own population growth. That result turned Hollerith's machine from a clever engineering demonstration into an obvious commercial product, and in 1896 he founded the Tabulating Machine Company to build and lease the equipment to the Census Bureau and to other large organizations that needed to process bulk records, railroads tracking freight, insurers tracking policies, anyone drowning in paper tallies the way the Census Bureau had been. In 1911 the Tabulating Machine Company merged with three other manufacturers, a scale company, a time-clock company, and a meat-slicer and coffee-grinder company, to form the Computing-Tabulating-Recording Company, usually abbreviated CTR, a diversified maker of office and industrial machinery with Hollerith's tabulators as one product line among several. In 1924, under the leadership of Thomas J. Watson Sr., CTR was renamed International Business Machines, IBM, and the punched-card tabulating business that began as Hollerith's answer to a slow census became the founding product line of what would grow into one of the most consequential companies in the entire history of computing.

Hollerith's machine answered a real, pressing bureaucratic problem, counting a growing nation on schedule, using a punched card descended directly from Jacquard's loom, and the answer it gave was electrical: holes no longer just permitted a mechanical linkage, they closed a circuit that drove a counter, which is the moment pure mechanism in this series starts turning into something closer to electrical computation. That is still a long way from a logic circuit in the modern sense, Hollerith's dials counted and his sorting box sorted, but neither could evaluate a Boolean condition or combine signals the way a logic gate does, and the theoretical bridge from electrical counting to electrical logic would not arrive until 1937, when a young graduate student named Claude Shannon showed in his MIT master's thesis that George Boole's nineteenth-century algebra of true and false could be implemented directly with electrical switching circuits, turning "the circuit either closes or it doesn't" into a formal mathematics of computation itself.