hardware

Hardware in the sense of metal goods — nails, hinges, locks, tools — has been English since the 15th century. The hard distinguished metal goods from soft goods: cloth, leather, perishables. A hardware shop in a 19th-century English town sold the small iron and steel components that held houses and machinery together. The word was material and literal: these things were hard.

When electronic computing emerged in the 1940s, engineers working on ENIAC and its successors needed to distinguish the physical machine from the programs that ran on it. The machine — the vacuum tubes, the wiring, the switches — was called hardware, borrowed from the metal-goods tradition. The programs were called software, a playful inversion coined by the mathematician John Tukey in 1958. Hard and soft, tangible and intangible, permanent and changeable.

The hardware/software distinction encoded a fundamental division in computing: hardware sets the limits of what is physically possible; software explores the space of what is logically possible within those limits. The same physical hardware (a processor) can run completely different software — a word processor, a game, an operating system. The machine is constant; the programs are variable. The nails are nails; the building they create can be anything.

Modern hardware includes processor chips built at scales measured in nanometers — the transistors on an Apple M2 chip have features 3 nanometers wide (smaller than a strand of DNA). The word hardware persists for objects nearly incomprehensible to 15th-century nail-sellers, objects so precise and small that their manufacture requires the most advanced machinery ever built. The hard objects got smaller. The word got larger.