From gears to electricity
Calculating machines existed long before electronics. Mechanical adding machines and desk calculators were common in offices by the early twentieth century, and the nineteenth-century designs of Charles Babbage showed that a machine could, in principle, follow a changing list of instructions. Babbage’s engines were never completed in his lifetime, but the idea of a general-purpose calculating machine stayed in circulation.
By the 1930s, demand for faster calculation was growing. Scientists, engineers, governments and insurers all needed large tables of numbers. Much of that work was done by people, sometimes called computers themselves, using desk calculators. Some of the most important problems were also being tackled with analog computers, which represent quantities with physical measurements such as rotations or voltages rather than with digits.
The relay stage
The first big technical step away from purely mechanical devices used electromechanical relays: small switches opened and closed by electromagnets. Relays were already familiar from telephone exchanges, so the components were available and understood.
Several relay machines were built in the late 1930s and 1940s. In Germany, Konrad Zuse built a series of machines in the late 1930s and early 1940s, including the Z3, which is often described as an early programmable, automatic calculator. In the United States, Bell Telephone Laboratories built relay calculators, and Howard Aiken, working with IBM, led the project that produced the Harvard Mark I, which began operating in 1944. These machines were slow by later standards because moving parts have to physically switch, but they showed that long calculations could run automatically without a person handling each step.
The electronic step
Replacing relays with vacuum tubes, which switch using flows of electrons instead of moving metal, made much higher speeds possible. The trade-off was complexity and reliability: a machine with thousands of tubes had thousands of parts that could fail, and early builders worried about exactly that. Careful engineering and operating practices made it workable.
Several groups worked on electronic calculation at about the same time, often without knowing about each other:
- John Atanasoff and Clifford Berry at Iowa State College built a special-purpose electronic calculating machine around the early 1940s, designed to solve sets of linear equations. It used binary numbers and electronic switching, but it was not programmable in the general sense.
- A British team at Bletchley Park and the Post Office Research Station, with engineer Tommy Flowers prominent among them, built the Colossus machines in the mid-1940s to help with codebreaking. They were electronic and used large numbers of tubes, but they were designed for a specific task rather than general computation. Their existence stayed secret for decades after the war, which affected how the history was told.
- John Mauchly, J. Presper Eckert and colleagues at the University of Pennsylvania built ENIAC, completed in 1945 and shown to the public in 1946. It was much more general in purpose and is often the machine people mean when they say the first electronic computer, though that label depends on how the term is defined.
Because the definitions differ (electronic or not, programmable or not, general-purpose or special-purpose), different books name different machines. The question of who counts as the computer inventor has no single clean answer, and the honest picture is of collaborative invention across countries, with wartime secrecy hiding some of the work.
The stored-program idea
Early electronic machines were mostly set up for each problem by rewiring or adjusting switches, which could take days. A major conceptual shift was the stored-program concept: keeping the instructions in the same kind of memory as the data, so that a machine could change programs quickly and even modify its own instructions.
A 1945 draft report on a proposed machine called EDVAC, associated with John von Neumann and discussed within the Penn group, spread this idea widely. The thinking behind it drew on contributions from several people, and credit has been debated ever since. The theoretical groundwork laid by Alan Turing in the 1930s on what a general computing machine could do is often mentioned alongside this period, though how directly it shaped the hardware designs is still discussed by historians.
The first working stored-program computers appeared in the late 1940s. The Manchester “Baby” ran a program in 1948, and the Cambridge EDSAC began operating in 1949. Both were built at British universities, and both belong in any account of this period.
Memory was the bottleneck
Fast arithmetic was no use without somewhere to keep numbers and instructions. Early designers tried several approaches to main memory: mercury delay lines, in which pulses circulated in tubes of liquid; cathode-ray-tube storage that held patterns of charge on a screen; and magnetic drums. Magnetic core memory, which used tiny rings of magnetic material, arrived in the early 1950s and became standard for a time. Each choice affected speed, cost and reliability, and the shape of early machines owed much to what memory was available.
What the 1940s left behind
The 1940s computers were large, expensive, power-hungry and usually served one institution. Most were funded by governments or universities and aimed at military and scientific calculation, such as ballistics tables, codebreaking and physics problems. They were not yet products.
That changed in the following decade, when a handful of designs turned into machines that could be ordered and delivered. The path ran through ENIAC’s descendants, university projects in Britain and the United States, and eventually the first commercial computers. Meanwhile, replacing vacuum tubes with transistors, which began in the 1950s, would lead the technology in a very different direction again.
A cautious summary
Electronic computing did not have a single birthday. It grew out of mechanical calculators, telephone relays, radio and radar electronics, wartime funding and a few unusually creative teams. Naming one winner tends to hide that. A fair account says that several groups contributed important pieces in the 1930s and 1940s, some of them in secret, and that the machines which emerged by the end of the decade made the later history of computing possible.