There is no single first
Historians do not agree on one “first digital computer”, and the disagreement is not a gap in the record. It comes from the word “computer”. A machine can be digital, electronic, programmable, general-purpose and able to store its own program, and the early machines each achieved some of these things but not all of them at once. Choose one definition and you get one answer; choose another and you get a different machine.
None of this happened in 1967. By that year, digital computers had existed for about a quarter of a century, and the machines of the late 1960s were a different kind of object from the pioneers described below. The sections that follow set out the main claims, what each machine is credited for, and where 1967 stood.
The main claims
| Machine | When | What it is credited for |
|---|---|---|
| Atanasoff-Berry Computer | prototype 1939, working machine about 1942 | Electronic digital calculation using binary numbers |
| Zuse Z3 | 1941 | A working, automatically controlled, programmable machine built from relays |
| Colossus | 1943 to 1944 | Large-scale electronic digital machine, built for codebreaking |
| ENIAC | completed 1945, unveiled 1946 | A large, general-purpose electronic computer |
| Manchester Baby | June 1948 | Running a stored program on an electronic machine |
Atanasoff-Berry
John Atanasoff and his student Clifford Berry built their machine at Iowa State College. It used vacuum tubes, binary arithmetic and a form of regenerative memory, and it was designed to solve systems of linear equations. It was not programmable in the modern sense, and it was never put to wide use. Its claim rests on being an early electronic digital calculating device.
Zuse
Konrad Zuse, working in Germany largely on his own, built a mechanical machine (the Z1) in the late 1930s and then the Z3, completed in 1941. The Z3 used telephone relays, so it was electromechanical rather than electronic, but it could follow a sequence of instructions read from punched film. The original was destroyed in wartime bombing, and Zuse’s work was little known outside Germany for years afterwards.
Colossus
Colossus was built in Britain by a team that included the engineer Tommy Flowers, for the codebreakers at Bletchley Park, who were attacking German teleprinter ciphers. The first machine began working at the end of 1943 or early in 1944, and an improved version followed in mid-1944; reportedly around ten were in use by the end of the war. It was electronic and digital and could be set up for different tasks with switches and plugs, but it was built for one family of problems. Because the project was kept secret until the 1970s, its role was not widely understood for decades.
ENIAC
The ENIAC computer was built at the University of Pennsylvania by J. Presper Eckert and John Mauchly for the US Army, originally to calculate artillery firing tables. It used roughly 17,000 vacuum tubes and was shown to the public in February 1946. It was programmed by setting switches and rearranging cables, which could take a long time. For many years it was treated as the first electronic computer, partly because Colossus was secret.
The credit was later disputed in court. In a 1973 US federal ruling in the Honeywell and Sperry Rand case, the main ENIAC patent was declared invalid, with the judge finding that some of its ideas had been derived from Atanasoff’s earlier work. Historians still weigh the competing contributions differently.
The Manchester Baby
On 21 June 1948 a small experimental machine at the University of Manchester, built by Frederic Williams, Tom Kilburn and colleagues, ran its first program. It was tiny, but it held its program in electronic memory alongside its data, which is how nearly every computer since has worked. The stored-program idea had been discussed in the mid-1940s, including in the EDVAC report associated with John von Neumann, and the Baby was the first working demonstration of it on an electronic machine. Cambridge’s EDSAC, running in 1949, followed soon after.
Choosing a definition
If “digital and electronic” is the test, Atanasoff-Berry and Colossus come first. If “programmable and working” is the test, Zuse’s Z3 is a strong candidate. If “general-purpose electronic” is the test, ENIAC. If “stored program” is the test, the Manchester Baby. A careful answer names the criterion before naming the machine.
What computing looked like by 1967
Twenty years after the Baby, computing had become an industry. The IBM System/360, announced in April 1964, was a family of machines of different sizes that could run the same software, which let customers upgrade without rewriting everything. Such IBM mainframes handled payroll, billing and record-keeping for large companies and government agencies. At the other end of the scale, minicomputers such as the DEC PDP-8 were cheap enough for a laboratory or a factory floor.
Smaller components made much of this possible. Transistors had replaced vacuum tubes, and integrated circuit chips were beginning to appear in new designs. Computers also guided spacecraft in the space race, and in the United States an early ARPANET plan for linking computers was being drawn up in 1967. Banking was starting to change too, with the first cash machine appearing in London in June 1967.
None of these machines was “the first”. They show how long the road from a handful of experimental machines to everyday business tools actually was, and why the earliest claims are best understood as stages rather than as a single invention.