When Was the First Computer Made? The 2,000-Year Answer Most Timelines Get Wrong

The first computer was made around 100 BCE: the Antikythera mechanism, a bronze gearwork device Greeks built to predict eclipses and planetary positions. If "computer" means a programmable, general-purpose machine, the answer is 1941 (Konrad Zuse's Z3) or 1945 (ENIAC), depending on which definition you accept. No single date is correct, because the answer tracks whatever definition you apply.

That ambiguity is not wordplay. It is the central fact about the history of computing. Every "first computer" claim rests on an implicit definition, and the definitions conflict. This explainer walks through the mechanisms, the dates, the edge cases, and the misconceptions, so you can answer the question precisely for whatever definition you actually care about.

What "computer" has meant across history

For most of the word's life, "computer" was a job title, not a machine. From roughly the 1600s into the mid-1900s, a computer was a person who performed calculations, often in teams, often women, often for astronomy, artillery, or insurance. The Oxford English Dictionary traces the machine sense to the late 1800s, but it did not displace the human sense until the 1940s.

This matters because the question "when was the first computer made" silently assumes the machine sense. Once you accept that, you still face a second split: does the device need to be programmable, or merely automatic? A tide-predicting analog machine computes without being programmable. A programmable machine follows stored instructions. These are different inventions, and they arrived at different times.

A third distinction is general-purpose versus special-purpose. The Antikythera mechanism was exquisitely complex but dedicated to one task: celestial prediction. ENIAC was built for artillery firing tables, yet it was rewired into a general-purpose machine. Generality, not complexity, is what most historians treat as the dividing line.

How the earliest computing mechanisms actually worked

The Antikythera mechanism, recovered from a shipwreck in 1901 and dated to roughly 100 BCE, used at least 30 interlocking bronze gears. A hand crank drove a train of gears that modeled the Metonic cycle (19 years), the Saros eclipse cycle, and the irregular motion of the Moon. The genius was mechanical: a pin-and-slot coupling approximated the Moon's variable speed, a real astronomical effect, without any electronics.

Its mechanism is a form of analog computation. Each gear ratio encodes a mathematical relationship, and turning the crank evaluates that relationship continuously. The device did not store a program. It embodied one in metal. That is why some historians call it the first analog computer and others call it an astronomical calculator, and both are defensible.

Programmability arrived much later, and it came from textiles. Joseph-Marie Jacquard's loom, demonstrated in 1801, used punched cards to control which threads were lifted. The cards encoded instructions, and the machine executed them. Charles Babbage saw the loom and adapted the idea: his Analytical Engine, designed from 1837 but never completed in his lifetime, specified punched-card input, a separate "mill" for arithmetic, and a "store" for memory. That architecture is recognizably a computer, which is why Babbage's design, not his hardware, earns the credit.

The 1940s: where most timelines converge

The modern answer usually lands in a five-year window. The table below shows the machines that compete for the title, with the defining feature each one adds.

ENIAC is the machine most general audiences mean. It contained roughly 18,000 vacuum tubes, weighed about 30 tons, and filled a room of about 1,800 square feet. It could perform roughly 5,000 additions per second, thousands of times faster than a human computer. But ENIAC's "program" was a patch-cable configuration, so changing tasks meant physically rewiring the machine, sometimes over days.

The stored-program concept, usually credited to John von Neumann's 1945 draft report, put instructions in the same memory as data. The Manchester Baby proved the idea in practice in 1948. That shift, not raw speed, is why historians often treat 1948, not 1945, as the birth of the computer as we know it.

The counterintuitive detail: ENIAC's first major job, a hydrogen bomb feasibility calculation in 1945, was commissioned before the machine was fully documented, and the programmers were six women who learned its wiring from circuit diagrams. Their work was treated as clerical support for decades.

Edge cases that break every simple answer

Mechanical calculators precede all of the above. Blaise Pascal built a working adding machine in 1642, and Gottfried Leibniz improved the design around 1673 with a stepped drum that could multiply. These are automatic calculators, not computers, because they execute a single fixed operation. The line is programmability.

Analog computers form a parallel lineage. The Antikythera mechanism, the differential analyzer built by Vannevar Bush in 1931, and the mechanical tide predictors used into the 1960s all compute continuously rather than digitally. If you define computer as "a device that computes," the first one is ancient. If you require digital logic, the date moves to the 1940s.

The Z3 deserves special attention because it precedes ENIAC by four years. Konrad Zuse built it in Berlin in 1941 using about 2,000 telephone relays. It was programmable via punched film and fully automatic. It lacked conditional branching, which limits its generality, and it was destroyed in a bombing raid in 1943. Its wartime isolation kept it out of most Western histories for decades.

Red-flag warnings for anyone trying to settle the debate:

Any single date claims more precision than the evidence supports; definitions, not facts, decide the answer. "First electronic" and "first programmable" are different titles held by different machines. Wartime secrecy delayed recognition of Colossus and the Z3 by decades, so publication dates mislead. Replicas and reconstructions (the Z3 rebuilt in 1960, the Antikythera reproductions) are not the originals. Human "computers" performed the same calculations earlier, so "first computer" depends on whether people count.

Misconceptions worth correcting

The most common error is treating the abacus as a computer. The abacus, used in Mesopotamia by roughly 2400 BCE, is a calculating tool. The user does the arithmetic; the device stores state. No mechanism executes instructions. Calling it a computer stretches the word past usefulness.

A second error is assuming the first computer was electronic. The Z3 was electromechanical, and the Harvard Mark I used electromechanical relays and rotating shafts. Electronics mattered for speed and reliability, but programmability came first.

A third error is placing the first computer in the 1970s with the personal computer. The PC is a packaging and affordability milestone, not a first. The microprocessor (Intel 4004, 1971) made computers small, but the architecture dates to the 1940s.

A subtler error is crediting a single inventor. The stored-program design emerged from conversations among von Neumann, J. Presper Eckert, John Mauchly, and others at the Moore School. The 1945 draft report bore von Neumann's name alone, which distorted credit for decades.

How to answer the question for your own purpose

If you need a defensible answer, pick the definition first, then the date. The steps below map common purposes to defensible answers.

If the audience is general and the goal is a memorable fact, lead with the Antikythera mechanism, about 100 BCE, and call it the oldest known analog computer. If the goal is the first programmable digital machine, use the Zuse Z3, 1941. If the goal is the first general-purpose electronic computer, use ENIAC, 1945. If the goal is the first stored-program electronic computer, use the Manchester Baby, 1948. Always name the definition in the same sentence as the date; the definition is the answer. When citing a source, check whether it distinguishes analog from digital and programmable from automatic.

Threshold guidance, reconciled: treat "computer" as requiring programmability unless your audience expects ancient history, in which case say "analog computer" explicitly. For school contexts, ENIAC is the safest answer. For history-of-technology contexts, the Antikythera mechanism and the Z3 both earn mention. These are conventions, not facts, so state them as conventions.

The timeline below compresses the sequence into one view.

Limits of the evidence

The Antikythera mechanism survives as fragments, roughly a third of the original, and its reconstruction relies on imaging and gear-train modeling rather than a complete artifact. The Z3 was destroyed, so its specifications rest on Zuse's documentation and a later reconstruction. ENIAC's early history was shaped by patent disputes, which colored the published record.

Definitions also drift. As quantum and neuromorphic machines enter the field, the boundary of "computer" will shift again, and the first of those categories may already be contested. The honest position is that the question has a small set of correct answers, each tied to an explicit definition, and no single answer outranks the others without that definition stated aloud.