From counting aids and Babbage's gears to ENIAC, transistors, the web and AI: this guide follows the chain of ideas behind modern computers, explains why "first computer" has no single answer, and shows the numbers, with diagrams and worked examples.
The phone in your pocket can do in a blink what once needed a machine that weighed as much as several elephants and filled a large room. Yet no single person invented the computer. It grew from a chain of ideas, each one solving a problem left behind by the one before: how to calculate without error, how to give a machine instructions, how to make its switches faster, smaller and cheaper, and how to connect the machines together. This guide follows that chain from counting devices to smartphones and artificial intelligence, explains why "who built the first computer?" has no single answer, and shows the numbers behind the story.
Helpful background: How Computers Work explains binary, processors and memory, which makes several milestones below easier to follow.
The word computer originally meant a person who carried out calculations, and it kept that meaning for hundreds of years. Teams of human computers produced navigation tables, astronomical tables and, later, artillery firing tables. Machines were invented to do this work faster and with fewer mistakes.
Calculating aids are very old. The abacus is thousands of years old. The Antikythera mechanism, recovered from an ancient Greek shipwreck in 1901 and dated to roughly the second or first century BCE, used a system of gears to model the movements of the sun, moon and other heavenly bodies. In the 1600s, the French mathematician Blaise Pascal built a mechanical adding machine, and the German mathematician Gottfried Leibniz designed one that could also multiply. These machines did one fixed kind of arithmetic. They could not be given a new set of instructions.
In 1804, Joseph Marie Jacquard introduced a loom in which punched cards controlled the weaving pattern. Where a card had a hole, a hook could pass through, and where it did not, the hook was blocked. Changing the cards changed the pattern without rebuilding the loom. This was an early example of controlling a machine with changeable instructions stored as holes and no-holes.
A hole and a no-hole are two states, just like the 1 and 0 of binary. Suppose one row of a card has 8 positions, and the pattern is: hole, no hole, hole, hole, no hole, hole, no hole, hole. Write 1 for a hole and 0 for no hole: 10110101.
Step 1: Write the place values for 8 bits, from left to right: 128, 64, 32, 16, 8, 4, 2, 1.
Step 2: Add the place values where there is a 1: 128 + 32 + 16 + 4 + 1.
Step 3: 128 + 32 = 160; 160 + 16 = 176; 176 + 4 = 180; 180 + 1 = 181.
So the row of holes stores the number 181. Early punched-card systems did not always use pure binary like this, but the principle of storing information as patterns of holes is the same one that modern computers use with electrical on and off.
In the 1820s and 1830s the English mathematician Charles Babbage designed two machines. The Difference Engine was meant to calculate and print mathematical tables. The much more ambitious Analytical Engine, first described in 1837, was designed with a separate "store" for numbers, a "mill" to do arithmetic, and punched cards to supply instructions. That layout closely resembles memory, a processor and a program. Babbage never completed the Analytical Engine. However, a Difference Engine built in the 1990s from his drawings, at London's Science Museum, worked correctly, which showed that his designs were sound.
Ada Lovelace translated an article about the Analytical Engine in 1843 and added her own long notes, including a step-by-step method for calculating a sequence of numbers known as Bernoulli numbers. This is often described as the first published program written for a general-purpose machine. She also saw that such a machine could work on more than numbers, for example on music, if the symbols could be represented in the machine's terms.
In 1890, Herman Hollerith's punched-card tabulating machines were used to process the United States census, greatly speeding up the counting. His company later became part of the business now known as IBM.
In 1936 the British mathematician Alan Turing described an imaginary machine, now called a Turing machine, which reads and writes symbols on a tape by following a table of rules. He showed that one such machine could imitate any other, which gave a precise meaning to the idea of a general-purpose computer. His work also explored what can and cannot be computed at all. Modern computers are practical versions of this idea.
In the late 1930s and 1940s, several teams in different countries built machines that used relays or vacuum tubes. A relay is a switch moved by an electromagnet. A vacuum tube is a glass bulb that can switch or amplify electric current with no moving parts, which makes it much faster than a relay. Which of these machines was "the first computer" depends on how you define a computer:
| Machine | Years | What it was first (or among the first) to do |
|---|---|---|
| Analytical Engine (design) | 1837 | Design for a general-purpose, programmable machine; never completed |
| Zuse Z3 (Germany) | 1941 | Working programmable, automatic digital computer, built from relays |
| Atanasoff-Berry Computer (USA) | about 1939-42 | Electronic digital calculator for solving equations; not general-purpose or programmable in the modern sense |
| Colossus (Britain) | 1943-44 | Programmable electronic digital machine used for codebreaking; kept secret for decades |
| ENIAC (USA) | completed 1945, unveiled 1946 | Large general-purpose electronic digital computer |
| Manchester "Baby" (Britain) | 1948 | First electronic computer to run a program stored in its own memory |
Historians therefore say "first" only with a qualifier: first programmable, first electronic, first general-purpose, first stored-program. Anyone claiming a single winner is simplifying.
ENIAC, built at the University of Pennsylvania by John Mauchly and J. Presper Eckert, shows the scale of the era. It contained around 17,000 to 18,000 vacuum tubes, weighed about 30 tons, filled a large room, and could perform about 5,000 additions per second, which was far faster than any earlier machine. A team of six women, including Kay McNulty, Betty Jennings and Betty Snyder, taught themselves how to program it. To change its task, they had to rearrange cables and set switches by hand.
ENIAC's team knew that rewiring was a bottleneck. In 1945, a report on a proposed machine called EDVAC, written by the mathematician John von Neumann and building on ideas developed with the Pennsylvania team, described the stored-program design: keep the instructions in the same memory as the data, and let the processor fetch them one at a time. Changing what the computer does then needs only a different program, not new wiring. The ideas came from a team, though the report carried his name and the design is often called the "von Neumann architecture".
The first small stored-program computer to run a program was the Manchester "Baby" in June 1948, followed by the EDSAC at Cambridge in 1949, which was soon used for real scientific work. This design is the basis of almost every computer made since, from laptops to phones.
Programming itself also advanced. In the early 1950s Grace Hopper developed one of the first compilers, a program that translates human-readable instructions into the machine's own code. Languages such as FORTRAN (1957) and COBOL (1959) let people write instructions in a form closer to ordinary language. See What Is Coding? for how this works today.
Every computer is, at heart, an enormous number of tiny on-off switches. Faster, smaller, cooler and more reliable switches meant better computers.
In 1965, Gordon Moore, who later co-founded Intel, noticed that the number of components on a chip was doubling at a steady rate. In 1975 he adjusted his forecast to a doubling roughly every two years. This is called Moore's law, but it is an observation about industry, not a law of nature, and it was achieved only through huge engineering effort. In recent years the pace has slowed because components are approaching physical and cost limits.
Start with the Intel 4004 of 1971: about 2,300 transistors. Suppose the count doubles every 2 years for 50 years, until 2021.
Step 1: The number of doublings is 50 ÷ 2 = 25.
Step 2: Doubling 25 times multiplies the count by 225 = 33,554,432.
Step 3: 2,300 × 33,554,432 ≈ 77 billion transistors.
Meaning: steady doubling turns thousands into tens of billions. That is the same range as many of today's most advanced processors, which contain billions to tens of billions of transistors, so the pattern has held surprisingly well over 50 years. This shows how powerful repeated doubling is. Remember that it describes a trend that industry worked hard to sustain, and that it cannot continue forever.
With microprocessors, a computer no longer needed to fill a room or serve a whole organisation. In 1975 the Altair 8800 kit appeared on the cover of Popular Electronics and attracted hobbyists. In 1977 ready-made machines such as the Apple II, the Commodore PET and the Tandy TRS-80 reached ordinary buyers. IBM introduced its Personal Computer in 1981, and its design was widely copied, which helped make the "PC" a standard.
Early personal computers were controlled by typing commands. The graphical user interface, with windows, icons and a pointer, was developed at Xerox's Palo Alto Research Center in the 1970s, including the Alto of 1973, and brought to a wide public by Apple's Macintosh in 1984 and by Microsoft Windows from 1985. This made computers usable without specialist training. For how operating systems work, see Operating Systems.
In 1969, the US-funded ARPANET sent its first message between computers at two sites in California (UCLA and the Stanford Research Institute), starting with a very small network. Over the following years, networks multiplied, and in 1983 ARPANET adopted TCP/IP, the shared rules that let different networks exchange data. That is why 1983 is often given as the birth of the modern internet.
The World Wide Web is a different thing from the internet. Tim Berners-Lee, working at CERN in Switzerland, proposed it in 1989, and it opened to the public in 1991. It is a system of linked pages that runs on top of the internet. Graphical browsers such as Mosaic in 1993 made it easy for ordinary people to use, and search engines soon followed (see What Is the Internet, Really? and Search Engines & Web Browsers).
Smartphones combined the microprocessor, the graphical interface and the internet in a pocket-sized device. Phones with touchscreens and app stores became mainstream after Apple's iPhone (2007) and the first Android phone (2008). From the mid-2000s, cloud computing let people rent computing power and storage from distant data centres instead of owning every machine (see Cloud Computing).
Artificial intelligence research began in the 1950s, but progress depended on three things arriving together: cheap and powerful hardware, huge amounts of data, and better methods. In the 2010s, a style of machine learning called deep learning achieved large gains, for example in image recognition. In 2017 researchers introduced the Transformer design, which underlies many of today's large language models. In late 2022, chatbots based on such models became available to the general public. This is another step in the same long story: better tools built on earlier layers of hardware, software, networks and data.
These are approximate, rounded figures for comparison.
Weight: ENIAC weighed about 30 short tons, which is roughly 27,000 kg. A typical phone weighs about 0.2 kg. The ratio is 27,000 ÷ 0.2 = about 135,000 times heavier.
Speed: ENIAC could do about 5,000 additions per second. A typical high-end phone processor core runs at around 3 GHz, which means about 3,000,000,000 clock cycles per second. Comparing 3,000,000,000 with 5,000 gives 3,000,000,000 ÷ 5,000 = 600,000.
Reading the result: by raw clock rate alone, one phone core is hundreds of thousands of times faster, and modern chips also have several cores and can perform multiple operations in each cycle, so the real gap is even larger. A "clock cycle" is not exactly the same as an "addition", so this is a rough comparison, not a precise measurement.
Counting the years between some milestones in this article gives a feel for the pace. The choice of milestones is a judgement, so treat this as an illustration.
Observation: the gaps shrink from 108 years to 62, 24 and 19, which is consistent with new technology building on all the earlier layers. But the pairs overlap and were chosen by us, so this is an illustration, not proof of a rule, and it does not mean progress is guaranteed to keep speeding up.
Computing history is often told through a few famous names, but it was built by large teams, and many contributors were overlooked at the time. The human computers who did calculations by hand, the six women who programmed ENIAC, the codebreakers at Bletchley Park whose work stayed secret for decades, Grace Hopper's compilers, and the engineers who made chips smaller each year all shaped the story. A useful habit when reading any "who invented it?" claim is to ask what exactly was invented, and who else was involved.
Try each question first, then tap it to reveal the answer. A scored multiple-choice quiz follows below.
A person who carried out calculations by hand. Machines were built to do that work faster and with fewer mistakes.
Its design had a store for numbers, a mill for arithmetic and punched cards for instructions, which closely resembles memory, a processor and a program.
Early machines such as ENIAC had to be rewired for each new task. Keeping instructions in memory like data meant a new task only needed a new program.
Vacuum tube, then transistor, then integrated circuit (leading to the microprocessor).
10 ÷ 2 = 5 doublings, so 1,000 × 25 = 1,000 × 32 = 32,000.
Because "first" depends on the definition: first programmable, first electronic, first general-purpose or first stored-program. Different machines fit each.
Punched-card activity (no computers needed): give students a strip of card with eight positions and let them punch or mark a pattern, swap with a partner, and decode each other's number using the place values in Worked Example 1.
Doubling demonstration: ask students to fold a sheet of paper in half repeatedly and count how many folds they can make, then compare with the 25 doublings in Worked Example 2.
Who was first? debate: assign groups one machine from the table in section 4 and ask them to argue for what it was "first" at, then discuss why definitions matter.
Research prompt: have students choose one person from section 11 and find out what they contributed, using at least two sources and noting where the sources disagree.
This article is original EDUSAMBAM writing. The milestones follow standard histories of computing, and the details about ENIAC were checked against several accounts, including the Computer History Museum's catalogue record, which gives about 18,000 vacuum tubes, about 30 tons and 5,000 additions per second. Counts of ENIAC's tubes vary slightly between sources (roughly 17,000 to 18,000) and dates such as ENIAC's unveiling are given as 14 or 15 February 1946, so the article says "around" and "1946". The diagrams are original EDUSAMBAM schematics, and the calculations are arithmetic using rounded figures.
Histories of invention are debated and revised as new evidence appears, so check more than one source for any claim of who was "first". This article does not rank companies or products.
10 questions. Select an answer for each, then submit to see your score instantly.