Jack Kilby / The integrated circuit
How do you wire up thousands of parts?
That time, that place
By the late 1950s the transistor existed. The trouble was what came next. A computer needed tens of thousands of transistors, resistors and capacitors, each joined by hand with wire. More parts meant more joints, and more joints meant more failures. People called it the tyranny of numbers.
Why this question
A newly hired engineer at a Texas company had no vacation days yet. Alone in the empty lab, he turned the problem around: if joining separately made parts was the trouble, why not **make every part on one piece of semiconductor** so nothing needs joining?
What was found
In September 1958 he built a transistor, resistor and capacitor together on a fingernail-sized sliver of germanium and switched it on. A waveform appeared on the screen — the first time several parts worked as a circuit on a single piece.
Half a year later an engineer in California arrived at the same idea in a form easier to manufacture: coat the surface with a thin film and print the metal wiring on top. Today's chips descend from that version. Together the two are called the integrated circuit.
Before this, a circuit meant buying parts, plugging them into a board and soldering them together. One part was one object. After the integrated circuit a part is not an object but **a pattern on a single piece**.
A pattern can be printed many times at once. Soon dozens of chips were made together on one wafer, and adding parts barely raised the price. In 1965 came the observation that the number of parts per chip was doubling every year or so — a trend that held for half a century.
Parts per fingernail went from a handful to tens of billions. The engineer with no vacation received the Nobel Prize in Physics forty-two years later. The Californian died before it could be shared, but the company he co-founded opened today's processor industry.