THE INVENTION. Kilby proved a whole circuit could live in one piece. Noyce proved how to print one. The lawyers took the next ten years.
HERE IS A SMALL JOB. IT SHOULD TAKE ABOUT A MINUTE.
Four tiny parts in front of you — a transistor, two resistors, a capacitor — and a board to hold them. Every part has metal legs. The circuit only works if each leg is joined to exactly the right partner. So you pick up a soldering iron and connect the first pair. Then the next. Then the next. When you finish, you have a working circuit.
Now do it a thousand times. Now shrink the parts to the size of rice grains, then sand, then dust. The legs get thinner than a hair, the wires thinner still, and somewhere in that tangle one joint is slightly loose — and the whole machine quietly stops working.
That is not a thought experiment designed to make you feel clumsy. In the late 1950s it was one of the hardest walls in electronics.
Engineers already knew how to make better parts. What they could not do was connect enough of them.
So somebody asked a different question: what if you didn't wire the circuit after making the parts — what if you made the connections as part of the same process that made the circuit?
That question is the microchip.
IN JUNE 1958, A VICE PRESIDENT AT BELL LABS NAMED JACK MORTON GAVE THE PROBLEM ITS NAME.
He called it the tyranny of numbers. Systems needing hundreds, thousands, even tens of thousands of devices were running into it — because every connection was made by a person, and every joint had to be inspected, supported and kept from touching its neighbours.
The arithmetic is brutal. Assume a solder joint works 99.999% of the time. A machine with 100,000 joints still has at least one bad joint about six times out of ten. Better parts didn't help — the more parts you used, the more the joints, not the parts, decided whether the machine worked.
What changed everything was a switch. A transistor is an electrically controlled switch — a small voltage decides whether current flows. Chain them one way and the bulb answers "are both closed?" Chain them another and it answers "is at least one closed?" The arrangement does the calculating. Nobody does arithmetic.
December 1947, Bell Labs: John Bardeen and Walter Brattain, in a group led by William Shockley, built the first working transistor — a piece of germanium doing a vacuum tube's job. ENIAC, before it, had needed about 17,000 tubes and a room.
But a transistor had legs too. Every leg still needed a wire.
TEXAS INSTRUMENTS, 1958. A NEW ENGINEER NAMED JACK KILBY HAD JOINED IN MAY.
TI was working on the Army Signal Corps' micromodule program — stacking little wafers of parts into compact blocks. Kilby thought it was a half measure, because it made things smaller without reducing a single part or a single joint.
His idea was to remove the reason for the wires. Resistors and capacitors were cheap parts made of other materials — so why not make every part out of the same semiconductor, and carve them all from one piece? A transistor is already layers of treated semiconductor. A resistor is just a strip that resists current — longer and thinner resists more. A capacitor can be the boundary between two differently treated layers.
Late July 1958: he sketched it in his notebook.
Then he had to prove it, and he had just joined — so he was one of the few people not on vacation. The laboratory was nearly empty. Whatever he wanted to try, he built himself, by hand.
SEPTEMBER 12, 1958. COMPANY EXECUTIVES AROUND A BENCH.
On the bench: a sliver of germanium about seven-sixteenths of an inch long, glued to a glass slide to keep it flat. Kilby had etched it into a transistor, resistors and a capacitor, and wired it to an oscilloscope. He switched on the power — and a line on the screen began to rise and fall in a smooth, steady, repeating wave.
An oscillator. A circuit that makes a repeating signal — the steady heartbeat radios and clocks run on. A week later he showed an amplifier. For the first time, a complete working circuit had been made from a single piece of material.
Now look closer at that sliver, because this is where the story turns. The parts were in one piece — but they were still joined by wires. Hair-thin gold wires arched from one spot to the next, each attached by hand. They're called flying wires, because they actually flew over the surface. Kilby said years later that if he had known how long he'd be showing it off, he'd have tidied it up.
So the idea survived and the problem only moved. Instead of wiring separate parts, somebody still had to wire separate regions of one slab, one wire at a time.
THE REST OF THE ANSWER CAME FROM A DIFFERENT DIRECTION, AND IT STARTED WITH DIRT.
1957: eight young engineers left a semiconductor lab run by William Shockley and, with backing from a camera company, started Fairchild Semiconductor. Robert Noyce, Gordon Moore, Jay Last and a Swiss-born physicist named Jean Hoerni. They bet on silicon instead of germanium.
Their transistors were good enough that one was chosen for the guidance computer of the Minuteman missile. But they had a weakness: the junctions — the important boundaries inside a transistor — were exposed at the surface. A speck of dust, a trace of moisture, a loose flake of metal could short them out.
Engineers opened failed units and found flakes of metal rattling around inside the sealed case. That is a serious problem for a device sitting inside a missile's brain.
Hoerni's answer used something already in the factory. Heat a silicon wafer in oxygen and its surface turns into a thin layer of silicon dioxide — essentially glass. Bell Labs had found the glass could act as a stencil: cut windows in it and impurity atoms enter the silicon only through the windows. The usual practice was then to strip the glass away.
Hoerni asked why remove it. Leave the glass on, and the vulnerable junctions end up sealed beneath a layer that keeps out dust and metal. The finished surface was flat, so he called it the planar process. He wrote it down in December 1957 and had a working planar transistor by March 1959.
And this is why silicon won: its oxide is tough, stable and an excellent insulator. Germanium's oxide is weak and dissolves in water. That single material difference is most of the reason the industry moved.
NOW NOYCE LOOKED AT HOERNI'S GLASS AND SAW SOMETHING ELSE ENTIRELY.
Not a shield. An insulating floor.
JANUARY 1959, a notebook entry: isolate multiple devices on one slice of silicon. Cut small windows in the glass. Spread a thin film of metal across the whole surface — where there's a window, the metal touches silicon; everywhere else it lies on glass, held away from what's beneath. Then remove the metal except for narrow strips, and each strip becomes a flat wire running from one window to another, bonded to the surface.
That is the answer to the question at the top of this page. The wire was no longer something you add after the circuit is made. It was part of the pattern.
Making it real took a team. In 1959 Noyce asked Jay Last to build one, and Last assembled engineers. May 1960: a working prototype — a flip-flop built from four transistors and a handful of resistors, the circuit that remembers one bit by holding one of two states. By September they had a production version, and in March 1961 Fairchild announced a family of them called Micrologic.
PUT THE TWO SIDES BY EACH OTHER, BECAUSE THE DIFFERENCE IS THE WHOLE STORY.
Kilby's slab had its parts in one piece — but the wires were added by hand, and every unit was finished individually.
Fairchild's chip had its parts and its wires drawn onto the surface as a pattern. Adding a connection to Kilby's meant another human task, repeated for every chip. Adding one to Fairchild's meant adding a line to a pattern — and once the pattern existed, light could transfer it again and again across a whole wafer.
That is why the second approach could be manufactured in quantity, and why Texas Instruments soon adopted surface connections too.
And there was a bonus nobody planned: printed connections can be microscopic, and microscopic wires are short. A signal crossing a tiny chip arrives sooner and costs less energy than one travelling along a hand-soldered lead. Shrinking didn't just make it cheaper to build. It made it faster to run.
SO WHO INVENTED THE CHIP?
Kilby filed his patent application in February 1959. Noyce filed in July. Texas Instruments started a patent war. The two companies fought in court for years and settled in 1966 by agreeing to cross-license each other's patents.
Kilby shared the 2000 Nobel Prize in Physics. Noyce had died in 1990 — and Kilby said Noyce would have shared it.
The fairest summary is this: Kilby showed that a whole circuit could live in one piece. Hoerni, Noyce and Last showed how to make it in quantity.
AND NONE OF IT SOLD ITSELF. THE FIRST CHIPS WERE SLOW, REPLACED A HANDFUL OF PARTS, AND COST MANY TIMES MORE THAN THE SEPARATE PARTS THEY REPLACED.
Somebody had to be willing to pay, and the buyers cared most about small, light and reliable. The Air Force's Minuteman 2 used chips. So did NASA's Apollo guidance computer, designed at MIT — built from a single kind of Fairchild chip containing a simple circuit called a NOR gate, which lights its output only when none of its inputs are on.
The project bought roughly 200,000 of them at $20 to $30 each, and the Apollo work was the largest customer for chips through 1965, until Minuteman overtook it.
Orders that size gave manufacturers the reason and the money to keep improving yield — and prices fell.
1965: Gordon Moore, by then Fairchild's research director, wrote down the economics. Processing a wafer costs roughly the same whichever circuit it carries, so putting more components on each chip makes every component cheaper — until crowding in more starts to wreck the yield. He expected the number of components on the cheapest chips to roughly double every year, and later revised it to every two.
1968: Noyce and Moore founded Intel. 1971: the 4004, a processor with about 2,300 transistors. Today's hold tens of billions.
SPLASH'S READ: The story is usually told as an invention. It's really a story about paperwork, and that's why it belongs here.
Two men solved the same problem five months apart, in two states, without knowing about each other. One had the first working circuit and the wrong method. The other had the right method and filed second. The lawyers split the difference in 1966, and the Nobel went to one man because the other died first.
That's the shoe series again. New Balance kept the factories. Vans sold the machine. Yeezy never owned a mold. Here, the invention was real both times — and what decided who got paid was a filing date.
But the sentence that actually matters is technical, not legal. Go back to the wire arching over Kilby's germanium sliver. It marks the exact place where the circuit still depended on a person — someone had to hold it, attach it, and do it again.
No one can hand-wire billions of switches. For a chip to reach that scale, the connection had to stop being an act and become a pattern: flat metal on a glass floor, drawn by light, lined up layer after layer.
The wire never went away. It just stopped being something anyone had to hold.