Once transistors existed, the next problem was wiring them together. Circuits built from separate transistors, resistors and capacitors had to be connected by hand, and every joint added cost and a possible point of failure. The notion of putting a whole circuit in one piece of material appeared early. In 1949 Werner Jacobi of Siemens filed a patent for a five-transistor amplifier on a single semiconductor, and in 1952 the British engineer G. W. A. Dummer proposed electronic equipment made as a solid block without connecting wires. [1][2]

Several laboratories pursued the idea through the 1950s. Harwick Johnson at RCA designed an integrated oscillator in 1953, and in 1955 Ian Ross and colleagues at Bell Labs built a four-stage counter on silicon. None of these efforts produced a practical product, but they show that the integrated circuit was less a single flash of insight than a race that several teams were running at once. [1]

The eventual winners of that race were shaped by a company in Mountain View, California. William Shockley, co-inventor of the transistor, founded Shockley Semiconductor Laboratory in September 1955 with backing from Arnold Beckman. It developed Northern California’s first prototype silicon devices and recruited talented young engineers, but many disagreed with Shockley’s management and with his focus on a four-layer diode rather than commercial silicon transistors. [3]

In September 1957 eight of them resigned together and founded Fairchild Semiconductor in Palo Alto. The group was Julius Blank, Victor Grinich, Jean Hoerni, Eugene Kleiner, Jay Last, Gordon Moore, Robert Noyce and Sheldon Roberts. Fairchild became highly successful, and two of its founders would later start Intel. [3]

The first working demonstration came from Texas Instruments in Dallas. On 12 September 1958 Jack Kilby showed an oscillator in which a transistor, capacitor and resistors had all been formed in slices of germanium and connected with fine gold wires. Texas Instruments announced the concept, which it called a solid circuit, in March 1959, and a first commercial device followed in March 1960. [2][1]

Kilby filed his patent on 6 February 1959 and it was issued in June 1964. His demonstration proved that all the components of a circuit could be made from the same semiconductor. Its weakness was the flying gold wires: they had to be attached by hand, which limited how many circuits could be built and how complex they could become. [2][1]

The manufacturing breakthrough came from Fairchild. Standard practice at the time was to strip off the oxide layer that formed on a silicon wafer during processing. Jean Hoerni, a Fairchild physicist, proposed keeping it as a protective skin over the transistor’s junctions. He recorded the idea in December 1957, demonstrated a working planar transistor in March 1959 and filed a patent on 1 May 1959. Fairchild put its first planar transistor, the 2N1613, on sale in April 1960. [4]

Planar transistors had far lower leakage currents than earlier designs, and because every step was carried out from one flat side of the wafer, the process lent itself to making many devices at once. The Computer History Museum describes it as possibly the most important innovation in the history of the semiconductor industry, and Fairchild licensed it widely. [4]

Robert Noyce, Fairchild’s co-founder, saw that Hoerni’s oxide layer could do more than protect transistors. Aluminium lines could be deposited on top of it to connect components, removing the need for hand-attached wires. He filed a patent for this monolithic integrated circuit on 30 July 1959, and it was issued on 25 April 1961. Kurt Lehovec of Sprague Electric separately patented a method of isolating components electrically within one chip. [5]

Noyce assigned Jay Last to turn the concept into a product in August 1959. Last’s team produced the first working monolithic devices on 26 May 1960, a flip-flop with four transistors and five resistors designed by Robert Norman. Fairchild publicly announced the first of these chips, under the trade name Micrologic, in March 1961, and Texas Instruments announced its own Series 51 logic family in October 1961. [6]

The combination of Hoerni’s process and Noyce’s interconnect is the direct ancestor of every chip made today. A modern GPU is still, at heart, a planar integrated circuit: devices formed in one side of a silicon wafer and wired together by layers of metal above them.

Fairchild and Texas Instruments fought a long patent dispute over the invention. The courts ultimately ruled in Noyce’s favour, but by then the companies had already reached a commercial settlement based on cross-licensing, with a net payment to Fairchild. [5][1]

History has largely settled on shared credit. Kilby and Noyce both received the National Medal of Science in 1979. In 2000 Kilby received half of the Nobel Prize in Physics for his part in the invention of the integrated circuit, the other half going to Zhores Alferov and Herbert Kroemer for work on semiconductor heterostructures. [1][7]

Fairchild’s greatest legacy may have been the people who left it. By 1971 former Fairchild employees had started 23 semiconductor companies, and more than 100 more followed during the next decade. That year, on 11 January, the journalist Don Hoefler published a three-part series in Electronic News under the headline Silicon Valley U.S.A., the first time the name had appeared in print. [8]

The most consequential departure came in 1968. Noyce and Gordon Moore, who had been Fairchild’s director of research and development, founded Intel on 18 July 1968. The name was a contraction of integrated electronics, and the company began operating on 1 August 1968 from a conference room in a former Union Carbide building in Mountain View, with about a dozen engineers. [9][10]

Intel bet on MOS memory. Its 1101 RAM of 1969 used the silicon-gate process that Federico Faggin and Tom Klein had developed at Fairchild the year before, and its 1103 dynamic RAM of 1970, holding 1,024 bits at about one cent per bit, quickly displaced magnetic cores as main computer memory. In 1971 Faggin and Masatoshi Shima completed the Intel 4004, a 4-bit processor with 2,300 transistors on a single chip, based on an architecture conceived by Ted Hoff and Stanley Mazor. [11][12][13]

The first Micrologic flip-flop held four transistors. Today’s largest AI accelerators hold hundreds of billions, and they are made by descendants of the same planar process and interconnect scheme. The integrated circuit also created the industry’s business model: design and manufacturing costs are front-loaded, so each new generation must be sold in enormous volumes to pay back its development. That economics still drives how quickly new GPUs replace old ones.

For anyone renting or buying accelerated compute, the practical lesson is that chip generations turn over quickly and prices follow. On Kovara you can compare what providers charge for each GPU model, see how older and newer generations are priced side by side, and ask Kova to explain which hardware makes sense for a particular job.