The History of the Transistor

By Victor Hoang · Updated Oct 24, 2026

The problem was the vacuum tube. It worked, but it was hot, fragile and power hungry. The machines built from it showed the cost. Penn's engineering school says "the chief challenge was tube reliability" in ENIAC, the Electronic Numerical Integrator and Computer.

The transistor was first made to work by John Bardeen and Walter Brattain at Bell Labs on 16 December 1947 and announced to the public on 30 June 1948. It has been shrinking, changing material and multiplying ever since. Nobody set out to build a computer part. Telephone engineers wanted an amplifier that didn't burn out, and a 70-year trail of crystal rectifiers, failed patents and wartime radar work finally gave them one. ENIAC used about 18,000 tubes and drew roughly 150 kilowatts. For the wider story of the field, see The History of Computers, the computer timeline and who invented the computer.

Two terms first. A semiconductor is a material, such as germanium or silicon, whose ability to carry current can be changed by adding tiny amounts of impurities. A transistor is a semiconductor device with three connections, in which a small signal on one controls a larger current between the other two. That lets it amplify a signal or act as a switch.

Timeline at a Glance

The Crystal Years (1874-1939)

A crystal rectifier and two patents nobody could build

In 1874 Ferdinand Braun, a 24-year-old University of Berlin graduate studying crystals at Würzburg University, tested galena, a lead sulfide crystal, with the point of a thin metal wire. Current flowed one way only. The Computer History Museum calls his note the first written description of a semiconductor diode, and says he demonstrated it in Leipzig on 14 November 1876. It had no use until radio arrived. Then the "cat's whisker" crystal became the detector in early radio sets.

The people who made it practical were radio men. Jagadis Chandra Bose in Calcutta filed a US patent on 30 September 1901 for a point-contact rectifier, and Greenleaf Pickard, an AT&T engineer, tested thousands of minerals from 1902 to 1906 and filed a silicon detector patent in 1906. These were diodes, one-way valves. They could not amplify, and they were precursors, not transistors. Vacuum tubes replaced them in most radio sets by the mid-1920s.

The idea of a solid-state amplifier appears next, on paper. Julius Lilienfeld filed a patent on 8 October 1926 for a device with a copper sulfide film and a third electrode, meant to replace thermionic tubes with no heated filament and no vacuum. He filed a second in 1928. In 1935 Oskar Heil, who had worked at Cambridge, filed a similar British patent, granted the same year. The museum is careful here: both patents were issued, but no records show either man built a working device. In 1938 Robert Pohl and Rudolf Hilsch at Göttingen got a three-electrode crystal to amplify very low frequencies, which led nowhere.

By December 1939, William Shockley of Bell Labs had written in his notebook that a semiconductor amplifier seemed possible in principle. The next era is the story of why it took eight more years.

For teams today: an idea can be right and still be early. Lilienfeld's patents describe the right category of device decades before materials existed to make one.

The Bell Labs Push (1940-1947)

Radar purifies the materials and a Bell Labs lead assembles a team

The turning point is a war. Radar receivers needed rectifiers that worked at microwave frequencies, which tubes could not handle, so the Allies made millions of crystal diodes from silicon and germanium. The Radiation Laboratory at MIT drove purification work at the University of Pennsylvania and DuPont, and the museum says silicon of 99.999 percent purity was available by war's end. Karl Lark-Horovitz's group at Purdue worked on germanium.

At Bell Labs in Holmdel, New Jersey, the chemist Russell Ohl was testing silicon radar detectors. On 23 February 1940, according to the Computer History Museum, a slab with a seam through it jumped in current under bright light, and the two sides behaved oppositely. The seam separated a phosphorus-rich side from a boron-rich side. The p-n junction, the structure behind most later transistors and solar cells, had been found by accident in a purity experiment.

Peace brought a deliberate attempt. In early 1945, encouraged by executive vice president Mervin Kelly, Shockley began organizing a solid-state physics group at Bell Labs. That April he conceived a field-effect amplifier. It did not work. Around 1946, the theorist John Bardeen suggested that electrons trapped on the semiconductor's surface were blocking the electric field from entering, and he and the experimenter Walter Brattain began studying these "surface states."

On 16 December 1947, the two of them pressed two closely spaced gold contacts, held by a plastic wedge, against a slab of high-purity germanium. The signal came out amplified up to 100 times. They demonstrated it to lab officials on 23 December, and Shockley, who had not been part of the breakthrough, called it "a magnificent Christmas present". The museum adds that Brattain used Purdue's high-purity wartime germanium for the first point-contact transistor.

This era ended with a working device and a new problem. The point-contact transistor was delicate, and nobody was sure it could be made in volume.

For teams today: the breakthrough came from combining a theorist and an experimenter on one problem, with materials built for a different purpose. Check what your adjacent teams have already purified.

The Announcement and the Rivals (1948-1951)

A press conference, a jealous inventor and a better design

The turning point is a date with three versions. The device was invented on 16 December 1947, patent-filed on 17 June 1948 and granted on 3 October 1950 to Bardeen and Brattain, and announced publicly on 30 June 1948. At the New York press conference a Bell spokesman said it "may have far-reaching significance in electronics and electrical communication". Electrical engineer John Pierce supplied the name "transistor."

Almost at once, Shockley moved to replace it. On 23 January 1948, per the museum, he conceived a "distinctly different transistor" built on Ohl's junction. He was partly driven by professional jealousy over having missed the December breakthrough, and he doubted the point-contact design could be mass produced. His theory held that charge carriers could move through the bulk of the germanium, not only along its surface, and John Shive confirmed it on 16 February 1948. Shockley filed his junction transistor patent on 26 June 1948, granted 25 September 1951.

The invention was not made once. Two German physicists in Paris, Herbert Mataré and Heinrich Welker, had worked on radar crystals during the war. Mataré saw amplification in early 1948 and the pair independently built a "transistron", learning about Bell Labs' version about a month after June 1948. By mid-1949 thousands were in the French telephone system.

Shockley's design needed purer, more uniform crystal than anyone could buy. Chemist Gordon Teal built crystal-growing equipment himself with John Little and Ernest Buehler, using a method based on Jan Czochralski's 1917 work. In April 1950, Morgan Sparks and Teal dropped impurity pellets into the melt to form n-p-n layers, and Bell Labs announced grown-junction transistors on 4 July 1951. Shockley called the crystal growing "the most important scientific development in the semiconductor field in the early days," the museum reports.

The era gave the world a design it could manufacture and left the question of who would be allowed to make it.

For teams today: a first invention is rarely the one that scales. Shockley's rival design, and the crystal-growing method behind it, did the manufacturing work.

The Licensing Years (1952-1958)

Bell Labs shares the recipe and silicon takes over

The turning point is a policy of sharing. In April 1952, Bell Labs held a nine-day symposium for more than 100 people from 40 companies that had each paid a $25,000 licensing fee, including GE, RCA, Texas Instruments and Sony. Attendees took home a book of proceedings they nicknamed "Ma Bell's Cookbook."

The first buyers were not computer makers. In 1952 the Sonotone hearing aid, at $229.50, used one transistor and two tubes, and by 1954 the museum says 97 percent of hearing aids used only transistors. In October 1954 Texas Instruments and Regency launched the TR-1 radio at $49.95, with four TI junction transistors; it sold over 100,000 units and made "transistor" a household word. Computers came next. On 16 November 1953, Manchester University demonstrated a prototype with 92 point-contact transistors, and in 1954 Bell Labs built TRADIC for the US Air Force with about 700 transistors. For the machines these replaced, see the history of ENIAC and UNIVAC and the first commercial computers; for the next stage of computing design, see generations of computers.

Material and process changed as fast as the products. Morris Tanenbaum made a silicon transistor at Bell Labs in January 1954, but Bell did not pursue it. Gordon Teal, who had moved to Texas Instruments, led a team that made its first silicon transistor on 14 April 1954 and announced silicon devices as in production on 10 May. Silicon could handle heat that germanium could not. Meanwhile, Bell Labs chemists learned to diffuse impurities into silicon at high temperatures, and in 1955 Carl Frosch and Lincoln Derick found, after a furnace accident, that a glassy oxide layer could mask the wafer. That oxide would turn out to be the key to everything after.

In September 1955, Shockley agreed with Arnold Beckman to found a semiconductor laboratory in Mountain View, California. In December 1956 he shared the Nobel Prize in Physics with Bardeen and Brattain "for their researches on semiconductors and their discovery of the transistor effect". The next September, eight of his employees resigned and founded Fairchild Semiconductor, which later spun off AMD and Intel. Their story continues in the history of the integrated circuit.

This era made the transistor a product and a regional industry. The open problem was wiring them together.

For teams today: Bell Labs gave its recipe to competitors and the market grew faster than any one company could have grown it. Know when sharing know-how expands your own demand.

The MOS Turn (1959-1971)

A glass layer fixes the surface and the transistor learns to be a chip

The turning point is the old problem of the surface. The field-effect devices Lilienfeld and Heil had patented failed for decades because of surface states, the same trapped charge Bardeen had identified in 1946. At Bell Labs, M. M. "John" Atalla and Dawon Kahng found that thermally grown silicon dioxide cut those states at the silicon surface and demonstrated the first working MOS field-effect amplifier. The museum gives 1960 in its summary and 1959 for the first successful device, so the year is one or the other. Its name is its structure: a metal gate, an oxide insulator and a silicon channel.

Bell Labs did not push it. It was slow and did not answer urgent phone-system needs, and in a 1961 memo Kahng noted its "ease of fabrication and the possibility of application in integrated circuits." RCA and Fairchild researchers pursued it, and MOS devices reached the market in 1964. The Lilienfeld idea had been right in 1926. It took the oxide layer to make it real.

Around it, transistors were being joined onto chips. Jean Hoerni's planar process kept the oxide on the wafer, and Fairchild's first working monolithic circuits came in May 1960. Gordon Moore's 19 April 1965 article projected 65,000 components per chip by 1975 and later became "Moore's Law." Intel's 4004 followed in 1971, with 2,300 transistors designed by Ted Hoff and Stanley Mazor and implemented by Federico Faggin and Masatoshi Shima. The transistor was no longer the part. The chip was.

The museum says that today over 99 percent of microchips use MOS transistors, because they are smaller and use less power than bipolar devices. Software rode this hardware: see the history of operating systems.

The era moved the transistor from one part to millions on a chip. The next one was about keeping the shrinking going.

For teams today: a technology that your own lab sets aside can become the standard elsewhere. Kahng's memo saw the use that Bell's own phone business could not.

The Billions Era (1970s-present)

Planar scaling hits its limit and the transistor turns three-dimensional

The turning point is a leak. As planar MOS transistors shrank, they became hard to turn fully off. Chenming Hu, in the IEEE Spectrum account, describes it as "very difficult to turn off a transistor when it's very small." Engineers kept voltages high to preserve the on/off gap, and because power rises with the square of voltage, the cost became too much. In 1996 DARPA issued a request for transistors below 25 nm. Hu, Tsu-Jae King Liu and Jeff Bokor won the contract in 1997, and in 1999 Xuejue "Cathy" Huang built a working FinFET, a transistor whose channel stands up as a fin and is controlled on several sides.

Industry took years to follow. The 2002 International Technology Roadmap for Semiconductors listed FinFET and ultrathin-body silicon-on-insulator as likely successors to the planar MOSFET. In May 2011 Intel announced that its 22 nm chips would use three-dimensional "tri-gate" transistors, a FinFET variant, which IEEE Spectrum reports cut the smallest features from 32 nm to 22 nm and halved power consumption.

The scale is now hard to picture. Intel's 4004, the chip at the start of the history of the microprocessor, held 2,300 transistors in 1971. NVIDIA's Blackwell GPU, announced on 18 March 2024, contains 208 billion on a custom TSMC process. A 1971 CPU and a data-center accelerator are different kinds of part, but the ratio is about 90 million to one. The same silicon underlies AI chips and the data centers behind the history of cloud computing.

The era is still being written. Its open problem is the physical limit: each step now needs a new device structure, not just a smaller one.

For teams today: when the easy lever stops working, the next gain comes from changing the structure. Hu's team had to stop shrinking a flat design and stand it up.

Key People

Key Facts

• Bardeen and Brattain's patent was filed 17 June 1948 and granted 3 October 1950; Shockley's junction patent was filed 26 June 1948.
• The first device amplified a signal up to 100 times and was shown to Bell Labs officials on 23 December 1947.
• Forty companies paid a $25,000 licence to attend Bell Labs' April 1952 symposium.
• The Regency TR-1 sold for $49.95 and over 100,000 units.
• The 1956 Nobel Prize in Physics was shared in thirds by Shockley, Bardeen and Brattain.
• NVIDIA's Blackwell GPU has 208 billion transistors; Intel's 4004 had 2,300.

The Pattern Behind the History

The same five moves repeat in the transistor's history. We call the sequence the Crystal to Chip Cycle. It is our synthesis of the history, not a research finding.

1. A patent runs ahead of the materials. Lilienfeld and Heil described field-effect devices that nobody could build.
2. A side project supplies the purity. Radar work gave Allied labs the silicon and germanium that made 1947 possible.
3. The first design is replaced by a manufacturable one. The point-contact device gave way to Shockley's junction design.
4. Sharing grows the market. Bell Labs licensed its recipe, and consumer products arrived within two years.
5. The surface keeps deciding the winner. Surface states blocked the early devices, and the oxide layer solved them for MOS.

What It Teaches Teams Today

• Separate the invention from the product. The transistor was invented in 1947, announced in 1948 and made in volume only after 1951.
• Let two disciplines meet. A theorist and an experimenter produced what a single specialty had missed for 20 years.
• Do not assume your best design is your final one. Shockley's rival design beat the original.
• Share what expands your market. Bell Labs' licensing set up the industry that bought its parts.
• Ask what your own lab ignores. Atalla and Kahng's MOS device was set aside at Bell and became the standard elsewhere.

The transistor shows how a stubborn reliability problem, a wartime supply of pure crystals and a few people willing to share a recipe turned a laboratory curiosity into the basic unit of computing. When your own system seems stuck, look for what a neighboring field has already purified.