A Transistor Is a Switch — A Tiny Faucet You Open and Close with Electricity

A Switch Inside a Fingernail-Sized Chip

Mobile news frequently features stories about chips. The headlines read something like "New chip, faster and with longer battery life." That chip is a fingernail-sized piece, and the part inside that handles calculations is a switch that lets electricity flow or stops it.

A transistor is a tiny switch turned on and off by electricity. A wall light switch is pressed with a finger, but a transistor is pressed with electricity. A single chip today contains an uncountable number of these switches.

When we think of a switch, a light switch comes to mind first, but this article explains it using a single water faucet. A faucet has a water channel, a stopper that blocks the water channel, and even water that drips stubbornly when fully closed. Everything needed to understand a transistor is contained in these three. Water is the flow of electricity, namely electric current. We will carry this faucet all the way to the end of the article.

A single transistor is a single water faucet opened and closed by electricity. A chip is an uncountable number of those faucets woven together.

A Single Faucet — The Watercourse and the Stopper

To build a faucet, you first need materials. The material for a chip is silicon, which is silicon obtained from sand. Silicon is a semiconductor. A semiconductor is a material somewhere between a metal, which conducts electricity well, and glass, which hardly conducts it at all, and it either conducts or blocks electricity depending on conditions. It must be able to be made to conduct or block to become a switch.

Pure silicon does not conduct electricity well if left as it is. So it is tweaked. Doping is the process of mixing tiny amounts of other atoms into silicon to change how well it conducts electricity. Depending on the type of atoms mixed, the particles carrying electricity become either electrons or empty spaces where electrons used to be. Either way, mixing them makes electricity flow better.

A transistor divides a single piece of silicon into three zones. The end where current enters and the end where it exits are doped so electricity flows well. In the middle between them, atoms of the opposite type from both ends are mixed. Then a threshold forms at the boundaries where both ends meet the middle, so under normal circumstances, the particles carrying electricity cannot cross that boundary. To put it in terms of a faucet, there is an intake pipe and an outflow pipe, with the watercourse between them blocked. The channel is the path through which electric current passes between the two ends of the transistor. It is the watercourse of the faucet.

Now a stopper is needed. The gate is a piece placed over the channel with a thin film in between, and the voltage applied to it opens and closes the channel. When no voltage is applied to the gate, the channel is blocked and almost no current flows. When voltage is applied, particles carrying electricity gather in the blocked middle to form a thin path, and current flows between the two ends.

The thin film between the gate and the channel is an insulating layer laid down to block electricity (DOI: 10.1088/0034-4885/69/2/r02). Therefore, almost no current flows into the gate itself. In terms of a faucet, turning the handle uses no water. The usefulness of this switch is that a large flow can be opened and closed with a small force.

There is one spot where the analogy goes awry. A real faucet has a moving stopper that opens the watercourse. A transistor has not a single moving part, and the voltage of the gate creates and erases paths in the channel. Even when this article says the stopper "presses" or "wraps around" the watercourse, nothing is moving. It means the force of the voltage applied to the gate reaches the channel from one side, from three sides, or from all sides.

A transistor is a switch that opens and closed a blocked path between two ends using the voltage applied to the gate.

0 and 1, and Small Judgments

There are two things you can do with a single faucet: water either flows or it does not. The chip attaches the numbers 1 and 0 to these two states. In this article, we will count water flowing as 1 and not flowing as 0. Photos, music, and the letters of this article are all converted into long strings of 0s and 1s inside the chip.

Using two faucets creates a judgment. Let us connect two faucets in series on a single water channel. Water comes out at the end only when both the front and back faucets are open. If even one is closed, it does not come out. This is the judgment "both must be open."

This time, we split the water channel into two branches and place faucets side by side. In this case, water comes out even if just one of the two is open. This is the judgment "if even one is open, it comes out." The two examples are illustrations created by this article, but circuits inside a chip also place switches in series or side by side to make such judgments.

Stacking these small judgments endlessly becomes addition, and stacking additions becomes calculation. What a chip does is ultimately a matter of countless switches turning on and off by receiving each other's results. Therefore, the more switches there are, the more things a chip can do at once.

The first transistor came out of Bell Labs in the United States in November 1947 (DOI: 10.1109/4.643644). What is inside a chip is a different method—namely, opening and closing the channel with the gate's voltage, as we saw earlier. A MOSFET is a transistor that opens and closes a channel using the voltage applied to the gate, and it gets its name from the layered structure of metal, oxide, and semiconductor.

The conception of the MOSFET—the idea of changing how well a semiconductor conducts electricity using the power of electricity applied from the side—first appeared in a patent application in 1928. However, due to problems with the silicon surface, a usable device did not emerge for some time, and that problem was solved in 1958. Afterward, computers needed such switches, and the technology of packing multiple components onto a single piece paved the way. That is how the MOSFET became the basic component of digital devices (DOI: 10.1049/esej:19980509).

When the path connected by switches is open it is 1, and when blocked it is 0, and weaving together a few switches makes a small judgment. The chip stacks those judgments to calculate.

Smaller, Still Smaller — What Was Gained

The number of components in a single chip has doubled and doubled again, following a trendline that started from a single component in 1959 (DOI: 10.1109/tsm.2010.2096437). This flow, which doubles every fixed period, is commonly called Moore's Law. Meanwhile, each individual switch has grown progressively smaller.

Just how many more can fit when things get smaller can be counted using faucets. Imagine packing faucets densely across one side of a wall. If you cut both the width and height of each faucet in half, the area occupied by a single faucet becomes one-fourth. You can fit 4 times as many on the same wall. This arithmetic is the math of this article. Chip factories transition every few years to a new manufacturing method that shrinks the switches by a little bit all at once, and each of these steps is called a process generation. In actual chips, the amount shrunk in a single generation is smaller than this arithmetic (DOI: 10.1142/s0129156417400018).

When things were made smaller, other benefits came along with the count. Along with fitting more in the same area by shrinking the switches, the switches turned on and off faster, and the electricity consumed to turn them on and off once decreased. This does not mean the electricity used by the entire chip decreased. The industry has expected these PPA (performance, power, and area) benefits with every new generation of the process (DOI: 10.1142/s0129156417400018, DOI: 10.54254/2755-2721/2026.ka30036). There was a time when simply making things smaller made everything better.

Those benefits are dwindling. As switches became extremely small, three problems arose. The gate fails to properly block the shortened channel, power concentrates in narrow spaces, and signals are delayed in the wires connecting the switches. These three are eating away at the benefits gained by making things smaller (DOI: 10.54254/2755-2721/2026.ka30036). What happened as the benefit on the power side became blocked is dark silicon. This refers to a state where, due to limits on heat and power, not all the switches inside a chip can be turned on at the same time, and some must be left turned off. The detailed story is followed by the article covering dark silicon.

Making switches smaller allowed more, faster, and less electricity used at once, but as they became too small, those benefits began to shrink.

Even When Tightly Shut, It Drips

An old faucet drips even when turned off tightly. The same thing happens to transistors. Leakage current is the current that leaks even when the transistor is in the "off" state.

The leakage grew larger as the switches became smaller. As a result of continuously shrinking the channel length, the thickness of the film between the gate and the channel, and the voltage at which the transistor starts to turn on, leakage current became a non-negligible share of the power the chip uses (DOI: 10.1109/jproc.2002.808156). In terms of a faucet, the shorter the water path and the thinner the membrane under the plug are made, the more water drips from the closed faucet.

There are several paths for the leak. The first is the flow seeping right through the shut channel itself. To see this flow, one must know one thing first. Even while the switch is off, a voltage is applied between both ends trying to push current across. This is like the water inside a pipe continuing to push against the plug even when the faucet is closed. When the channel is long, this voltage at the exit end cannot reach the spot the gate is blocking. When the channel becomes short, the voltage applied at the exit end affects the spot the plug is blocking, loosening the plug (DOI: 10.1109/jproc.2002.808156). It is akin to the voltage on the opposite side helping to push open a door that the gate alone should be holding shut, so current leaks even when it is turned off. With a structure that places a gate on top of a flat channel, it is difficult to maintain the grip of the gate on the channel as the switches continue to shrink (DOI: 10.1109/soi.2009.5318794).

The second is the flow leaking by piercing through the membrane under the plug. The membrane continued to get thinner. This is because the thinner it is, the better the gate grips the channel (later calculations also show this, DOI: 10.1063/1.3043662, simulation). However, when the membrane used under the gate thinned down to 1.4 nanometers—the thickness of a single piece sliced from a human hair about 70,000 times—the current leaking through the membrane became too large. Therefore, research continued to replace the membrane with another material that could be laid down thicker while maintaining the gate's force just the same (DOI: 10.1088/0034-4885/69/2/r02).

The leaking electricity also comes from switches waiting while turned off. Even if the water dropping from a single faucet is one drop, the situation is different if a building has 1 billion faucets. If all of them drip one drop per second, for the entire building that is 1 billion drops per second. This number is an illustration for this article, intended to show that while the amount leaking from a single switch inside a chip is very small, the number of them is countless. Reducing leakage is especially important in uses where electricity must be conserved, such as mobile devices (DOI: 10.1109/jproc.2002.808156). A 2001 review considers that the limit on how small switches can be made comes mainly from this leaking current, yet concludes that its end point is not a single dot but multiple different points depending on the application (DOI: 10.1109/5.915374).

If you simply make transistors smaller using the same flat structure, the electricity that leaks even when they are turned off increases. The limit to shrinking switches is also considered to come mainly from this leakage.

A Plug Wrapping the Water Channel — From FinFET to Gate-All-Around

A plug that presses down on a water channel from only one side struggles to block the gap far from the plug as the channel gets shorter. The solution was to have the plug grip the water channel from more sides. FinFET is a structure where the channel stands like a fin so that the gate wraps around three sides, and Gate-All-Around is a structure where the gate completely surrounds the channel on all four sides.

In FinFET, the channel does not lie flat but stands like a wall. The gate covers both sides of that wall, and commonly the top as well. With the gate on multiple sides, FinFET handled shorter channels better than flat transistors, making it possible to shrink the switch further (DOI: 10.1155/2014/365689).

This shape made its way into actual factories. A manufacturing process unveiled in 2012 reported that by building a chip for mobile devices using transistors wrapped on three sides along with a front-side film of another material, it tightly controlled short channels and achieved a record-low leaking current (DOI: 10.1109/iedm.2012.6478969). This was reported by the group that developed the manufacturing process.

The next step is for the plug to wrap all the way around the water channel. In Gate-All-Around, the channel is shaped into thin sheets stacked vertically, and the gate wraps around each individual sheet from all four sides. These thin sheets are called nanosheets. In the transition from a flat structure to FinFET and then to Gate-All-Around, the gate area touching the channel increased, allowing the gate to control the channel better and pass more current in the same footprint (DOI: 10.1093/nsr/nwae008).

Gate-All-Around with stacked nanosheets was proposed as a strong successor to FinFET at a 2017 conference presentation. It was reported to have a wider path for current in the same footprint and better performance than FinFET (DOI: 10.23919/vlsit.2017.7998183). A study comparing the two through computer simulation also showed that the nanosheet design passed more current when turned on within the same area (DOI: 10.1109/jeds.2018.2866026, simulation). To put it in terms of a faucet, it is shaped like stacking multiple water channels in one spot with a plug wrapping entirely around each layer.

As the sides where the gate wraps around the channel increase from one to three and then to all four, less current leaks when turned off, and more current flows in the same space.

When You Read the Headline "GAA Mass Production"

Now let us read a news headline: "Mass Production of Next-Generation GAA Transistors Begins." At first glance, it is a headline with only three letters. Based on just what we have seen so far, we can ask three questions.

First, what is GAA? It is an acronym for gate-all-around. It is a transistor in which the stopper called a gate surrounds the channel, which is the path for water, from all four sides. The immediately preceding generation, FinFET, surrounded it on three sides.

Second, why surround it? As the switch gets smaller and the channel becomes shorter, current leaks even when it is turned off, and the gate's grip on the channel weakens. Increasing the surrounded sides makes that grip stronger and reduces the leaking current. It is reduced, not made zero. Even a structure surrounding all four sides has separate leakage paths that must be blocked (DOI: 10.23919/vlsit.2017.7998183).

Third, what do you gain? That you can pass more current in the same space is the benefit shown by those who proposed this structure and by simulation research. How much of that benefit actually emerges in a real product is not told by the headline.

If a number like "3 nanometers" is attached to the same headline, what that number is the length of is yet another question. That question is answered separately by the article covering "What Is "3 Nanometers" the Length Of?".

The three letters "GAA" mean that the stopper surrounds the water path from all four sides, and the reason is the current leaking from the shrunken switch.

Where to Look Next

The exact number that will surround all sides has not yet been decided. A 2024 review paper reports that once we move past the generation called 1 nanometer as a process name, we can no longer expect progress through the method of reducing size, and points to the structure of stacking switches vertically as the next path (DOI: 10.1093/nsr/nwae008). Here, the nanometer number is not a length measured from any part inside the chip, but a name attached to a generation.

• "Why Did the Clock Stop at 4 GHz — Dark Silicon and the Power Wall" — What happened when the power gains obtained by making things smaller hit a wall.
• "After Moore's Law — What a "Nanometer" Came to Mean" — Current leakage and changes in gate structure, one step deeper.
• "2D Semiconductors — Graphene's Betrayal and the Real Candidate for the "Post-Silicon" Era" — Attempts to make channels out of materials just a few atoms thick instead of silicon.
• From the same primer series, What Is "3 Nanometers" the Length Of?, Printing with Light — Lithography, and Fetching Is Slower Than Computing — The Memory Bottleneck and HBM cover the number in the process name, the light that carves switches, and the work of bringing data from outside the chip, respectively.

Source Papers

• "The other transistor: early history of the metal-oxide semiconductor field-effect transistor" (1998) — DOI: 10.1049/esej:19980509 — The idea of controlling semiconductor conduction with an electric field first appeared in a 1928 patent, and the MOSFET became a basic component of digital electronics after surface issues were solved in 1958, driven by the needs of computers and integrated circuits.
• "A history of the invention of the transistor and where it will lead us" (1997) — DOI: 10.1109/4.643644 — Bardeen and Brattain discovered the transistor at Bell Labs in November 1947.
• "Fifty Years of Moore's Law" (2011) — DOI: 10.1109/tsm.2010.2096437 — The trendline of doubling components per chip starts from a single component in 1959.
• "High dielectric constant gate oxides for metal oxide Si transistors" (2005) — DOI: 10.1088/0034-4885/69/2/r02 — The gate insulating film SiO₂ thinned down to 1.4 nm, causing leakage to become too large, requiring a switch to a high-k oxide that can be used at a greater physical thickness.
• "Dielectric scaling of a top gate silicon nanowire on insulator transistor" (2008) — DOI: 10.1063/1.3043662 — In three-dimensional quantum simulations, the thinner the insulating film, the better the gate controls the channel potential.
• "Scaling Challenges for Advanced CMOS Devices" (2017) — DOI: 10.1142/s0129156417400018 — Every new process generation demands scaling of chip power, performance, and area.
• "Nanomaterials for Next-Generation Semiconductors: From Transistor Scaling to Heterogeneous Integration" (2025) — DOI: 10.54254/2755-2721/2026.ka30036 — In the gate-length regime of 10 nm, short-channel effects, power density, and interconnect delay eat away at the traditional gains in performance, power, and area.
• "Leakage current mechanisms and leakage reduction techniques in deep-submicrometer CMOS circuits" (2003) — DOI: 10.1109/jproc.2002.808156 — As threshold voltage, channel length, and oxide thickness shrink, leakage has become a large share of power consumption, with various leakage mechanisms such as drain-induced barrier lowering being especially important in low-power applications.
• "Device scaling limits of Si MOSFETs and their application dependencies" (2001) — DOI: 10.1109/5.915374 — Scaling limits stem from tunneling and subthreshold leakage, and the endpoint is not singular but varies by application.
• "Review of FINFET technology" (2009) — DOI: 10.1109/soi.2009.5318794 — As planar CMOS was miniaturized, maintaining gate-channel control became difficult, prompting the proposal of multi-gate devices like FinFETs.
• "FinFETs: From Devices to Architectures" (2014) — DOI: 10.1155/2014/365689 — FinFETs with gates on two or three sides handle short-channel effects better than planar devices, keeping miniaturization going.
• "A 22nm SoC platform technology featuring 3-D tri-gate and high-k/metal gate, optimized for ultra low power, high performance and high density SoC applications" (2012) — DOI: 10.1109/iedm.2012.6478969 — Announcement of a manufacturing process reporting exceptional short-channel control and record-low leakage through 3D tri-gates.
• "New structure transistors for advanced technology node CMOS ICs" (2024) — DOI: 10.1093/nsr/nwae008 — Moving from planar to FinFET to GAA increased the gate sides, improving control and current, and vertical stacking rather than geometric shrinkage is pointed to as the path forward after the 1 nm node.
• "Stacked nanosheet gate-all-around transistor to enable scaling beyond FinFET" (2017) — DOI: 10.23919/vlsit.2017.7998183 — Stacked nanosheet GAAs have a wide effective width per unit area and better performance than FinFETs, making them a candidate to replace FinFETs beyond 5 nm.
• "Systematic DC/AC Performance Benchmarking of Sub-7-nm Node FinFETs and Nanosheet FETs" (2018) — DOI: 10.1109/jeds.2018.2866026 — In TCAD simulations, nanosheets of the same area have a larger effective width, resulting in a higher on-current than FinFETs.