What Is a Transistor? How Modern Transistors Work

A transistor is a semiconductor device that controls current or voltage. It can act as an electronic switch, an amplifier, a current source, or part of an oscillator. Digital chips use vast networks of transistors to represent and transform information.

The useful mental model is control, not just on and off. In a MOSFET, gate voltage controls a channel between source and drain. In a bipolar junction transistor, a smaller base-emitter input controls a larger collector current.

What Is a Transistor?

A transistor uses semiconductor regions, insulating layers, metal contacts, and electric fields to regulate charge-carrier motion. The two main families are bipolar junction transistors, or BJTs, and field-effect transistors, especially MOSFETs.

Family	Control input	Main terminals	Common strengths
BJT	Base current and base-emitter voltage	Emitter, base, collector	Analog gain, RF, current drive
MOSFET	Gate-source voltage	Source, gate, drain, body	Low static gate current, dense CMOS logic, power switching
IGBT	Insulated gate controlling bipolar conduction	Emitter, gate, collector	High-voltage power conversion
Specialized FETs	Electric field	Varies	RF, sensors, high-power GaN/SiC, memory

Transistor choice depends on voltage, current, switching speed, gain, noise, temperature, fabrication process, and cost. “MOSFETs replaced BJTs” is too broad. CMOS MOSFETs dominate digital logic, while BJTs and hybrid devices still matter in analog, radio-frequency, and power circuits.

A Short History

John Bardeen and Walter Brattain demonstrated the point-contact transistor at Bell Labs in December 1947, with William Shockley leading the solid-state group and soon developing the junction transistor. The three shared the 1956 Nobel Prize in Physics.

Mohamed Atalla and Dawon Kahng demonstrated the MOSFET at Bell Labs in 1959. Its insulated gate, scalable geometry, and compatibility with complementary circuits made modern integrated electronics possible. The Intel 4004 arrived in 1971 with roughly 2,300 transistors; leading processors now integrate tens or hundreds of billions.

How a MOSFET Works

In an enhancement-mode n-channel MOSFET, a positive gate voltage creates an electric field through a thin dielectric. Above the threshold region, electrons form a conductive inversion channel between source and drain. A drain-source voltage then drives current through that channel.

$$I_D\approx\frac12\mu_n C_{\mathrm{ox}}\frac{W}{L}(V_{GS}-V_T)^2$$

This square-law relation is a useful long-channel saturation model, not an exact equation for modern nanoscale devices. Mobility degradation, velocity saturation, short-channel electrostatics, series resistance, quantum confinement, and leakage all change real current.

• Source: supplies carriers to the channel.
• Drain: collects carriers from the channel.
• Gate: controls channel charge through an electric field.
• Gate dielectric: insulates the gate so steady gate current is very small.
• Body: the semiconductor region that forms junctions and affects threshold voltage.

How CMOS Makes Logic

Complementary metal-oxide-semiconductor, or CMOS, pairs n-channel and p-channel MOSFET networks. In a simple inverter, one device pulls the output high while the other pulls it low. Ideally, little direct current flows in a stable state.

$$P_{\mathrm{dynamic}}\approx\alpha C V^2 f$$

Dynamic power grows with switched capacitance \(C\), activity factor \(\alpha\), frequency \(f\), and especially the square of supply voltage \(V\). Leakage and short-circuit currents add static power. This is why modern performance is constrained by heat and energy, not only by how many devices can fit.

How a BJT Works

A BJT has emitter, base, and collector regions arranged as NPN or PNP. Forward bias at the base-emitter junction injects carriers through a thin base. Most are swept into the collector, so a smaller base current controls a larger collector current.

$$I_C\approx\beta I_B$$

The current gain \(\beta\) varies with device, current, voltage, and temperature, so precision designs do not treat it as a fixed constant. In active operation, an exponential relation between collector current and base-emitter voltage is often more fundamental.

$$I_C\approx I_S e^{V_{BE}/V_T},\qquad V_T=\frac{k_{\mathrm B}T}{q}$$

That thermal voltage is about 25.9 mV at 300 K. The Boltzmann constant guide explains the \(k_{\mathrm B}T\) energy scale behind this temperature dependence.

Transistors as Switches and Amplifiers

Use	Operating idea	What the circuit designer controls
Digital switch	Choose voltage regions representing logic states	Noise margins, delay, leakage, energy
Analog amplifier	Bias in a region where small input changes produce useful output changes	Gain, bandwidth, noise, linearity
Power switch	Alternate low-loss off and on states	Breakdown voltage, on-resistance, switching loss, heat
Oscillator	Combine gain with frequency-selective positive feedback	Frequency stability, phase noise, startup

A physical transistor is never a perfect switch. The off state leaks, the on state has resistance, gates and junctions store charge, and transitions take time. Circuit design manages those non-ideal behaviours.

From Planar MOSFET to FinFET and GAAFET

Shrinking planar MOSFETs made it harder for the gate to control the channel against the drain’s electric field. FinFETs raised the channel into a fin so the gate could control several sides. Gate-all-around FETs wrap the gate around nanosheets or nanowires for still stronger electrostatic control.

Geometry	Gate control	Why it appeared
Planar MOSFET	One main channel surface	Simple, scalable foundation of CMOS
FinFET	Gate around three sides of a fin	Better short-channel control
GAAFET/nanosheet	Gate surrounds the channel	Stronger control and tunable effective width
CFET research	Stacked n- and p-type devices	Potential area reduction beyond lateral CMOS layouts

CFET is still a research and development direction, not a guaranteed consumer-node deadline. Imec’s CFET research announcement presents stacked complementary devices as a candidate beyond gate-all-around nanosheets.

What Process-Node Names Mean

A label such as 3 nm or 2 nm is a process generation name, not the literal length of every transistor feature. Foundries choose different naming conventions, cell architectures, pitches, materials, and density targets. Comparing node labels alone can be misleading.

A meaningful comparison asks about transistor density, performance at a specified power, leakage, standard-cell height, interconnect, memory density, yield, and packaging. The device geometry is only one layer of the product.

How Many Transistors Fit in Modern Chips?

Apple announced the base M4 in May 2024 with 28 billion transistors on a second-generation 3-nanometre process. NVIDIA announced the Blackwell GPU with 208 billion transistors across two reticle-limited dies connected by a high-speed link.

Those official figures come from Apple’s M4 announcement and NVIDIA’s Blackwell announcement. The count is not a direct speed score. Architecture, memory bandwidth, clocks, power, software, and workload decide useful performance.

Why Scaling Is Harder Now

• Electrostatics: a shorter channel is harder for the gate to control.
• Leakage: thin barriers and small geometries permit tunnelling and subthreshold current.
• Interconnect: wire resistance and capacitance can dominate delay and power.
• Variability: small material and dimension differences matter more at tiny scales.
• Heat: more switching activity in less area raises power density.
• Economics: advanced lithography, masks, design, and fabrication plants are extremely expensive.

The industry’s answer is not one miracle transistor. It is co-optimization: new device structures, backside power delivery, better interconnect, specialized accelerators, chiplets, advanced packaging, 3D stacking, and software that uses hardware efficiently.

Common Misconceptions

• A transistor stores a bit forever: logic transistors switch; memory cells use transistor-capacitor or other device arrangements with specific retention behaviour.
• A smaller node is a literal gate length: modern node labels are product-family names.
• More transistors guarantee a faster chip: many may be memory, cache, I/O, or specialized units.
• MOSFET gates draw no current: ideal static gate current is zero, but real devices have leakage and charge/discharge current.
• CFET is already the universal next consumer device: it is a promising research path with major integration challenges.
• Silicon is disappearing immediately: silicon CMOS remains central while GaN, SiC, and other materials grow in specific roles.

Related Physics Guides

Use these next if you want to connect this result with the surrounding physics:

• Boltzmann constant and thermal energy
• Coulomb’s law and electric force
• Physics basics study notes

Key Takeaways

• A transistor controls current or voltage and can switch, amplify, oscillate, or regulate power.
• MOSFETs use gate electric fields; BJTs use carrier injection through a base region.
• CMOS logic limits stable-state current but still consumes dynamic and leakage power.
• FinFET and GAAFET geometries improve gate control as devices shrink.
• Modern progress combines transistor design with interconnect, architecture, packaging, and software.

Frequently Asked Questions

The transistor count is easy to market. The hard engineering is delivering useful work within power, heat, yield, and cost limits.

Browse more in Physics.