The distinctive electrical behavior of a semiconductor arises from its atomic crystal lattice and the quantum-mechanical arrangement of its electrons into energy bands.¹³¹⁴ In any solid crystalline material, electrons are restricted to specific ranges of energy.¹⁵ The highest energy range that is completely filled with electrons at absolute zero temperature is called the valence band, where electrons remain bound to individual atoms within covalent bonds. Above the valence band lies the conduction band, where electrons are detached from their parent atoms and can move freely through the crystal lattice to carry an electric current.

The energy difference separating the top of the valence band from the bottom of the conduction band is known as the band gap. In conductors like metals, the valence and conduction bands overlap, allowing electrons to flow effortlessly. In insulators, the band gap is wide—typically several electron volts (eV)—preventing thermal energy at room temperature from exciting electrons across the divide. Semiconductors possess a moderate band gap, such as 1.12 eV in silicon or 1.42 eV in gallium arsenide at room temperature. When thermal energy or incident photons with energy exceeding the band gap strike an intrinsic (pure) semiconductor, electrons jump from the valence band into the conduction band.¹⁶ Each excited electron leaves behind a vacant state in the valence band called an electron hole.¹⁷ Neighboring valence electrons can shift into this vacancy, causing the hole itself to migrate through the lattice as if it were a positively charged particle.

Semiconductor materials are broadly categorized into elemental semiconductors, composed of a single species of atom, and compound semiconductors, synthesized from two or more elements across different groups of the periodic table.¹⁹ Historically, researchers refer to elemental germanium and silicon as first-generation materials, direct-bandgap III-V compounds as second-generation materials, and wide-bandgap compounds as third-generation materials.

Germanium was the primary material used for the earliest diodes and transistors during the late 1940s and early 1950s because it has a lower melting point (937 °C) than silicon and was easier to refine to high purity.²⁰²¹ However, germanium devices exhibit high leakage currents in the off state and cease functioning reliably above 70 °C. Once chemical refinement techniques improved in the mid-1950s, silicon supplanted germanium and came to account for more than 95 percent of all semiconductor devices manufactured globally. Silicon operates across a much wider temperature range (-55 °C to 125 °C), is the second-most abundant element in Earth's crust in the form of silica, and naturally grows a chemically stable insulating layer of silicon dioxide ($SiO_2$) on its surface when heated in oxygen—an attribute crucial for planar circuit fabrication.²²²³

Compound semiconductors address the physical limitations of silicon. Gallium arsenide (GaAs) and indium phosphide (InP), formed from Group 13 and Group 15 elements, possess electron mobilities up to six times higher than that of silicon, allowing transistors to switch at microwave and millimeter-wave frequencies used in satellite communications, radar, and cellular base stations. Furthermore, unlike silicon's indirect band gap, GaAs and InP feature a direct band gap, meaning electrons can transition directly from the conduction band to the valence band and emit energy as photons without requiring lattice vibrations (phonons) to conserve momentum. This property makes direct-bandgap compounds essential for laser diodes, fiber-optic transceivers, and high-efficiency space photovoltaics.²⁴

A growing segment of the industry relies on wide-bandgap semiconductors, principally silicon carbide (SiC) and gallium nitride (GaN), whose band gaps exceed 3 eV. Silicon carbide has a band gap of approximately 3.2 eV, a thermal conductivity three times that of silicon, and a breakdown electric field strength ten times higher. These characteristics allow SiC metal-oxide-semiconductor field-effect transistors (MOSFETs) to handle voltages of 1,200 V to 1,700 V with minimal energy loss, making them the standard in electric-vehicle traction inverters, fast-charging stations, and photovoltaic grid converters.²⁵²⁶²⁷ Gallium nitride excels in high-frequency switching and mid-voltage power conversion (typically around 600 V) as well as solid-state LED lighting. Researchers are also developing ultra-wide-bandgap materials, including gallium oxide and synthetic diamond, which offers five times the thermal conductivity of SiC for extreme-temperature aerospace and defense electronics.

Scientific observation of semiconducting behavior predates atomic theory. Although Alessandro Volta used the term "semiconducting" as early as 1782, the first documented experimental finding occurred in 1833, when English physicist Michael Faraday reported that the electrical resistance of silver sulfide decreased when heated, contrary to the behavior of metallic conductors. In 1873, English engineer Willoughby Smith discovered photoconductivity after observing that light lowered the electrical resistance of selenium. Three years later, William Grylls Adams and Richard Evans Day found that illuminating a junction of selenium and platinum generated an electric current, demonstrating the photovoltaic effect and enabling American inventor Charles Fritts to construct the first rudimentary selenium solar cell in 1883. Meanwhile, in 1874, German physicist Karl Ferdinand Braun observed that pressing a thin metal wire against metallic sulfide crystals allowed current to pass in only one direction. This point-contact rectification became the basis of "cat's-whisker" crystal radio detectors in the early twentieth century.

The word "semiconductor" (German: Halbleiter) was introduced in its modern context by Josef Koenigsberger and Jakob Weiss in 1911. Theoretical understanding remained incomplete until the formulation of quantum mechanics in the late 1920s. In 1929, German physicist Walter Schottky confirmed the existence of potential barriers at metal-semiconductor junctions, and in 1931 British physicist Alan Herries Wilson published The Theory of Electronic Semiconductors, applying quantum band theory to explain conduction, insulation, and the role of impurities. In 1940, American physicist Russell Ohl at Bell Laboratories stumbled upon the p-n junction while examining a cracked sample of purified silicon that produced a voltage when exposed to a flashlight.

Modern semiconductor manufacturing transforms raw mineral quartz into wafers packed with billions of nanoscale devices through hundreds of sequential chemical, optical, and thermal steps conducted inside ultra-clean facilities known as foundries or "fabs".³²³³ The process begins by reducing quartzite sand ($SiO_2$) with carbon in a furnace at roughly 1,800 °C to produce metallurgical-grade silicon of about 98 percent purity.³⁴ This crude silicon is chemically converted into liquid trichlorosilane gas, distilled, and redeposited as electronic-grade polycrystalline silicon (polysilicon) reaching a purity of 99.9999999 percent—commonly called "nine nines"—or higher.³⁵

To form the continuous crystal lattice required for electronic devices, polysilicon is melted inside a high-purity quartz crucible at temperatures just above silicon's melting point of 1,414 °C via the Czochralski process.³⁶ A precisely oriented single-crystal seed mounted on a rod is dipped into the molten silicon and slowly withdrawn while rotating. The liquid solidifies around the seed into a cylindrical single-crystal ingot, or boule, which is then sliced by diamond wire saws into thin circular discs called wafers, lapped flat, chemically etched, and polished to a mirror finish.

Inside the fabrication plant, the polished wafer undergoes repeated cycles of deposition, photolithography, etching, and doping. First, an insulating oxide film or conductive layer is grown or deposited on the wafer surface. The wafer is then coated with a light-sensitive chemical called photoresist. During photolithography, a specialized scanner projects ultraviolet light through a patterned template, or photomask, onto the photoresist, transferring the geometric layout of a circuit layer.³⁷ Exposed regions of the resist are dissolved away, and reactive plasma gases or wet chemicals etch the underlying exposed material.³⁸ Ion implantation machines then fire accelerated beams of dopant ions into designated regions of the silicon, followed by thermal annealing to repair the crystal lattice.

For the most advanced logic and memory chips—such as 3-nanometer (nm) and 2 nm process nodes—chipmakers rely on extreme ultraviolet (EUV) lithography machines produced exclusively by the Dutch manufacturer ASML.³⁹⁴⁰ An EUV system fires a high-powered laser at falling droplets of molten tin 50,000 times per second inside a vacuum chamber, creating a plasma that emits light at a wavelength of 13.5 nanometers, which is focused onto the wafer by atomically smooth reflective mirrors.

Semiconductors are divided commercially into four major product categories: logic chips (including central processing units, graphics processing units, and application-specific processors), memory chips (such as volatile Dynamic Random-Access Memory and non-volatile NAND flash), analog integrated circuits (which translate continuous real-world signals like sound, temperature, and power into digital data), and optoelectronics, sensors, and discrete components.⁴¹⁴²

Driven by the rapid build-out of artificial intelligence (AI) data centers, cloud infrastructure, and vehicle electrification, global semiconductor industry revenues expanded by 25.6 percent in 2025 to reach $791.7 billion, up from $630.5 billion in 2024, according to the Semiconductor Industry Association (SIA) and World Semiconductor Trade Statistics (WSTS). Monthly global sales reached $82.5 billion in January 2026, putting the industry on pace to approach $1 trillion in annual revenue. Demand for accelerated AI computing has reshaped market shares: companies specializing in AI logic processors, such as Nvidia, and high-bandwidth memory (HBM) manufacturers, including SK Hynix, Samsung Electronics, and Micron Technology, recorded the steepest revenue gains in 2025 and 2026, even creating memory allocation shortages for traditional personal computer and smartphone makers. Meanwhile, the automotive semiconductor segment is projected to nearly double from $68 billion in 2024 to $132 billion by 2030 as automakers adopt electric drivetrains and advanced driver-assistance systems.

The industry operates through a specialized global division of labor. "Fabless" design companies—such as Nvidia, Apple, Qualcomm, and AMD—design integrated circuit architectures using electronic design automation software but own no manufacturing plants. They contract fabrication out to pure-play foundries, led by Taiwan Semiconductor Manufacturing Company (TSMC), which entered volume production of its 2 nm nanosheet (gate-all-around) transistor process in late 2025.⁴³ Integrated device manufacturers (IDMs) such as Intel, Samsung Electronics, and Texas Instruments both design and manufacture chips, while specialized vendors supply lithography equipment, silicon wafers, and outsourced assembly and testing.

In 2026, leading-edge manufacturers diverged in their adoption of next-generation High Numerical Aperture (High-NA) EUV lithography—machines costing roughly $370 million each that shift the optical numerical aperture from 0.33 to 0.55 to print features at an 8 nm resolution. Intel Foundry pioneered High-NA EUV in high-volume production on selected layers of its 18A process node for Core Ultra Series 3 ("Panther Lake") processors in 2026. Conversely, Samsung committed to deploying High-NA EUV for DRAM production by 2028, while TSMC opted to manufacture its 2 nm and subsequent generations with conventional EUV, deferring High-NA volume production until 2030 alongside a joint industry initiative to transition from 6-inch to 12-inch photomasks by 2033.⁴⁴

Because over 50 percent of global semiconductor revenue is concentrated in the Asia-Pacific region—with advanced logic fabrication heavily centered in Taiwan and memory fabrication in South Korea—governments in North America, Europe, and Asia treat semiconductor supply chains as matters of national security.⁴⁵ Following pandemic-era chip shortages and escalating technological competition, the United States and the Netherlands instituted export controls restricting the sale of advanced AI processors and EUV lithography systems to China. Simultaneously, governments launched major industrial subsidy programs, such as the U.S. CHIPS and Science Act and India's Semicon India program, to incentivize domestic wafer fabrication.

The physical supply chain also remains vulnerable to geographic bottlenecks and climate shocks. In September 2024, Hurricane Helene temporarily halted mining operations in Spruce Pine, North Carolina, a single locality that supplies roughly 70 percent of the world's ultra-high-purity quartz required to manufacture the crucibles used in Czochralski silicon ingot pulling. Furthermore, a 2024 World Economic Forum analysis noted that 40 percent of existing semiconductor fabs and 40 percent of facilities announced since 2021 are located in watersheds projected to face high or extremely high water stress between 2030 and 2040.

Chip fabrication is resource-intensive, requiring vast quantities of ultrapure water to rinse wafers between hundreds of chemical steps, as well as substantial electrical energy to run cleanroom climate controls and lithography tools. Manufacturing processes also rely on hundreds of per- and polyfluoroalkyl substances (PFAS)—synthetic compounds with strong carbon-fluorine bonds used in photoresists, anti-reflective coatings, and etching fluids because they withstand harsh chemical and thermal conditions.⁴⁶⁴⁷ Because PFAS persist in the environment and have been linked to adverse human health outcomes, wastewater discharges from expanding fabrication plants have drawn regulatory scrutiny in Europe and the United States. While semiconductor trade groups argue that immediate bans could disrupt global electronics production due to a lack of qualified drop-in chemical substitutes, environmental researchers estimate that global expenditures to treat and remove PFAS from semiconductor and electronics wastewater could reach $16.5 billion to $50 billion annually by 2035.⁴⁸