What Is a Silicon Wafer? How Wafers Are Made, Step by Step
A silicon wafer is a round plate of silicon, less than 1 mm thick, made of a single crystal. That means its atoms line up the same way across the whole plate. The surface is polished to a mirror finish. Hundreds to thousands of chips are built on one wafer at the same time. Wafers are round because the crystal is grown as a cylinder, rotated while it is pulled out of molten silicon, and then sliced.
What you'll learn
- What a wafer is, and why chips are built on a round plate
- How rock becomes extremely pure silicon
- How a single crystal is grown and polished to a mirror finish
- Why flatness and freedom from particles matter
- Why wafers have grown larger, and who makes them
Prerequisites: none. It helps to know that chips are made together on one plate. "How Are Chips Made?" covers the whole process.
In everyday life
News about chip factories often gives output as a number of wafers, such as "tens of thousands of wafers a month." A wafer is the thin, round plate that chips are built on. When it arrives at the factory, it has no circuits yet. It is just a shiny disc that looks like a mirror.
A small number of specialist companies make most of the world's wafers. Two of the five main makers are Japanese (they are listed near the end). If you join a chip company, you will often hear wafer words such as "12-inch," "notch" and "epi." This article explains them as they come up.
What a wafer is: a round plate made of one crystal
The largest wafers used in volume production are 300 mm across, about the size of a large dinner plate. They are about 0.775 mm thick, close to a credit card (about 0.76 mm). By calculation, one wafer weighs about 130 g.
The key point is that the whole plate is one crystal. A crystal is a solid whose atoms are arranged in a fixed, repeating pattern. In a wafer, that pattern points the same way from one edge to the other. This kind of crystal is called a single crystal.
Imagine a plate made of many small grains, each pointing a different way. The way electricity flows would change at the boundaries between grains. A chip holds billions of transistors (electrical switches), and they all need to behave the same. That is why the wafer must be a single crystal, with the same arrangement everywhere.
Circuits are built only in a very shallow layer at the surface. The rest of the thickness is support, so the wafer does not crack or warp during processing. After the circuits are made, the back of the wafer is often ground thinner. Then the wafer is cut into chips.
Why chips are made together on a round plate
Chips are tiny. Making them one at a time would take too much work, and their quality would vary. Instead, identical chips are laid out in a grid on one plate and processed together. A 300 mm wafer holds about 600 chips that are 10 mm square.
Wafers are round because the crystal is rotated as it is pulled up during growth. It grows into a cylinder, and slicing a cylinder gives round plates. Square chips on a round plate leave partial chips at the edge. Even so, this growth method is the best way to make large, high-quality single crystals.
From rock to ultra-pure silicon
Silicon is everywhere around us. Sand and rock contain it as silicon dioxide, which is silicon bonded to oxygen. The raw material for wafers is quartzite, a quartz rock that is almost pure silicon dioxide. Beach sand is not used as it is.
First, quartzite and carbon are heated to a high temperature in an electric furnace. The carbon takes away the oxygen, and silicon is left behind. This is called metallurgical-grade silicon. It is about 98–99% pure, which is not nearly enough for chips.
Next, this silicon is reacted with hydrogen chloride. That turns it into a compound that is liquid at room temperature. As a liquid, it can be distilled again and again to remove impurities. The cleaned compound then goes into a reactor with hydrogen. There, silicon slowly builds up on the surface of thin, heated silicon rods. This method is called the Siemens process.
The silicon that builds up is made of many tiny crystal grains. Each grain points a different way, so it is called polycrystalline silicon, or polysilicon. It is extremely pure, but it cannot be used for chips in this form.
Purity is written with a row of nines, like 99.9999999%. Many sources put the semiconductor grade at 9 to 11 nines. A major Japanese materials maker treats 11 nines ("eleven nines") as the expected level (news report, 2026). That means no more than one impurity atom per 100 billion atoms.
Why so pure? So that tiny, deliberate amounts of impurities can set how easily the silicon conducts electricity. "What Is a Semiconductor?" explains this.
Growing a single crystal: the CZ method
Polysilicon is placed in a heat-resistant quartz container called a crucible. It is melted at just over 1,400 °C. A tiny amount of boron or phosphorus is added at this stage. This gives the crystal the intended electrical conductivity.
A small piece of single crystal is then dipped into the surface of the molten silicon. This piece is called the seed crystal. The seed is rotated and slowly pulled upward.
Where the molten silicon touches the seed, it cools slightly and freezes. As it freezes, its atoms line up the same way as the seed's atoms. This repeats over and over, so a thick, long single crystal grows downward with the seed's arrangement. The pulling speed and temperature are adjusted to keep the diameter constant.
This growth method is called the Czochralski (CZ) method. It is based on a method discovered in 1916 by Czochralski, a Polish-born chemist. It was first used to study metal crystals and was later applied to silicon.
The grown cylinder is called an ingot. Both ends, where the diameter is uneven, are cut off. The outside is then ground so the diameter is the same all along.
In the CZ method, the quartz crucible slowly dissolves into the molten silicon. As a result, the crystal picks up a little oxygen. The float-zone (FZ) method uses no crucible. Because the silicon never touches a container, the crystal takes in less oxygen and other contaminants. However, it is hard to make large-diameter crystals this way. FZ is used for some applications, such as power semiconductors that handle high voltages. Most production wafers are made by the CZ method. "CZ vs FZ silicon" compares the two.
Slicing, grinding and polishing to a mirror finish
Turning an ingot into wafers takes many steps. Here they are in order.
- Mark the orientation: A small V-shaped cut is made in the edge. It shows which way the crystal faces and is called a notch. Factory tools use it to turn every wafer the same way.
- Slice thin: A saw with hundreds of thin, moving parallel wires cuts the ingot into many round plates at once. Each plate is less than 1 mm thick.
- Round the edges: Sharp edges can chip when they hit something, which leads to cracks and particles. So the edges are ground round.
- Even out thickness and flatness: Both faces are ground so every plate has the same thickness and is flat.
- Remove the damaged layer: Cutting and grinding leave fine scratches and strain near the surface. This layer is dissolved away with chemicals.
- Polish to a mirror finish: The surface is polished with a soft pad and a liquid containing very fine particles. The liquid reacts with the silicon surface to make it easier to remove. The particles then wear it away little by little.
- Clean and inspect: Small particles and dirt are washed off, and flatness and thickness are measured.
Inspection shines laser light on the surface. Any tiny particle scatters the light, which shows how many particles there are and where. Wafers that pass are shipped in special containers that keep particles out.
For some uses, another thin single-crystal layer is grown on top of the polished wafer. The new layer copies the crystal arrangement below it, so this method is called epitaxial growth. A wafer made this way is an epitaxial wafer, which people in the industry simply call "epi." The top layer can have fewer defects, or a different conductivity.
There are also wafers with an insulating film under a thin top layer of silicon. "SOI vs bulk silicon" compares them with ordinary wafers.
Why wafers must be flat and particle-free
In a chip factory, circuit patterns are printed onto the wafer with light. This step is called exposure, part of lithography. The machine uses lenses or mirrors to project a shrunken image of the pattern onto the wafer. The range of heights where the pattern is in sharp focus is very narrow.
This range is called the depth of focus. Printing finer patterns needs lenses that gather light from wider angles, and that makes the depth of focus shallower. Wherever the wafer surface has a tiny bump or dip, the pattern falls out of focus and blurs.
The machine measures the height at each spot and refocuses. Even so, it cannot correct small bumps within the area printed at once (a few centimeters across). So the wafer itself must be extremely flat. "What Is EUV Lithography?" explains how fine patterns are printed.
Particles are another enemy. Where a particle sits, the pattern may not print correctly, and wires may short or break. That chip becomes defective, and the share of good chips (yield) drops. That is why wafers are checked strictly before shipping, down to the number of particles.
Why wafers have grown larger
Early wafers were only a few centimeters across. Since then they have grown in stages, mainly 100 mm, 150 mm, 200 mm and 300 mm. Volume production on 300 mm wafers began around 2001. In factories, 300 mm wafers are also called "12-inch" wafers (strictly, 12 inches is about 305 mm).
Going bigger gets more chips from each wafer. Area grows with the square of the diameter, so a 300 mm wafer has 2.25 times the area of a 200 mm wafer. Many steps process wafers one at a time or in batches. So the bigger each wafer, the more chips each step produces.
The number of chips grows even faster than the area. Let's count 10 mm square chips with this site's calculator. A 200 mm wafer gives about 260, and a 300 mm wafer about 610, about 2.4 times as many. The reason is that a bigger wafer wastes a smaller share at the edge. The partial chips lie in a thin band along the edge. The length of that band (the circumference) grows only in proportion to the diameter. Try different wafer diameters and chip sizes in the die per wafer and yield calculator.
On the other hand, changing the diameter means rebuilding almost all of a factory's equipment. For the next size, 450 mm, a joint development group (G450C) was set up in the US in 2011. Intel, TSMC, Samsung Electronics and others took part. But the member companies decided it was not yet time to move to the next phase. The group ended its work at the end of 2016 (EE Times, January 2017). As of 2026, the largest wafers in volume production are still 300 mm.
Smaller wafers have not disappeared either. Power semiconductors, sensors and analog parts are still made in 200 mm fabs, according to the industry group SEMI. Parts that do not need the finest features suit familiar, proven equipment better.
Who makes wafers?
Chipmakers usually do not make their own wafers. They buy them from specialist wafer makers. A wafer maker handles everything from growing the crystal to polishing, inspection and shipping. The circuits are built by the chipmaker that buys the wafers, or by a contract manufacturer (a foundry).
Five companies are often named as the main silicon wafer makers:
- Shin-Etsu Chemical (Japan; its group company Shin-Etsu Handotai makes the wafers)
- SUMCO (Japan)
- GlobalWafers (Taiwan)
- Siltronic (Germany)
- SK siltron (South Korea)
Market share figures vary with the research firm and the year, so this article does not list them. The companies that make the raw polysilicon, and those that build crystal-growing equipment, are separate businesses again.
Common misconceptions
Myth: Wafers are made by melting beach sand
In fact, the raw material is quartzite, which is almost pure silicon dioxide. It has the same main ingredient as sand, but it is not simply melted. It is first turned into metallurgical-grade silicon, then purified as a liquid compound, and finally grown into a single crystal. "Chips are made from sand" is a short way of saying the ingredient is the same.
Myth: Circuits fill the whole thickness of the wafer
In fact, circuits are built only in a very shallow layer at the surface. The rest of the thickness is support, so the wafer does not crack or warp during processing. When the chips are finished, the back is often ground thinner. Cross-section drawings usually enlarge the circuit layer, which is why this belief is common.
Myth: Bigger wafers make better-performing chips
In fact, the wafer diameter mainly changes how many chips you get per wafer, and the cost. A chip's speed and power efficiency depend on how fine its circuits are and on its design. Power semiconductors and sensors made on 200 mm wafers are important parts in their own fields. Because leading-edge chips are made on 300 mm wafers, people tend to link size with performance.
By the numbers
| Number | What it is |
|---|---|
| 300 mm | Diameter of the largest wafers in volume production (as of 2026) |
| About 0.775 mm | Standard thickness of a 300 mm wafer (SEMI standard) |
| About 130 g | Weight of one 300 mm wafer (calculated from its thickness and the density of silicon) |
| 2.25× | Area of a 300 mm wafer compared with a 200 mm wafer |
| About 260 → about 610 | Number of 10 mm square chips per wafer, 200 mm → 300 mm (rough estimate from this site's calculator) |
| 98–99% | Purity of metallurgical-grade silicon made from quartzite |
| 9 to 11 nines | Purity level of polysilicon for semiconductors (major makers treat 11 nines as standard; news report, 2026) |
| Just over 1,400 °C | Temperature at which silicon melts (melting point about 1,410 °C) |
| 1916 | Year the method behind the CZ process was discovered |
| About 13 billion square inches | Area of silicon wafers shipped worldwide in 2025 (SEMI, February 2026). That is about 8.4 km², or about 120 million 300 mm wafers |
Put the numbers side by side, and you can see why wafers are hard to make. A single crystal 30 cm across is grown from material with many nines of purity. It is sliced about as thin as a card and polished so flat that the printed image stays in focus. In 2025, more than 100 million of these plates, counted as 300 mm wafers, went to factories around the world.
Learn more
- Read about the next step: "What Is EUV Lithography?" covers how circuit patterns are printed, and "How Are Chips Made?" covers the whole process.
- See the differences in a table: CZ vs FZ silicon, SOI vs bulk silicon, Epitaxy vs CVD
- Try the numbers: the die per wafer and yield calculator works out how many chips you get from a wafer diameter (150–450 mm) and a chip size.
- Look up terms: crystal orientation (how the direction of a crystal is described), gettering (techniques that trap metal contamination), MCZ method (the CZ method with a magnetic field applied to the crucible)
- See it in a figure: a CZ diagram with the equipment parts labeled, and more, are in all figures.