What do 12-inch and 8-inch wafers mean, and why did wafers get bigger?

The number in "12-inch wafer" or "8-inch fab" is the diameter of the wafer: a 12-inch wafer is 300 mm (30 cm) across, and an 8-inch wafer is 200 mm (20 cm). A wafer is the thin, round slice of silicon that chips are made on: the chips are laid out across it, made all together, and cut apart at the end (What is a chip?). So when the news says "a 12-inch fab", it means a factory that makes chips on 12-inch wafers. Many people just say "300 mm wafer" instead; it's the same thing.

Are those real inches? Why is 300 mm called 12 inches?

They are inches, but only roughly. An inch is 25.4 mm, so a true 12 inches is 304.8 mm; 300 mm is really 11.8 inches. The same goes for 8 inches: 200 mm is about 7.9 inches.

Wafers are specified in millimeters. The industry's public standard gives the diameters as 300 mm and 200 mm and sets the thickness too, and wafer makers and tool makers both build to it. The very earliest wafers really were sized in inches, such as 2 and 3 inches. From about the 100 mm size on, sizes were set in millimeters, but people kept calling each one by the nearest whole number of inches. So "12-inch" is a name, not a measurement that comes out at exactly 12 inches.

How did wafer sizes grow?

Wafers didn't start this big. Roughly:

• 4 inches (100 mm) and smaller: earlier decades, growing one size at a time.
• 6 inches (150 mm): the 1980s.
• 8 inches (200 mm): in volume production from the early 1990s.
• 12 inches (300 mm): from around 2000; most new leading-edge lines use it today.

From 6 inches to 12, that's about one new size every ten years. Each time, every tool in the fab, the boxes that hold the wafers and the way they get carried around had to change too. At 12 inches a full pod weighs 7 to 9 kg, and carrying pods went over to OHT vehicles on ceiling rails (How do wafers move around a fab?).

How much bigger is a 12-inch wafer, and how many more chips fit?

The diameter is 1.5 times as large (300 ÷ 200). But chips cover the whole surface, so the 1.5 counts twice: 1.5 × 1.5 = 2.25 times the area. One side of a 12-inch wafer is about 707 cm², a bit more than a sheet of A4 paper; an 8-inch wafer is about 314 cm², roughly half a sheet.

Even more than 2.25 times as many chips fit. Square chips laid on a round wafer leave incomplete spots at the edge that go to waste, and the bigger the wafer, the smaller that share (Why are wafers round? covers this). Here is a rough formula common in university courses, worked out for a chip 1 cm on a side, about the size of a fingernail:

Wafer	Diameter	Area	Fits about
8-inch	200 mm	about 314 cm²	about 270 chips
12-inch	300 mm	about 707 cm²	about 640 chips

640 ÷ 270 is about 2.4 times. That's how many fit, not how many are good; the share that comes out good is the yield (What does "yield" mean in chip news?).

Because the gap is this big, capacity is often counted in "200 mm (8-inch) equivalent", where one 12-inch wafer counts as a little more than two (What is a wafer start?).

Why did wafers get bigger?

Think of baking cookies. The oven bakes one tray at a time. Make the tray bigger and you bake several more cookies at once, but the baking time and the person minding the oven don't grow nearly as much.

A fab works the same way. Many steps handle one wafer, or one batch, at a time, so a bigger wafer means more chips every time a step runs. But the tools, the building, the people, the power and the water don't become 2.25 times as much. Spread over each chip, the cost can come down.

"Can", because bigger also costs more: 12-inch tools, wafers and fabs are all more expensive, and it only pays when you make enough to keep the tools busy. That's why chips made in huge numbers, such as the processors and memory in phones and computers, are mostly made in 12-inch fabs.

So why didn't wafers keep getting bigger?

The next size was developed. In the 2010s a joint research program set out to scale wafers up to 450 mm, called 18-inch, and test wafers and some tools were built. But within a few years the work had been shelved, and volume production still runs on 12-inch wafers. The reasons usually given:

1. Every kind of tool has to be redone: lithography, etch, thin film, cleaning, metrology, transport. Each has to be redesigned and requalified, and the development bill is enormous.
2. Everyone has to switch together: wafer makers, tool makers and fabs only gain if they all change. Today's 12-inch fabs are hard to simply refit, too.
3. The gain is smaller than it looks: some steps work one small patch at a time anyway. A scanner prints one small field at a time (What is photolithography?), so a bigger wafer just means more fields to print, and that step saves almost nothing.
4. Bigger is harder to make: a larger wafer is heavier and has to be thicker to support itself, and the silicon crystal it's sliced from has to be grown wider, which is harder too.

With that much money at stake and an uncertain payoff, the industry stayed at 12 inches.

What do 8-inch fabs make today?

Many 8-inch fabs are still running. They make chips that don't need the smallest switches:

• Power management: turning the power from a battery or a wall socket into the voltages each part needs.
• Power devices: switches for large currents and high voltages, in cars and motors.
• Analog chips: handling signals that vary smoothly, like sound, temperature and brightness.
• Sensors and MEMS: MEMS are tiny moving structures built in silicon, such as the sensor that can tell when you turn your phone.

Older generations of process do these chips well. Many lines have run their tools for years, the tools are long paid off, and production is steady and cheap (The news says "mature nodes"). Switching the whole set to 12-inch doesn't always pay.

Most wafers are silicon (why is covered in What is a semiconductor?), but not all. Silicon carbide, a material for power chips, is much harder to grow as a crystal than silicon; its common wafers are still 6-inch, with 8-inch on the way (The news says "wide-bandgap semiconductors").

How thick is a wafer?

Less than a millimeter. Under the industry standard, a 12-inch wafer is about 0.775 mm (775 micrometers) thick and an 8-inch wafer about 0.725 mm, close to a credit card (0.76 mm).

Bigger wafers are thicker because they have to support themselves: when they're carried, heated or spun fast, a wafer that's too thin bends or cracks easily. Work out the weight from the density of silicon and a 12-inch wafer comes to about 130 g, an 8-inch one about 50 g.

After the chips are made, and before they're packaged at a test-and-assembly house, the wafer is ground thinner from the back (What is chip packaging?). So the chip inside your phone is much thinner than the wafer it started on.

In fab24hr

The Wafer glossary card says this fab uses 30 cm (12-inch) wafers, 25 to a box. The card for Hot lot H-2417 says the same: 25 wafers go into one tightly closed box called a FOUP. In the production fab, tap the etch tool ETCH-03 and then "Take it apart": the wafer in the chamber is labeled 300 mm across. At the test-and-assembly house, tap the dicing saw, and its card says the wafer is first ground thin from the back, then cut apart along the scribe lines (dicing).

There's no 8-inch fab in the game, and the game doesn't say how many nanometers its fab runs at. The companies, people, lot numbers and tool numbers are fictional; wafer sizes, thickness and why wafers got bigger follow public, general sources and industry standards. For other jargon, see Chip fab jargon in plain English.