how much does it cost to build a chip fab

A leading edge fab costs many billions of dollars. Most of that goes on machines rather than on the building. A plant that makes older, simpler chips costs a small fraction of that, which is why more countries can build one. Both kinds also need years of work before they run well.

what is the difference between a fab and a foundry

A fab is a single factory that makes chips. A foundry is a company whose business is making chips to other people's designs, and it may own several fabs. Every foundry has fabs. Some fabs belong to companies that make only their own chips, so they are not foundries.

why do chip fabs use so much water

Wafers are rinsed after almost every step, and the water has to be far cleaner than drinking water. A large fab therefore uses millions of litres a day. Many plants now recycle a big share of it. Water supply is one of the first things a new site is judged on.

can my country build its own chip fab

A mature fab making simple chips is within reach for a country with steady power, clean water and a plan to train staff. A leading edge fab is much harder, because the machines, suppliers and skills are scarce and are controlled by trade rules. Many countries start with packaging and testing plants instead, which cost far less and employ a lot of people.

What is a chip fab, and why are there so few?

Picture a small shop that prints T-shirts. A worker lays a stencil on the cloth, pushes ink through the gaps, then lifts the stencil away. The pattern appears.

A chip fab plays the same trick. However, the pattern is far thinner than a human hair. The cloth is a polished disc of silicon. And the shop prints on the same disc over and over, layer on layer, for weeks.

Each layer has to land on the one below it with almost no error. That is the whole job. Everything else in a fab exists to protect that one act.

What a chip fab actually is

Fab is short for fabrication plant. It is the factory where raw silicon becomes a working chip.

A fab starts with a wafer. A wafer is a thin round slice cut from a single crystal of very pure silicon. Modern logic fabs use wafers 300 millimetres across, about as wide as a dinner plate. One wafer carries hundreds of chips. At the end they are cut apart.

Why start with a big disc at all? Because most steps cost the same whether the disc holds one chip or nine hundred.

Silicon is the classic semiconductor, a material that carries current only when you tell it to. A fab is where that switch gets built, billions of times over, on a surface you could cover with one hand.

How it works, step by step

The core loop is short. A fab repeats it again and again.

Grow a silicon crystal, slice it into wafers, and polish each wafer flat.
Coat the wafer with a light-sensitive liquid called photoresist.
Shine light through a mask. The mask holds the pattern for one layer.
Wash the wafer. The light changed the resist, so the pattern is now open in some places and covered in others.
Etch away material, add a new film, or fire in atoms that change how the silicon conducts.
Clean, polish flat again, and start the next layer.
Test every chip on the wafer, cut the wafer up, and package the ones that pass.

A modern chip needs many layers. The wiring alone can take more than ten. Count every coat, print, etch and clean, and the total runs into the hundreds of steps. Therefore one wafer can spend weeks inside the building. For the most advanced chips it can spend a few months.

Why a speck of dust ruins a chip

Hold a sheet of paper up to a window and you will see specks floating in the light. Ordinary indoor air is full of them. On a wafer, one of those specks is a boulder. What happens when a boulder lands on a printed line? The line breaks, and the chip under it is scrap.

Therefore a fab is built as a cleanroom. Filtered air pours down from the ceiling all day. Workers wear full suits so their skin and hair stay sealed in. The floor sits on a slab that soaks up vibration from the road outside. Heat and humidity are held steady. Water is cleaned far past drinking standard before it touches a wafer.

Even so, some chips fail. Engineers track the share of chips on a wafer that work. That share is called yield. YIELD IS WHAT A FAB REALLY SELLS. A plant with poor yield spends the same silicon, the same power and the same weeks of work as a plant with good yield. It simply has less to sell at the end.

Why one fab costs so much

Most of the money goes into machines, and the machine that prints the pattern is the dearest of all.

The finest printing today uses extreme ultraviolet light, known as EUV. Its wavelength is about 13.5 nanometres. Glass and air both soak up that light, so the machine cannot use ordinary lenses. It works in a vacuum and steers the light with mirrors. The light itself is made by hitting tiny drops of molten tin with a laser. Only one company in the world sells these machines, and each one ships in many crates.

The building is costly too. It needs a floor that does not shake, huge air handling, a water plant, gas lines, and power that never wavers. A new leading edge fab costs many billions of dollars and takes several years to finish. Then it has to learn to make chips well, which takes longer still.

Every shrink has made this harder. That is one reason Moore’s Law has slowed.

Why there are so few of them

Money is the first answer. It is not the only one. Ask what a new entrant would need.

It would need the machines, and the queue for the best ones is long. It would need thousands of trained people, and that skill is learned on the job over years. It would need suppliers close by for chemicals, gases, masks and spare parts. It would need customers willing to bet a product on an unproven line.

Then comes the hardest part. Yield improves with practice. The plant that runs the most wafers learns the fastest, so it wins more orders, so it learns faster still. This is why the leading edge keeps folding into a few sites in a few countries.

Politics narrows it further. As of 2026, the most advanced printing machines cannot be sold freely to every country. Governments treat them as strategic goods.

Older fabs matter more than people think

Not every fab chases the smallest features. Many run mature recipes that are decades old, and they are busy.

These plants make power chips, radio chips, image sensors and display drivers. They also make the small controllers that run a washing machine, a water pump or a prepaid electricity meter. A car holds hundreds of chips, and most of them are simple ones. The shortage that stopped car factories in the early 2020s was mostly a shortage of these older chips.

This matters for any country weighing up a plant of its own. A mature fab costs a fraction of a leading edge one. It is a realistic goal. A plant making the very newest logic is a different order of problem.

Who designs the chip and who makes it

Three words explain most of the industry. A fabless company designs chips and owns no fab. A foundry owns fabs and makes chips to other people’s designs. An integrated maker does both under one roof.

So the chip in your phone usually has two parents. One company drew it. Another printed it. The NPU that reads your face at the lock screen is a good example of that split.

There is a third stage as well. After the wafer is cut up, the pieces have to be tested, wired and sealed into a package. That work often happens in a separate plant in a separate country. It is also where a lot of new progress sits. Makers now cut a design into smaller pieces called chiplets and join them in one package, sometimes stacking them upward. Packaging plants cost far less than fabs, and more countries can build them.

What to check

The next time you read about a chip shortage, ask one question. Which kind of chip?

If the answer is a controller or a power chip, the fix sits in mature plants, and many countries can build those. If the answer is the most advanced logic, the fix sits in a handful of buildings on Earth, and it will take years. The word “chip” hides that difference, and the difference is the whole story.

Just Out Tech explains new research in plain language. This article was drafted with AI assistance and checked by a human against the original source.