What does each part of the semiconductor supply chain do?

The semiconductor supply chain (also called the semiconductor value chain) is the long line of companies a chip passes through on its way from nothing to a finished product: materials and tools, design software, chip design, photomasks, the wafer fab, assembly and test, and finally the companies that put chips into products. When the news talks about "upstream" and "downstream", it's cutting this line into stages; where the cuts go differs from place to place, and this article covers the two common ways.

"Semiconductors 101" sums up these stages in one table. This article walks the chain stop by stop, and at each stop answers three questions: what happens there, why it's often a different company, and what happens when it's missing.

What companies does a chip pass through?

Stop 1: materials and tools

• What happens: a wafer fab needs blank wafers (made by companies that grow silicon into crystal ingots, slice them and polish them), many gases and chemicals, light-sensitive photoresist to coat the wafers, and slurry for polishing them flat. A test-and-assembly house needs substrates to hold the chips, fine metal wire to connect them, and plastic to seal them. The tools come from equipment makers.
• Why another company: making a gas extremely pure, or building a scanner, is a specialty in its own right. A supplier that sells to many fabs can spread that know-how and equipment over all of them.
• When it's missing: run out of one commonly used gas or chemical, and every step that needs it stops, along with all the wafers queued behind it.

Stop 2: design software and IP

• What happens: chips are designed with specialized software called EDA (What is EDA?). Many chips also contain circuit blocks that someone else designed and that are bought and dropped in as they are, called semiconductor IP, or silicon IP (What is semiconductor IP?). Think of building a house: you draw the plans with software you bought, and buy the doors and windows ready-made.
• Why another company: the same software, or the same block, can be sold to many design companies, so each one doesn't have to write it from scratch.
• When it's missing: without design software, there's no way to draw all those switches in a chip, or to check them.

Stop 3: chip design

• What happens: decide what the chip will do, design the circuit on a computer, check it many times, then hand over the final design. That handover is called tape-out (What is tape-out?, How are chips designed?).
• Why another company: design needs people and software, not a factory, so even small companies can do it. Many design companies have no fab of their own; they're called fabless.
• When it's missing: if the checks miss a mistake and it turns up after tape-out, the design has to be fixed and the masks remade. That's expensive, and it adds months.

Stop 4: photomasks

• What happens: the design is turned into a set of photomasks, the "negatives" used in exposure. Each layer of the chip's pattern needs at least one, and a set often has dozens (What is a photomask?).
• Why another company: a mask is written stroke by stroke with an electron beam or laser, then inspected and repaired, which takes its own tools and skills. So masks are made in specialized mask shops. Some belong to a wafer fab; others make masks for anyone.
• When it's missing: until the masks arrive, the fab can't start. One extra defect on a mask shows up in the same spot of every field.

Stop 5: wafer fabrication

• What happens: following the masks, the wafer fab takes wafers through hundreds of steps, building the circuit layer by layer. It takes weeks to months (How are computer chips made?).
• Why another company: this is the most expensive stage of the whole chain; building an advanced fab takes years. One company's chips often can't fill a fab, so there are foundries that make chips for many companies, and there are IDMs that design and make their own (The news says "fabless", "foundry" and "IDM").
• When it's missing: when a fab stops, every customer's chips inside it run late. Making more can't be rushed either: it starts over from the wafer start (How long does a wafer take to make?).

Stop 6: assembly and test

• What happens: finished wafers first go through wafer sort, where every chip is tested. Then they're cut apart, each chip goes into a package, and it's tested again once it's sealed. Some chips are packed together with small parts like resistors and capacitors into a tiny module (SiP; see What is SiP?). This stage is called assembly & test, and it's often done at a test-and-assembly house (What is an OSAT?).
• Why another company: the tools and materials are very different from a wafer fab's, and there are many kinds of package. A test-and-assembly house that takes work from many customers keeps its tools busy.
• When it's missing: with no tool to package them and no substrates to use, finished wafers just wait. Skip the tests, and bad chips end up with customers.

Stop 7: into the product and into your hands

• What happens: the packaged chips or modules are soldered onto circuit boards, which go into phones, computers and cars. The assembly is often handed to specialized electronics contract manufacturers.
• Why another company: the chips in one product come from many companies, and assembly takes a different kind of factory.
• When it's missing: a car has many chips in it. Miss one cheap chip, and the car can't be finished.

Put together: each stage needs very different amounts of money (design spends mostly on people and software, the wafer fab costs the most), and the know-how is completely different. A company that does one stage and serves many customers can spread the cost. How these companies hook up with each other is in "How does a chip design company work with a foundry?".

How is the chain split into upstream, midstream and downstream?

"Upstream" and "downstream" borrow from a river: a chip flows from design and manufacturing down to the product. Where to cut it into three differs from place to place. There are two common ways:

	Upstream	Midstream	Downstream
One: by how a chip gets made	Chip design (some include IP)	Wafer fabrication	Assembly and test
Two: by the whole industry	Materials, tools, design software, IP	Making the chip: design, fabrication, assembly and test	Putting chips into products: phones, computers, cars

The first is common in news reports in some places and looks only at the three stages of making a chip; materials and tools are either counted separately or folded into upstream. The second stands further back: the stages that make the chip are all in the middle, whoever supplies them is upstream, and whoever uses the chips is downstream. So when you see "upstream", check which one the writer means: it could be a design company, or a materials supplier.

Why does one weak link slow the whole chain?

The chain runs one stage after another: if one stage hasn't delivered, the next has nothing to work on, and the slowest stage decides when the product is done. A broken link is also hard to fix fast:

• You can't just switch: before a new material, tool or test-and-assembly house is used, it has to be tried in small runs to confirm the results come out the same (What is change control in a chip fab?).
• Catching up takes a long time: a wafer takes weeks to months from wafer start to finish, and building a new fab takes years.
• One missing chip holds everything up: during the chip shortage a few years ago, carmakers in many places cut production because they were short of chips (The news says "mature nodes").

In fab24hr

The main story, The life of a chip, follows a made-up hearing-aid chip called "Glowmere" through several stops on this chain. In the prologue, "Tape-out", a small company hands its design to a wafer fab, and it becomes a set of photomasks. Act one is the pilot run on the R&D line, act two the production fab, and act three the test-and-assembly house, where it's tested and built into a module smaller than a fingernail. In the finale it goes into a hearing aid and into Mr. Shen's hands. The chip passport in the codex lists the same five stages.

The R&D line also has a vendor field service engineer, who looks after a new etch tool from checkout to the customer's signed acceptance: tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "R&D line" and swipe to that job. In the production fab, as the litho or CMP process engineer or in Facilities, you sometimes run into trouble with something a supplier sent: a new batch of photoresist, or a gas cylinder whose certificate of analysis doesn't match. The companies, chips and people in the game are fictional; how the chain divides the work comes from public, general sources, and every company does it differently. For other jargon, see Chip fab jargon in plain English.