The news says "mature nodes". Why do older chips still matter?

The "mature nodes" in the news are older generations of chipmaking that have been in mass production for years. "Mature-node shortage", "mature-node oversupply" and "expanding mature-node capacity" are all about chips made this way. Names like "2 nm" and "3 nm" stand for a generation of chipmaking (The news says "2 nm". What exactly is 2 nm? explains where the names come from). The newest few generations are called advanced nodes; the ones several generations back, which factories have made for a long time, are mature nodes. The news also calls them legacy chips.

How many nanometers counts as "mature"? Where does advanced begin?

There is no single line. One common way to put it: 28 nm and the generations older than that are mature nodes. Some people draw the line at a newer generation.

The line also moves with time. Today's advanced node will slowly become a mature node a few years from now, once newer generations arrive.

"Mature" means the tools, the materials and every step have been in use for years. Most of the problems that can happen have already been found and fixed, production is steady, and the share of chips that come out good (the yield) is high. A brand-new generation is the opposite.

Why not make everything on the newest generation?

The newest generation fits more switches (transistors) in the same space, so chips run faster and use less power. The processors that do the computing in phones and computers, and AI chips, need that most.

But many chips have a different job. They don't need that many switches. They need other things:

• High voltage and large currents: chargers and power chips turn the electricity from a wall socket or a battery into the voltages each part needs. High voltage and large currents call for bigger switches, with a thicker insulating layer inside each one. The newest switches are tiny and thin and can only work at very low voltages; a higher voltage can break straight through them.
• Heat, cold and long years of use: a chip in a car has to work parked outside in winter and next to the engine in summer. Cars, factory machines and medical equipment stay in use for many years.
• Cheap, steady and still on sale: an appliance or a car uses a lot of chips, and each one has to be cheap. The same chip also has to stay available for years, because swapping in a different chip often means testing the whole machine all over again.

Older generations do all of this well, steadily and cheaply.

Why don't analog chips like to shrink?

One big family of chips is analog. An analog chip doesn't handle signals that are only 0 or 1. It handles things that change smoothly: how loud a sound is, how hot it is, how bright the light is, how much charge is left in a battery. Sensors, power chips and chips that send and receive radio signals are full of analog circuits.

Making analog circuits smaller doesn't help much, and sometimes makes them harder:

• The newest generations run at very low voltages, so a signal has less room to swing up and down, and a little noise looks big. It is like someone speaking very quietly: a little background noise and you can't make out the words.
• Analog circuits use parts like resistors and capacitors. Their size depends on the value they need to have, and they don't shrink when the switches do.

So many analog chips stay on older generations.

Why are older ways of making chips cheaper?

There are three main reasons:

1. The tools have already been written off. Fab tools are very expensive and lose value every day after they are bought (Why do chip fabs run 24 hours a day? explains this). Many lines that make older-generation chips have been running for years, their tools are already fully depreciated, and each wafer carries much less of that cost.
2. Fewer steps. The newer the generation, the more steps a chip has to go through and the harder it is to make. Older generations take a shorter road.
3. Experience. Yields are high and surprises are rare, so little gets scrapped.

Many lines that make power, analog and sensor chips still use wafers 20 cm (8 inches) across, one size down from the 30 cm wafers the newest lines usually use.

How many things use them?

A great many. Cars, home appliances, chargers, factory machines, medical equipment and network gear are full of chips made on mature nodes. A newer car can hold over a thousand chips (depending on how you count), and many of them are made this way.

Phones too. The processor that does the computing is often on the newest generation, while many of the chips around it that handle power, charging and sensing are made on mature nodes.

Why does the news swing between shortage and oversupply?

Because demand changes fast and capacity changes slowly.

Demand can jump all at once: everyone buys more of some product at the same time, or everyone wants to keep a few more chips in stock. Capacity can't follow right away. Adding a line in an existing fab means buying the tools, installing them and getting them stable, which often takes many months or more; building a new fab takes even longer, measured in years. Even with free tools, a wafer takes several weeks to a few months from wafer start to finish.

A car missing one small, inexpensive chip can't be finished. That is why a mature-node shortage often hits very big things.

It also works the other way. During a shortage, everyone adds lines, and they all come online a few years later. If demand has slowed by then, there is more capacity than work, fabs compete for orders, and the news says "oversupply". Shortage and surplus often take turns like this.

Will mature nodes be phased out?

No. Mature and advanced nodes split the work; the new ones are not replacing the old. Chips that need the most switches use advanced nodes; chips that need to handle voltage, last and stay cheap use mature nodes. One product often has both.

Mature processes are still being improved, too, with versions that use less power, last longer or add another function. The advanced packaging in the news can also make the part that needs it most on the newest generation and the other parts on older ones, then package them together (The news says "advanced packaging" explains this).

Next time you see it in the news

• "Mature-node chip shortage": chips made the older-generation way are wanted in larger numbers than the lines can make. Cars and appliances may be finished later because a few small chips are missing.
• "Mature-node oversupply": there are more lines that can make these chips than there is demand, and fabs compete for orders.
• "Expanding mature-node capacity": opening more lines that make chips the older-generation way, the kind used in cars, appliances and power supplies.
• "Advanced nodes and mature nodes": on one side, the newest generations that pack in the most switches; on the other, older generations made for years, steady and cheap. The two make different chips.

In fab24hr

The fab in the game is fictional and never says how many nanometers it is. The main story, The life of a chip, follows a hearing-aid chip called "Glowmere". In A day in the fab, play the furnace R&D process engineer on the R&D line, and in the afternoon shift you read one line: Glowmere uses a mature process generation, not the smallest transistors. What this chip cares about is something else. In chapter 10, "Standby Current", Leo, the customer, says standby current is the single most important number for Glowmere, because one battery has to last several more days. Standby current is the current that still flows while the chip is on standby, including what leaks through switches that are off. Even on a mature generation, the R&D line still has to tune new conditions and try new steps for it. The companies, chips and people in the game are fictional; what mature nodes are and why they still matter come from public, general sources, and where there is no single agreed number, only the general picture is given. For other jargon, see Chip fab jargon in plain English.