How long does a new chip technology take to go from a paper to mass production?

The news says things like "researchers build a 1 nm transistor", "a major breakthrough in 2D materials", "quantum chips" and "computing with light". It sounds as if your phone will have them in a year or two. In fact, a new chip technology often takes ten years or more to go from its first paper to mass production. Making one in a lab is only the first hurdle: it also has to be made the same every time, made cheaply and made to last, and a whole set of tools and materials has to catch up.

"Made in research", "in mass production", "widespread": what's the difference?

These three phrases in the news are very different distances from your phone:

What the news says	Where	How many
Made in research	A lab	One or a few; working is enough
In mass production	A production fab	Many wafers every day, sold to customers
Widespread	Many makers, many products	Most things people buy use it

What hurdles stand between a paper and mass production?

Step 1: Make one in the lab

A research team uses a new material or shape to make one working switch or memory cell, and writes a paper. This step proves the idea works. But it's often made by a method that can only build one, slowly, so nobody can say yet whether it could become a chip.

Step 2: Make many, all the same

Next, many have to be made across a whole wafer, all behaving about the same. A phone chip holds more than ten billion transistors, and they have to switch in step. If some drift too far, the whole chip may not work.

Step 3: Fit in with hundreds of other steps

A chip goes through hundreds of steps, and the new one has to slot in without clashing with what comes before or after: a new material may not survive the heat of later steps (all the heat a wafer takes along the way adds up to a total called the thermal budget), or it may contaminate other tools. When wiring switched from aluminum to copper, for example, a tiny amount of copper getting into the silicon makes transistors leak, so a thin film is laid around the copper first, and tools that touch copper are kept apart from the front end (see "How is the wiring inside a chip made?"). Fail this hurdle and, however good the new step is, it damages the steps around it.

Step 4: Tools and materials catch up

A lab tool only needs to handle a few wafers a day; a production tool has to process many an hour, around the clock, and every tool has to give the same result. So the companies that make tools and materials have to build production versions of the equipment, gases and chemicals, and measurement tools have to be able to measure the new structure. If a tool is too slow, each wafer costs too much; if you can't measure it, you can't tell whether it's right.

Step 5: Pilot runs and reliability tests on the R&D line

Then the whole chip is made on the R&D line with the real process flow, a pilot run; a lot is often split into groups with one condition changed in each (a split lot) to find the best one. The chip also has to pass reliability tests, such as running nonstop for 1000 hours at high temperature and raised voltage (HTOL) to see whether it fails early (see "What is burn-in?"). Ship without passing, and chips may fail in users' hands.

Step 6: Hand it to the production fab and bring the yield up

Once it's stable, the whole set of conditions is written up as the standard process (the POR), and it goes to the production fab only after a transfer review. This is called technology transfer (see "How is an R&D line different from a production fab?"). The production fab has to get several identical tools to make the same thing, and the yield has to climb step by step (see "Why do bigger chips have lower yield?"). If the yield is too low, each good chip carries too much cost to sell.

Step 7: Become widespread

At first, a new technology is often used only in the chips that need it most and can afford it; as more makers use it and the tools and materials get cheaper, it spreads to more products. Many chips keep using the methods of older generations, because they're good enough and cheap (see "The news says "mature nodes""). So on the day the news says "mass production", the things in your hands mostly don't use it yet.

Three examples: how long did each take?

Technology	Made in research	Mass production began	After that
Transistors stood on edge	A similar structure in the late 1980s; a university paper in the late 1990s	Early 2010s	By the mid-2010s, it was the main design in the most advanced generations
EUV lithography	1980s	Late 2010s	Still used only for the finest layers of the most advanced chips
Memory stacked upward	Second half of the 2000s (the approach in production today)	First half of the 2010s	More and more flash memory switched to it

• Transistors stood on edge: the channel the current flows through is stood up, and the gate grips it from three sides, so even a very small switch shuts off tightly. For how it's built, and how it later changed to a gate on all four sides, see "What are FinFET and GAA?".
• EUV: light with a wavelength of about 13.5 nm, which can print finer patterns. Prototype tools were built well before mass production, but the light source wasn't strong enough, so too few wafers could be printed in an hour, and it took years more to catch up (see "What is extreme ultraviolet (EUV) light?").
• Memory stacked upward: flash memory couldn't shrink much more on a flat surface, so the cells are stacked layer by layer and then drilled through from the top; it now has more than a hundred layers. Of the three, this one moved fastest; the idea of stacking upward had been proposed even earlier.

Will phones use new ways of computing soon?

Most are still in research. The work in phones and computers is still done by transistors (see "What is a transistor?").

• Quantum computers: experimental machines exist, and some only work when cooled close to absolute zero. They're good at a few particular problems; they aren't meant to replace the chip in your phone.
• Computing with light: using light to "send" data is already done in data centers (see "The news says "silicon photonics""); using light to "compute" is mostly still in research.
• 2D materials: materials only one layer of atoms thick, which might make even smaller switches. But growing them evenly across a whole wafer and connecting them well to metal are still being worked out.

Is there a scale for how far a technology has come?

Yes. A common one is the technology readiness level (TRL), from 1 to 9: around level 3 the idea has been shown to work, level 4 means it has been validated in a lab, and level 9 means it's in use in a real system and proven. It started in space programs and is now used by many government research programs too. There's also a separate scale just for "is it ready to manufacture?", the manufacturing readiness level, from 1 to 10; one of its levels is "shown that it can be made on a pilot line".

Next time you see the news, read it like this

• "Researchers build a 1 nm transistor": step 1 is done; it proves it can be made. That "nm" isn't the same thing as the one in "a 2 nm process" either (see "The news says "2 nm"").
• "A major breakthrough in 2D materials": usually progress on steps 1 and 2, such as larger or more even films. Several hurdles remain before mass production.
• "Pilot production begins": it has reached step 5: making small numbers, finding problems and running reliability tests.
• "Mass production begins": step 6, the production fab starts making it in volume; the yield may still be climbing, and becoming widespread is further off.

In fab24hr

In fab24hr, tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "R&D line" and swipe to the R&D integration engineer. The evening shift is the night before the transfer review: two of the eight checklist items aren't done, the reliability test needs three more weeks, and the customer wants the chips two weeks early. Whether to recommend a risk start is up to you. The R&D line also has a vendor field service engineer, who looks after a newly arrived etch tool from acceptance testing to sign-off. Chapters 1 to 5 of the main story, The life of a chip, are on the R&D line.

The game plays out the last few steps of this road: from the first 12 wafers starting to the transfer review takes about four months; the research before that isn't in the game. The companies, chips, people, tools and schedules in the game are fictional; how long a new technology takes and which hurdles it passes come from public, general sources, with dates given only to the decade. For other jargon, see Chip fab jargon in plain English.