Welcome back! Today we’re going to zoom around again in some odd directions, and give a roundabout introduction to the semiconductor industry, touching on some of the following questions:

• How do semiconductors get designed and manufactured?
• What is the business of semiconductor manufacturing like?
• What are the different types of semiconductors, and how does that affect the business model of these manufacturers?
• How has the semiconductor industry evolved over time?
• How do semiconductor manufacturers approach risk-taking in their strategic decisions?

This last question on risk-taking is especially relevant to the chip shortage, because semiconductor manufacturing involves a lot of expensive, high-stakes decisions with consequences that arise several years later — something that influences the choice of how much manufacturing capacity to build. (Robert Palmer, a member of AMD’s Board of Directors, allegedly compared building semiconductor fabs to Russian roulette: “You pull the trigger, and four years later you learn whether you blew your brains out or not.”)

In part one of this series, I talked about how bad the chip shortage is, I presented a two-minute summary of some of the factors that led to the chip shortage, and I gave a sneak preview of some of its nuances. We looked at toilet paper shortages. And played Lemonade Stand. Yes, really.

Part two’s focus is mostly centered around a kind of 1980s-1990s “retro” theme. If you don’t remember the USSR, the Berlin Wall, Yugoslavia, or parachute pants, consider this a history lesson of sorts.

Every good history of Silicon Valley starts with how William Shockley left AT&T’s Bell Laboratories, where he co-invented the first working transistor with John Bardeen and Walter Brattain in 1947, to start Shockley Semiconductor, located in a Quonset hut in Mountain View, California, in 1956, and how the Traitorous Eight left Shockley Semiconductor and started Fairchild Semiconductor in 1957 with a secret meeting in the Clift Hotel, where they signed dollar bills as a symbolic contract.

This is not that kind of history.

Think of it, instead, as a sort of magical journey with an invisible, attention-span-deficient Ghost of Semiconductor Past, starting in 1983 — ooh, look, there’s Huey Lewis and Cyndi Lauper! — and bouncing around among the past few decades. A set of hyperactive case studies, in a matter of speaking. (This is a long article. A very, VERY LONG article — no, you don’t understand; I’ve written long articles before, but this is a record for me — which I hope is worth the time needed to read it. It could probably be four or five articles instead, but I have my reasons for not splitting it up.) We’ll be covering these sorts of topics:

• some impacts of the prevailing customer market from around 1975-1985, during the Microcomputer Revolution
• the value (or lack thereof) of certain media commentary on the chip shortage
• an introduction to semiconductor fabrication with MOS Technology and Commodore
• some concepts of microeconomics
• capital expenditures
• the history and economics of DRAM manufacturing
• purchases of new and used semiconductor fabrication plants

But first, back to 1983. Why 1983? I have some personal reasons — for example, I first played Lemonade Stand in the fall of 1983 — but, more importantly, that year marked a major turning point in the rise of the personal computer. There are two pieces of evidence I’d like to share: a quantitative graph, and one little excerpt from a newspaper article of the time.

Interlude, 1983: A Brief Introduction to the Microcomputer Revolution and the Semiconductor Industry

The graph is from Jeremy Reimer’s excellent writeup of three decades of the personal computer industry in Ars Technica in 2005. The sales growth of the Commodore 64 and IBM PC in 1983 was just insane. (Go read the whole piece if you have time.)

The excerpt is from an article in the December 10, 1983 issue of the New York Times, titled Under 1983 Christmas Tree, Expect the Home Computer:

This is the year in which the home computer will join the sled and the bicycle under the Christmas tree.

In numbers that outstrip even the most optimistic predictions, Commodores, Ataris and Colecos are being snapped up from the shelves. Americans have embraced the home computer as their favorite gadget for a Christmas present, replacing the food processors and video games of Christmases past.

“Last year, computers were new, unique and expensive,” said Egil Juliussen, president of Future Computing Inc., a market forecasting concern that expects 2.5 million home computers to be sold this Christmas, twice as many as last year. “This year, they’re cheap, and they have become the gift.”

Only six months ago, a fierce price war erupted among home computer manufacturers, sending many into a tailspin from which it appeared some would not recover. This year, the industry will lose almost $1 billion.

Price War Aids Consumers

But the same price war that badly hurt the industry has brought home computers within the reach of millions of families. At the same time, industry advertising has helped make people feel more comfortable with the notion of a computer in their home or, alternatively, apprehensive that their children will fail without them. Suddenly, the home computer glut has given way to a shortage.

So there you have it. Boom! Home computer explosion. Gluts and shortages. Sounds familiar.

1983 began with Time Magazine’s January 3, 1983 “Man of the Year” issue dedicated to the computer. This issue contained a special section on computers, including a feature titled The Computer Moves In along with articles on programming and a computer glossary. While the Man of the Year title (actually Machine of the Year) was judged based on the status at the end of 1982, I really think of 1983 as the Year of the Personal Computer. Sure, the preceding years each brought new and exciting developments:

• 1977: the Apple II, the TRS-80, the Commodore PET, and the Atari VCS
• 1978: the Intel 8086 processor and Space Invaders
• 1979: the Atari 400 and 800 and TI-99/4
• 1980: Pac-Man and the Sinclair ZX-80 and Commodore VIC-20
• 1981: the IBM PC and the arcade game Donkey Kong
• 1982: the Intel 80286 processor and Commodore 64

There was certainly a sense that The Future Is Here! and articles in various magazines and newspapers heralded a new age of technology. The New York Times had an article in December 1978 predicting that “The long‐predicted convergence of such consumer electronic products as television sets, videotape recorders, video games, stereo sound systems and the coming video‐disk machines into a computer‐based home information‐entertainment center is getting closer.”

But I stand by my choice of 1983 as the year that personal computers really took off, both in an economic and cultural sense. Here are a few other reasons:

The movie WarGames came out in summer 1983, and with the later September 5, 1983 issue of Newsweek about Neal Patrick and the Milwaukee 414s, we were all introduced to the subculture of “hackers”.

Compute’s Gazette began publication, bringing articles and free software (to those willing to type it in from code listings) to Commodore computer owners, as a spinoff of Compute! Magazine. As it explained in its premier issue:

Where is the demand coming from? Well, we estimate that Commodore is currently selling over 100,000 VIC-20s and 64s a month. Dozens of software and other support vendors are rushing to supply products for these rapidly growing markets. Personal computing power is now expanding at a rate far past that predicted by industry observers. With the recent price decreases in the VIC-20 and 64, we expect this trend to continue its dynamic escalation.

Electronic Arts published its first games in 1983; we’ll take a closer look at one of them.

In summer 1983 I went to a “computer camp” for a week where we got to try out programming with VIC-20 computers and saw WarGames at the movie theater. Later that year, my family bought a Commodore 64. (Like I said, personal reasons. Okay, you don’t really care about either of those events, but they made a big impression on me.)

The video game crash of 1983 resulted from a tidal wave of interest in hardware and software companies racing to fill a mass demand for video games in the United States. It got filled, all right. Price wars dropped the cost of video game consoles and low-end computers like the Commodore 64 to the point where they were more affordable — good for consumers, but bad for the manufacturers. Atari was one of the most heavily-hit companies, with so many unsold games and consoles that they actually buried this excess inventory in a New Mexico landfill. The glut devastated the video game console market for two years, until the 1985 release of the Nintendo Entertainment System — and was almost entirely predictable. As Wikipedia states:

Each new console had its own library of games produced exclusively by the console maker, while the Atari VCS also had a large selection of titles produced by third-party developers. In 1982, analysts marked trends of saturation, mentioning that the amount of new software coming in would only allow a few big hits, that retailers had devoted too much floor space to systems, and that price drops for home computers could result in an industry shakeup.

In addition, the rapid growth of the videogame industry led to an increased demand, which the manufacturers over-projected. In 1983, an analyst for Goldman Sachs stated the demand for video games was up 100% from the previous, but the manufacturing output had increased by 175%, creating a significant surplus. Atari CEO Raymond Kassar recognized in 1982 that the industry’s saturation point was imminent. However, Kassar expected this to occur when about half of American households had a video game console. Unfortunately, the crash occurred when about 15 million machines had been sold, which soundly under-shot Kassar’s estimate.

Oddly enough, in 1983 there was a semiconductor shortage:

Demand, which has picked up sharply in recent months mostly because of the boom in personal computers, is now exceeding supply for most kinds of silicon chips used in computers, telecommunications equipment and other electronic devices.

At a conference here, industry executives said customers had to wait as much as 20 weeks or more for delivery of their orders. On the one hand, the shortage is a blessing for the companies, which are selling all they can make. Price-cutting has ended, profits are improving, and company stock prices have risen.

Both gluts and shortages have related dynamics, which are relevant to understanding this year’s semiconductor shortage. The computer and semiconductor industries are both examples of cyclical industries, which undergo boom and bust periods somewhat akin to oscillations in an underdamped, marginally stable control system.

Why do we get these sort of business instabilities, especially in the high-tech world, and especially if we can see them coming? Part of the answer comes from time delays, and part of it from corporate motivation to manage the pain of capital expenditures, something that becomes very clear if you look at the business of DRAM (dynamic RAM).

Disclaimers

I need to get through a few disclaimers:

I am not an economist. I am also not directly involved in the semiconductor manufacturing process. So take my “wisdom” with a grain of salt. I have made reasonable attempts to understand some of the nuances of the semiconductor industry that are relevant to the chip shortage, but I expect that understanding is imperfect. At any rate, I would appreciate any feedback to correct errors in this series of articles.

Though I work for Microchip Technology, Inc. as an application engineer, the views and opinions expressed in this article are my own and are not representative of my employer. Furthermore, although from time to time I do have some exposure to internal financial and operations details of a semiconductor manufacturer, that exposure is minimal and outside my normal job responsibilities, and I have taken great care in this article not to reveal what little proprietary business information I do know. Any specific financial or strategic information I mention about Microchip in these articles is specifically cited from public financial statements or press releases.

There you go; you’ve been informed.

One other topic I want to cover, before returning to the early days of computing, is some recent gossip, to remind us of a few very important points.

Interlude, 2021: Tech Gossip (Blame, Excuses, Prognostication, and Silver Bullets)

It’s always interesting to hear What Some People Say about the chip shortage… bad news is a time when everyone gets to weigh in on the topic, like a relative repeating gossip at a funeral. (“Oh, I heard he wrote Jimmy out of his will for dating his ex-wife, such a shame, if only he’d cut back with the cigarettes maybe he would have beat the cancer.”)

There is a lot of this sort of thing in the news lately. You need to realize that no one has a satisfying answer, so in a vacuum, here’s what we get instead, a superficial coverage of the problem that touches on half-truths, wishful thinking, and speculation.

Pat Says

Intel’s CEO, Pat Gelsinger, was quoted in an article in Fortune titled “Chipmakers to carmakers: Time to get out of the semiconductor Stone Age” that automotive manufacturers should be making ICs on more modern technology processes:

“I’ll make them as many Intel 16 [nanometer] chips as they want,” Intel chief executive Pat Gelsinger told Fortune last week during his visit to an auto industry trade show in Germany.

Carmakers have bombarded him with requests to invest in brand-new production capacity for semiconductors featuring designs that, at best, were state of the art when the first Apple iPhone launched.

“It just makes no economic or strategic sense,” said Gelsinger, who came to the auto show to convince carmakers they need to let go of the distant past. “Rather than spending billions on new ‘old’ fabs, let’s spend millions to help migrate designs to modern ones.”

And in his keynote speech at that IAA Mobility trade show: (at 17:50)

I’m also excited today to announce our Intel Foundry Services. In March of this year, I announced IDM 2.0 that we are foundering our products, but we’re also opening wide the doors of Intel to be a foundry for others’ products as well. Today, we’re announcing the European foundry services at Intel 16 and other nodes out of our Ireland facility. And we believe this has an opportunity to help expedite the end to this supply shortage, and we’re engaging with auto and other industries to help build on those, uh, capabilities.

But I’d also say some might argue, well, let’s go build… most of those auto chips are on old nodes. Don’t we need some old fabs for old nodes? Do we want to invest in our past or do we want to invest in the future? A new fab takes four or five years to build and have production-worthy.

Not an option to solve today’s crisis. Invest in the future, don’t invest backwards. Instead, we should be migrating old designs onto new, more modern nodes, setting them up for increased supply and flexibility into the future.

Lisa Says

AMD’s CEO, Lisa Su, said in an interview at Code Conference 2021

If you think about the semiconductor industry, we’ve always gone through cycles of ups and downs where, you know, demand has exceeded supply or vice versa. This time it’s different. And what’s different this time is every industry needs more, and so, you know, the confluence of that means that there is a, there is an imbalance. I will say that there’s a tremendous amount of investment going in, so, uh, there are, you know, over 20 new factories are coming online this year… and, and, you know, 20 more, you know, 20+ more in… um… uh… in planning and, um, so it’s still gonna be tight, you know, this year’s tight, first half of next year likely tight, but it’ll get better, as we get into 2022.

…

… we’ve seen, um, some stuff about automotive shortages ‘cause there were some supply chain interruptions there, so— It’s just every market has seen the demand go up, and um, and the key here with these complex supply chains is you may need thousands of chips, you know, only one of them being short is going to cause you to not ship your system, and so there’s just a lot of, let’s call it, um, mixing and matching of, of these things… But— you know, what I will say is, it will be solved. Okay, we, heh, [chuckling] the semiconductor industry has been through these things and it, it will absolutely, um, you know, uh, normalize, uh, supply to demand.

[Kara Swisher, host, interrupting] And when do you expect that to happen?

[Su] Uh, I would say it gets better next year. You know, not, not, um, not immediately, but it will gradually get better, as, uh, more and more plants come up, and it takes, you know, we’re an industry that just takes a long time to get anything done, so, you know, it might take, you know, 18 to 24 months to put on a new plant, and in some cases even longer than that, and, so, you know, these investments were started, uh, perhaps a year ago, and so they’re, they’re coming online, you know, as we, as we go through, uh, the next couple of quarters.

Elon Says

Elon Musk said, in a tweet in August 2021:

Tesla makes cars for export in first half of quarter & for local market in second half.

As publicly disclosed, we are operating under extreme supply chain limitations regarding certain “standard” automotive chips.

Most problematic by far are Renesas & Bosch.

Musk also stated in Tesla’s Q4 2021 earnings call, when asked for some “color” on the supply chain situation — Wall Street analysts can never get enough color — near the end of the call:

Elon Musk: Well, last year was chip hell of many chips, uh, so, silicon carbide inverters — were certainly one of them, but, uh, um —

Drew Baglino (Senior VP, Powertrain and Energy Engineering): Honestly, there’s a lot of annoying very boring parts.

Elon Musk: Yeah. It’s a ton of very simple control chips that, run-of-the-mill, literally, you know —

Drew Baglino: [Inaudible] you know.

Elon Musk: Yeah, basic chips to control.

Drew Baglino: Voltage references, oscillators, so they’re very boring things.

Elon Musk: Yeah, exactly. Like the little chip that allows you to move your seat back and forth. [Chuckling] That’s, actually, was a big problem.

Drew Baglino: Yeah.

Elon Musk: … the… couldn’t make seats. Um, so, I — like — but a lot of these things are alleviating. I think there’s, there’s some degree of the toilet paper problem as well, where, um, you know, the toilet paper shortage, uh, during COVID, and uh, like obviously it wasn’t really a, a suddenly, a, a tremendous enhanced need for ass wiping. Um, it’s just people panicked in order to — and got every paper product you probably, you could possibly wipe your ass with, basically. Um, and I wasn’t sure, is this like a real thing or not? I actually took my kids to the H-E-B and Walmart in Texas to just confirm that this was real.

Indeed, it was. Um, and there was, there’s plenty of food and everything else, but just nothing, no paper products, um, that didn’t cause a split up. So, um, an odd choice for people to panic about. Um, those, those things are — so, end of the world’s coming, I think toilet paper is the least of your problems.

Um, so, so, I think we, we saw just a lot of companies over-order chips, uh, and, and they buffer the chips, um, and so we should see — we are seeing alleviation in that… almost every area, but the output of the vehicle is — uh, goes with the, the least lucky, um, you know, um, what, whatever, whatever the most problematic item in the entire car is, and there’s like, at least, ten thousand unique parts in the car, so, uh, you know, way more than that if you go further up the supply chain, and it’s just — it’s just, which one is gonna be the least lucky one this time? It’s hard to say.

He Says That They Say That Somebody Says

Robert Carnevale wrote in a Windows Central article titled “The global chip shortage’s cause has been found — it boils down to one company, says report”

As reported by DigiTimes, some Taiwan-based tech manufacturers — think smartphones, PCs, and related gadgetry — have singled out Texas Instruments as being at the epicenter of the chip shortage’s widespread production pandemonium (via WinFuture). In case the name “Texas Instruments” sounds familiar to you, that’s because you very well may have used one of its calculators in your lifetime. That’s the company being accused of having a vice grip on global technology output.

This accusation is based on the fact that Texas Instruments manufactures analog chips that are essential for duties such as PC voltage regulation. Said chips are a fundamental part of much computing technology, and are in a more dire supply situation than the advanced, specialized chips the likes of TSMC and co. produce.

The aforementioned Taiwan-based sources say Texas Instruments’ inability to ramp up production capacity is the fundamental problem underpinning everything else. The question now is whether this supposed culprit identification will have any impact on the U.S. government’s shortage-combatting plans.

The DigiTimes articles — there are two of them, one mentioned above and another titled “Highlight of the Day: TI reportedly slow in expanding production”) — are behind paywalls now; I managed to stash the content from the Highlight of the Day article, back in November (and have faithfully reproduced its typographical errors here) for whatever it’s worth:

Pure-play foundry houses have been keen on expanding fab capacity, but the notebook industry continues to see serious chip shortage. Some industry sources are blaming Texas Instrument for its insuffcient supply. Networking device makers expect chip and component shortages will persist in 2022. But in the passive components sector, makers have seen demand slow demand.

IC shortage unlikely to ease until TI ramps up output: The global IC shortage is unlikely to ease until Texas Instruments scales up its output, according to sources in the notebook industry, which see supply-side constraints caused mainly by TI’s insufficient supply.

Enough of the Rumor Mill

There’s also Renée Says (Ampere’s Renée James: “I think what we thought was that we, I, I think this demand signal might have been read incorrectly in 2019 and 20 and in addition to that it uncovered, um, the fragility of the supply chain for semiconductors worldwide and the decline in U.S.-manufactured semis, which as you know is not just a supply chain issue, it’s also a national security issue for us, so, um, you know, one of the things that companies like mine, a small company, a startup only four years old, however I’ve been in this business for over 30 years, we know that there’s some systemic long-term things that we need to go after to build the health of U.S. semiconductors and that is what the Commerce Department is focused on talking with all of us about tomorrow and working on over the next five years. I mean, this is not a short-term thing, it’s something we need to get after as a national agenda.”) and Cristiano Says (Qualcomm’s Cristiano Amon: “I think in general, uh, everyone is complaining about, uh, the shortage of legacy technology and uh… but even those are gonna come online and uh… we’ll feel good about where we’re gonna be in 2022.”) and some others, but I’ve reached my limit.

This stuff is just frustrating to read, and I feel dirty repeating some of it, but I want to make a point: None of the tech gossip is really that useful.

Pat Gelsinger wants to help Intel sell its foundry services to the auto industry, and make them more resilient to capacity issues by manufacturing their ICs in fabs (Intel’s fabs, of course!) with newer technology which are seeing more investment — and we’ll talk about why that might or might not be useful in another article. Lisa Su is trying to paint a picture to reassure… but AMD doesn’t have any fabs, so it doesn’t control the situation directly; instead, it depends on foundries like TSMC to manufacture its processors. I trust she knows much more than I do about the situation, but at the same time, there’s no way to give a detailed answer without likely being wrong or sending messages to the financial world that can only get her or AMD into trouble. Elon Musk is rich enough that he can say what he wants and blame who he wants. If you want to learn something useful from CEOs, go listen to their company’s earnings calls and pay attention to concrete details that they are relating about their own company. If they’re talking about anyone else’s company, they’re no longer the expert.

As for the Windows Central and DigiTimes articles… bah!

I suppose there are shreds of truth in most of these sources, but you have to read enough in aggregate about the subject that you can distinguish the shreds of truth from what’s just a rumor, and unless there are supporting data or references to back up the assertions, they’re still just part of the rumor mill. Some say the chip shortage will be over soon, some say it will linger on; in fact, one day in mid-November of last year I spotted both back-to-back in my news feed:

The answers are all over the place. The uncertainty and confusion is enough to make you just throw up your hands in exasperation. What can you do?

I’ve been looking for good summaries of the situation for several months, and most of the time, whenever I think I’ve found something useful, in the same article or video there’s another detail that’s just totally out of left field and discredits the whole thing. Like a YouTube video (5 reasons why the world is running out of chips) that started off well, but then mentioned “minicontrollers” (are those just very large microcontrollers?) and towards the end slid into a lost-in-translation platitude that the chip-making industry “will need to keep innovating” and “will need to increase its capacity by working quicker, building square miles, and employing more people and machines.” Argh!!!

(If you do want some useful summaries, look at the links I listed near the beginning of Part 1.)

I’d like to leave the present gossip, and return to a historical look at earlier decades, instead of trying to explain today’s shortages directly. There are a few reasons. (Aside from the fact that today’s news is a bit demoralizing, whereas the 1970s and 1980s and 1990s exist at a safe, nostalgic distance.) One is that history tends to repeat itself, so we can look at a few of the past semiconductor business cycles, in their entirety, with the benefit of hindsight, instead of this mystery situation where no one really knows what is going to happen. The second is that the systems of the 1970s and 1980s are simpler to understand and explain. And the third is that I can actually find a lot of material on semiconductor history — there seems to be an unspoken statute of limitations on proprietary technology, so that if you’re an outsider and you want to know how today’s microcontrollers are designed, you’re out of luck, but people are willing to pull back the curtain and talk about the chips from the 1970s and 1980s, and all the crazy clever things they had to do to make them work.

So: the semiconductor industry. Yes.

Semiconductors! Meow!

I’m going to make the assumption that since this is on embeddedrelated.com, that you’ve worked with embedded systems, and so words like “op-amp” and “MOSFET” and “microcontroller” make sense to you. If not:

Semiconductors for People Who Are Not Electronics Enthusiasts

I think it would be easier for the average person to understand the semiconductor industry if it we just called it the microelectronics industry. “Semiconductor” makes people’s eyes glaze over, and just says something about the material these products are made of — something that conducts more than an insulator, but less than a conducting metal — not what they do, which is to control electrical signals. So there are a bunch of semiconductor manufacturers. They make chips, also known as “integrated circuits” (ICs), those little things in the black plastic packages that go in your phone or your computer. These things:

Inside the chips are lots and lots of really small things called transistors that switch on and off and can be used to form microprocessors that run software and figure out how to get information from other computers through the air and display them on a screen as cat videos.

For most of this article, you won’t need to know anything about how chips actually work. (For the rest, you’ll just have to catch up or work around the jargon and technical details.)

Semiconductors for Electronics Enthusiasts

If you have worked on embedded systems, you’ve probably bought ICs and soldered them on circuit boards and read datasheets and schematics. Maybe even designed your own circuit boards.

You can make your whole way through a career just on that kind of information, without knowing much about what goes on inside a chip or how these manufacturers were able to make the chips, and just thinking of them as different brands of little black boxes. (I know, because I was one of those people during a good portion of my career, even after working at a semiconductor manufacturer for several years.)

If you stop thinking of the chip as a black box, and instead as a container for transistors, which has been manufactured in a factory and designed by a company which may be different than the one who owns the factory — well, then, things get a bit interesting. I think you should probably know something about these aspects, which I mentioned at the beginning of today’s article:

• How do semiconductors get designed and manufactured?
• What is the business of semiconductor manufacturing like?
• What are the different types of semiconductors, and how does that affect the business model of these manufacturers?
• How has the semiconductor business evolved over time?
• How do semiconductor manufacturers approach risk-taking in their strategic decisions?

Here’s a short version:

• Semiconductor ICs contain a “die” (sometimes more than one) which has been manufactured as part of a circular wafer.

• The surface of the die has been built up in a bunch of layers, using equipment that adds or removes materials selectively from parts of the die.

• Some of the layers are made out of semiconductors, usually altered in some way to change their electrical properties and form transistors.

• Other layers are made out of metal, forming lots of very small wires.

• The selective adding or subtracting usually happens with the help of photoresist that has been selectively exposed using light that shines through a glass mask. (The most cutting-edge process, EUV lithography, uses mirrors that bounce around ultraviolet radiation produced from tin plasma.)

• IC designers figure out what kind of transistors will fit on the die, how to hook them together, and how to test them. Nowadays everything is computerized, and specialized computer software called electronic design automation does most of the grunt work, but it wasn’t that way in the 1970s or even the early 1980s.

• There are different types of semiconductors; market analysis of the industry usually divides them roughly into the following categories (see for example the World Semiconductor Trade Statistics Product Classification):

• Discretes — mainly diodes and transistors. Power diodes and transistors are growing in sales; signal diodes and transistors are getting rarer but are occasionally used as “glue” circuitry.

• Logic — this covers FPGA and “application-specific integrated circuit” or ASIC and the system-on-chip (SOCs) found in cell phones.

• Microcomponents (general-purpose processors)

• MPU — microprocessor unit. Here’s where your PC or server processor gets counted, probably also the GPU for driving displays.

• MCU — microcontroller unit. These are embedded processors which usually have memory built-in and specialized peripherals, and they go into almost everything electronic these days.

• DSP — digital signal processors

• Memory — SRAM, DRAM, EEPROM, Flash, etc.

• Analog — op-amps, comparators, voltage references, voltage regulators and other power management ICs, ADCs, DACs, etc., and sometimes interface ICs for communications.

• Optoelectronics — LEDs, optoisolators, image sensors, etc. These usually use materials other than silicon.

• Sensors — gyroscopes/accelerometers, pressure/force/strain sensors, humidity sensors, magnetic field sensors, etc. (Temperature sensors are probably counted as analog.)

• Others that may not fit into the above groups, depending on who defines the categories. (RF transceivers, for example)

• Technological development of manufacturing processes has improved rapidly over the decades to decrease the feature size (sometimes referred to “technology node” or “node” or “process node”) found on ICs, that permits more and more transistors per unit area.

• Terms like “leading edge” or “trailing edge” or “mature” are used to describe feature sizes relative to the state-of-the-art at the time. (Leading-edge are the smallest.)

• The different types of ICs require different manufacturing processes, and have different dynamics with respect to technological advance and obsolescence.

• Memory, logic, and MPUs are the bulk of what’s driving the leading edge. Designs in these categories last a few years and then become obsolete by newer designs.

• Analog and discretes are longer-lived technology manufactured on older processes.

• Manufacturing facilities or “fabs” are extremely expensive, getting more so as the technology advances.

• Cost per transistor has been getting less and less expensive. (until, perhaps, recently)

• Increasing capacity by building new fabs is a risky proposition that requires predicting demand several years into the future.

• The cost to design an integrated circuit has been increasing as the technology advances, and permits more complex ICs on smaller process nodes.

• Semiconductor companies now sometimes outsource their fabs, and they are organized into several categories:

• Foundries — these are companies that run fabs to manufacture IC die for other companies

• Integrated device manufacturer (IDM) — these companies run their own fabs

• Fabless — these companies do not own their own fabs and rely on foundries

• Fab-lite — these companies may rely on a mixture of their own fabs and on external foundries

• Semiconductor companies are in an extremely competitive business; here are some reasons that companies go out of business or get acquired by others:

• Not innovating enough to retain a technological advantage

• Not controlling costs

• Making poor predictions of demand

• Taking on too much debt

That’s the three-minute boring version.

If you’re interested in some more in-depth Serious Summaries of the semiconductor industry, here are a few resources for further reading:

• Accenture, Harnessing the power of the semiconductor value chain (December 2021) — not a bad read.
• Jan-Peter Kleinhans & Dr. Nurzat Baisakova, The global semiconductor value chain (October 2020)
• Clair Brown and Greg Linden, Chips and Change, 2009. Sadly out of print, but worth buying a used copy. Paul McLellan has a nice summary on his EDAgraffiti blog.

The technological pace of innovation is something that makes the semiconductor industry different from almost every other industry, except for maybe magnetic disk storage. You’ll hear the term “Moore’s Law” which means that each year we can cram more transistors onto a chip, and that each year we can see more cat videos faster, as long as we buy a new phone, because the old one is no longer supported. Moore’s Law is a double-edged sword; in addition to making new electronics better every year, it fosters obsolescence in its wake. As I mentioned in the beginning of this article, Bell Laboratories invented the transistor in 1947, when there were no cat videos, but Bell Laboratories is a shell of its former self, and is no longer in the semiconductor business. Shockley Semiconductor never really succeeded, and didn’t make it past the 1960s. There are a lot of other defunct semiconductor manufacturers that have failed, sold off or closed their semiconductor divisions, or were gobbled up, and did not make it to the age of cat videos; the early decades of the industry are full of them: Mostek, Rheem Semiconductor, Signetics, Sprague Electric, Sylvania, Westinghouse Electric, etc. To succeed in the semiconductor industry requires not only business acumen, and sometimes a lot of luck, but also a never-ending stream of innovation — if any of those ingredients are lacking, it can send a company off to the chopping block.

My three-minute boring description of the semiconductor industry is rather abstract; I’d prefer instead to give an example. So with that in mind, here is one particular chip, a groundbreaking design that powered almost all of those early 8-bit computers in 1983, the MOS Technology MCS 6502.

Semiconductor Fabrication and the 6502

The 6502 was first delivered to customers in September 1975. This was one of a few iconic microprocessors of the late 1970s and early 1980s. To understand how big of an impact this chip had, all you have to do is look at its presence in many of the 8-bit systems of the era, sold by the millions:

• Apple II
• Atari 400 and Atari 800
• Atari VCS (6507)
• BBC Micro
• Commodore PET and VIC-20
• Commodore 64 (6510)
• Commodore 128 (8502)
• Nintendo Entertainment System (Ricoh 2A03)

The eventual ubiquitousness of the 6502-based personal computer was the end result of a long process that began thanks to Motorola for its pricing intransigence of the 6800 processor, and to Chuck Peddle and Bill Mensch for getting frustrated with Motorola. In March 1974, Motorola had announced the 6800, but did not reach production until November 1974, initially selling the chip for $360 per processor in small quantities. Chuck Peddle had been giving marketing seminars to large customers in early 1974 — he’d smelled opportunity and tried to convince Motorola to pursue a lower-cost version for the industrial controls market, but they weren’t interested.[1 page 24-26] By August, Peddle had hatched a plan, leaving Motorola and setting off across the country to join MOS Technology, a scrappy little integrated circuit manufacturer located near Valley Forge, Pennsylvania. Mensch, one of the 6502 designers who went to MOS with Peddle, says this: “The environment was a small company where Mort Jaffe, John Paivinen, Don McLaughlin, the three founders, had created small teams of very capable calculator chip and system designers, a quick turn around mask shop and a high yielding large chip manufacturing team out of TI. So you go from Motorola with, relatively speaking, an unlimited budget for design and manufacturing, to an underfunded design team with very limited design tools for logic and transistor simulation. We had to manually/mentally simulate/check the logic and use very limited circuit simulation. In other words, it was really low budget. The datasheets and all documentation was done by the design team.”[2][3] Peddle persuaded Mensch and six other Motorola engineers — Harry Bawcom, Ray Hirt, Terry Holdt, Michael Janes, Wil Mathys, and Rod Orgill — to join him and a few others at MOS in designing and producing what became the MCS 6501/6502 chipset. “At MOS John Paivinen, Walt Eisenhower, and Don Payne, head of the mask shop, and mask designer Sydney Anne Holt completed the design and manufacturing team that created the high yielding NMOS depletion mode load process,” says Mensch. “The result was the MCS 6501/6502, 6530/6532 Ram, ROM Timer and IO combo and 6520/6522 PIA/VIA microprocessor family.”[3]

Some technical details of the 6502 are slightly fuzzy after so much time has passed — but I have chosen to focus on the 6502 because it is such a well-known processor, and at least some details are available. Semiconductor manufacturers are notoriously secretive, and it is hard to find detailed descriptions of how modern ICs are designed and manufactured. Whereas there are plenty of sources of information about the 6502.

(A word about the numbered notes: I don’t normally use such things, preferring instead a blogorrhific style of adding hyperlinks all over the place to point towards further information on various topics. But in this article, I have used notes to cite my sources a little more formally, for a few reasons. First, because there are inaccuracies about the 6502 floating about on the Internet, I’m trying to be a bit more careful. And since I’m not an expert in semiconductor manufacturing or economics, I feel like I have to point toward some specific accounts that back up my statements. Finally, a citation is a little more robust than a hyperlink in case an online publication becomes unavailable.)

EDN had a nice technical writeup of the 6502 in September 1975. BYTE magazine covered the 6502 in November 1975, with more of a focus on its instruction set than the physical aspects of the chip itself. Mind you, both the