Morris Chang and the Origins of TSMC

Book review of the autobiography of Morris Chang.

If you were making a list of the most important companies in the world, you’d find Taiwan Semiconductor Manufacturing Company, or TSMC, somewhere very close to the top. It’s one of a tiny handful of companies, along with Intel and Samsung, that can make the most advanced microchips, and it has a commanding lead in that market. TSMC makes the microprocessors and 5G chips for Apple’s iPhones and Macbooks. It makes chips for both Intel and AMD. And it makes the chips for Nvidia, Meta, and Amazon that are powering the AI boom. TSMC is so important that it’s sometimes called Taiwan’s “silicon shield”, the idea being that fear of damaging TSMC might dissuade China from trying to invade Taiwan.

TSMC has a fascinating origin story: it was founded in 1985 by Morris Chang, who, after working in the US for Texas Instruments for more than 20 years, was enticed to go to Taiwan and help develop their high-tech industries as head of Taiwan’s Industrial Technology Research Institute (ITRI). Shortly after Chang arrived, he created TSMC as an ITRI spinoff. Despite the fierce competitiveness of the semiconductor industry and the comparative technological backwardness of Taiwan, within a decade under Chang’s leadership TSMC was making a billion dollars a year, and it continued to grow from there, ultimately becoming the most advanced semiconductor manufacturer on the planet.

Given the importance of TSMC, the story of Morris Chang and how he managed to create the company is of obvious interest. And that story has been told, by him, in a two-volume autobiography. The first volume, covering his birth up to his early years at Texas Instruments, was released in 1998. The second volume, which covers the rest of his life up to the present day, was released in November of this year.

Unfortunately, both volumes have currently only been published in Chinese. And despite the fact that we live in a world of infinite, free machine translation, and that the founding of TSMC is among the most important events in modern history, nobody seems to have bothered to translate them into English.

This is a situation I would describe as “extremely stupid”. So I bought both volumes from a Taiwanese bookstore, and translated them.

Notes on the translation

I experimented with a few different translation methods, including Claude 3.5 Sonnet (the LLM I use day to day) and Otranslator (a sketchy Chinese GPT-4o-based translation service that I’ve used before for PDFs). But ultimately what gave the best results, and was also the simplest, was the built-in translation tools in Google Chrome.

The translation isn’t great; there’s a lot of language clunkiness and awkward phrasing (though I find that even human translated Chinese often reads pretty awkwardly if it’s not done really well). The machine particularly struggled with translating names, especially English names. It would often translate an English name into a phonetically similar cluster of syllables, or simply a similar-sounding Chinese name. The name “Haggerty” for instance, is variously translated as Hegdi, Haig, Haigidi, Haggard, Hagdi, and Hagrid.

But despite these issues, the books were surprisingly readable, and I found Chang’s story quite compelling. It’s an interesting, inside look at both the evolution of the semiconductor industry and the genesis of the ideas behind TSMC.

Chang’s early years

Morris Chang (whose Chinese name is “Zhang Zhongmou”) was born in 1931 in Ningbo, China. His family seems middle class, or maybe upper middle class: Chang’s father works in finance for the local government, and then later for a bank in the city of Guangzhou. As a child Chang moves around a lot due to war: when the Sino-Japanese War breaks out in 1937, Guangzhou gets bombed, and Chang’s family moves to Hong Kong. When Japan attacks Pearl Harbor in 1941, Japan invades Hong Kong the same day, and after a year living in occupied Hong Kong Chang’s father takes the family across battle lines to Chongqing in Free China. A few years after the war ends, the family is forced to flee again back to Hong Kong as the Chinese Civil war worsens and Communist forces advance.

By now, Chang is about to graduate high-school, and he’s not sure what he wants to do next. He hopes to be a writer, but his father convinces him that it’s hard to make a living as a writer, and suggests Chang pursue technology or engineering. At the time, Hong Kong has no good universities for science and technology, so Chang’s father encourages him to apply to college in the US. Chang’s uncle works as a professor for Northeastern University, and helps him apply to Harvard, where he’s accepted. In a freshman class of around 1100, Chang is the only Chinese student (Chang notes that in total there were only 14 minorities or foreign students in his class - 1 black, 1 japanese, 1 from Africa, 3 europeans, and 8 hispanic.)

Chang thrives at Harvard, and at the end of his first year he's in the top 10% of his class. But while Chang ultimately hopes to go back to China in the future, his plans change when the Communists win the war, forcing the nationalists to flee to Taiwan. During his first year Chang was most interested in studying the humanities and literature, but now he thinks he needs to plan more for a future career. Chang still has no real interest in engineering, but (per his father) he thinks it will offer good career prospects, so he transfers to MIT to study mechanical engineering. He works hard, getting his bachelors in mechanical engineering in 1952, and a masters in 1953.

Chang plans on continuing school to get a phD, as that’s what most Chinese students did at the time (apparently believing that it’s the only reliable way to have a stable career in the US.) Chang takes MIT’s phD entrance exam twice, and fails both times; MIT’s rules prevent him from taking it a third time.

Dejected, and with his plans in shambles, Chang decides to look for work. He applies and gets offers from several places (he has a masters from MIT after all), and he narrows his options down to two: doing research at Ford Motor company, or helping to automate transistor manufacturing at a company called Sylvania. Ford seems like the obvious choice; it’s a huge, successful company that would offer Chang job security and the potential for career advancement, and Chang gets along great with his future supervisor. Sylvania seems chaotic, his future supervisor seems indifferent to him, and Chang has no idea what the heck a transistor is. He decides to take the Ford job.

However, the Sylvania offer is $1 dollar a month higher than the Ford offer. Chang calls Ford and asks if they’ll match the offer. But the person he speaks to is rude and dismissive, refusing to negotiate at all, and Chang gets so angry that he decides to take the Sylvania offer instead.

Chang’s foray into semiconductors

Chang starts work at Sylvania in 1955. At the time Sylvania is one of a handful of companies that have licensed AT&T’s technology and are manufacturing transistors. At Sylvania’s factory, transistors are manually assembled one transistor at a time (a good worker can make “dozens of transistors per hour”). Chang observes Sylvania’s manufacturing process; he doesn’t know much about transistors, but he knows that semiconductors are very sensitive to impurities and chemical changes, and suspects that workers are inadvertently damaging the transistors while welding them, greatly reducing yields. Chang develops a new welding method that doesn’t require so much heating of the transistor; yields improve, and soon every production line has adopted his new method.

Chang throws himself into learning more about transistors and semiconductors. He begins studying William Shockley’s textbook about them, working through it one painstaking page at a time. He has a colleague who’s an expert on semiconductors, who spends every night drinking at a bar; Chang sits at the bar with him asking him questions about semiconductors, going home to study more, then going back to the bar if he has more questions.

Chang’s automation project is quickly forgotten, and he gets promoted to head of the R&D department to help develop new transistor products. Over the next several years, his department develops several different transistors made from germanium (while today most transistors are silicon, germanium was the major transistor material until the 1960s.) Chang continues to study semiconductors, and begins to publish his own research papers about them in journals.

But despite Chang’s efforts, Sylvania’s semiconductor division is doing poorly, due to (according to Chang) poor leadership. The division is led by sub-par people without any background in semiconductors, who can’t take advantage of all the technical talent they have. Chang realizes that Sylvania is falling behind, and when he’s forced to lay off some of his subordinates, Chang gets fed up. He quits and takes a job at Texas Instruments as an engineering supervisor, moving his family across the country from Boston to Dallas in 1958.

Chang joints Texas Instruments

When Chang arrives at Texas Instruments (TI), he’s immediately impressed. Almost everyone is young, less than 40, and everyone works very hard, 50 hours a week or more. Some employees work so much that they bring cots into the office to sleep. And while there’s an obvious chain of command, there’s few trappings of status or hierarchy: unlike at Sylvania, there are no special parking spaces for executives, executives and workers eat together in the cafeteria, and high-level managers and production-line workers converse freely. And while leadership at Sylvania had no real technical background in semiconductors, everyone at TI seems to be an expert on them.

Chang’s first task at TI is to help resolve manufacturing issues with a transistor being made for IBM. IBM itself has a very hard time making this transistor, achieving around 5% yield (the fraction of transistors that are acceptable) on average. But when the process gets transferred to TI it doesn’t seem to work at all: engineers joke that the yields are “zero and stable”. Chang works obsessively to try to solve the problem, working from 8 in the morning until midnight day after day. He concludes that IBM’s process specs are no good, and makes changes to the process based on his knowledge of semiconductor theory and his production line experience at Sylvania. Gradually, yields start to climb, first hitting IBM’s performance and then surpassing it. Eventually yields are greater than 30%, a very high percentage at the time (today, 30% yields would be considered very low). Of the four transistors TI is making for IBM, only Chang’s line is exceeding IBM’s own yield rates. Chang soon gets promoted to R&D manager for germanium transistors.

Over the next two years, Chang’s department grows to more than a dozen engineers. During this period Chang meets Jack Kilby, and learns of his plan to try to make an integrated circuit. Chang is skeptical—it's so hard to make even one transistor that surely manufacturing a device with many integrated ones will never be practical—but he nevertheless offers his expertise when Kilby asks him for advice.

Eventually Chang is approached by a manager who tells him that he could eventually rise to the position of Vice President of R&D, but to do so he would need to get a phD, and that TI will pay for Chang to get one. Chang feels like he has no choice but to accept, even though he’s somewhat worried that he’ll be giving up opportunities by putting his career on pause while he’s in school. In 1961, Chang applies and is accepted into Stanford’s electrical engineering graduate program. He again works hard and graduates in 3 years, but when he returns to TI he finds that many of his former colleagues have advanced several rungs up the ladder, while he’s back in his old position. But Chang hopes that his new knowledge, and the fact that the people he worked with previously are now high in the company, bodes well for his future.

Upon his return in 1964, Chang quickly gets assigned to help fix another germanium transistor manufacturing issue for IBM. He resolves it, and gets promoted to manager of the entire germanium transistor group. By now silicon transistors are starting to displace germanium ones, but the germanium department is still TI’s largest and most profitable division. As manager of an entire department, Chang has much more to worry about than the technical aspects of transistor manufacturing, and he resolves to learn all other aspects of the business—marketing, pricing, finance, strategy, accounting, and so on. Chang continues to excel and get promoted, first to head of the silicon transistor department in 1965, and then to the head of the integrated circuit department in 1966.

When he arrives as head of TI’s integrated circuit department, the division is struggling. While demand is growing rapidly, there are numerous competitors which have driven down prices and profit margins. Chang devises several strategies to try to combat this. First, he takes advantage of TI’s advanced manufacturing capabilities to produce integrated circuits with more transistors on them (so-called “Medium Scale Integration”, or MSI). While TI’s current integrated circuits are standardized commodities that almost anyone can produce, few manufacturers can achieve MSI; if customers can be convinced of the value of MSI, the profits for TI will be much higher. Chang also works to raise integrated circuit yields, which at the time were less than 20%. He decides to start a new fab in Houston, very near the existing fab in Dallas. The new fab soon achieves yields of 40%, and in response to the competitive pressure the Dallas fab manages to raise their yields as well.

While leading the integrated circuit division, Chang is introduced to learning curve theory by the Boston Consulting Group. The learning curve is the observation that production costs tend to fall by a constant percent every doubling of production volume. Chang quickly recognizes the power of this for semiconductor manufacturing: high market share and large production volumes will drive costs very low, creating a moat of efficiency that will be hard for competitors to overcome. With TI’s high yields, the company’s manufacturing costs are already lower than its competitors, so Chang decides to regularly reduce the price of integrated circuits to put the screws to the competition and try to capture market share.

During this period, TI also starts setting up overseas factories, which leads to two other important developments for Chang. One is that Chang begins to visit Taiwan regularly after TI sets up a factory there. The other is that Chang observes that Asian semiconductor factories are able to achieve much better yields than American factories.

Chang’s downward slide at Texas Instruments

In 1972, Chang gets promoted again, to head of the entire semiconductor division. But at this point he begins to struggle, and his relationship with his boss and TI’s leadership becomes strained. He fails with a project to develop 16 kilobyte RAM, and Chang’s bosses undermine him by setting up a parallel project, outside of Chang’s control, to copy a competitor’s 16k product. As head of a division in a rapidly growing industry one of Chang’s main tasks is hiring, but because TI is located in Texas, far from Silicon Valley, he has an increasingly difficult time attracting talent. This is made worse by the fact that Chang’s boss decides to build a new fab in Lubbock, Texas, essentially in the middle of nowhere. Chang refuses an order from his boss to preferentially allocate semiconductor production capacity to Texas Instrument products, instead opting to divide it fairly between internal and external customers. As tensions grow, TI’s leadership begins to publicly criticize Chang.

Chang is also increasingly unhappy with the state of technical knowledge and decision making at TI. His immediate boss, Fred Bucy, has no semiconductor background, and the knowledge of the executive leadership is increasingly out of date (something they seem to be ignorant of). Chang tries to convince leadership to start buying manufacturing equipment from outside vendors, which is more advanced and can achieve higher yields than TI’s internally-designed equipment, but he’s unsuccessful. He also fails to convince them to increase R&D spending, which Chang views as a short-sighted decision to prioritize short-term profits over long-term competitiveness.

Increasingly unhappy, in 1977 Chang asks for a transfer, and gets moved to head of TI’s consumer products division. But things don’t get better. Consumer products is a relatively recent initiative at TI to sell semiconductor-based products directly, rather than simply being a supplier to other companies, but the company soon finds that selling to consumers is a much different business than selling to businesses. When Chang takes over, TI has three product lines: simple consumer calculators, more advanced scientific calculators, and electronic watches. Only the scientific calculators are profitable, but they’re a very small source of revenue; the entire division is losing money.

Chang isn’t able to do much to change this; he thinks the entire foray into consumer products is misguided. The only successful consumer product released during Chang’s leadership is the Speak and Spell in 1978: because it requires special speech synthesis chips that are difficult to design and fabricate, it can’t be easily copied by low-cost competitors like most of TI’s other products. But the profits from the Speak and Spell are relatively modest, and the overall trajectory of the division doesn’t change.

In 1979, TI releases a personal computer for the home, the TI-99/4. Chang thinks this is a mistake, and he’s proved right: the TI-99/4 sells poorly, fewer than 20,000 units, and is described as “an embarrassing failure” (though its successor, the TI-99/4A, is much more successful). As head of the consumer products group, Chang is blamed for this failure, and in 1981 is moved to a new position, Director of Quality and Productivity, what Chang later describes as “being put out to pasture.” His bonus and level of seniority are reduced, and he thinks it’s only a matter of time before his stock options are taken away and he’s asked to retire. Though Chang has great loyalty to TI, he’s finally fed up, and he quits in 1983.

General Instrument

Chang looks around for a new job, at one point considering a position as head of Kodak’s new digital photography department, but ultimately takes a job as president of General Instrument, an electronics manufacturer. He moves to New York City, into the 53rd floor of Trump Tower (Trump himself is his next door neighbor).

But Chang isn’t any more successful at General Instrument than he was in his final years at Texas Instruments. His subordinates are indifferent or hostile to him, and he soon finds that his vision of his new role doesn’t line up with the CEO who hired him. Chang thinks he’s supposed to help build up General Instrument’s most successful existing businesses, but he gradually realizes that his boss wants him to acquire small companies, reorganize their operations, and then turn around and sell them, similar to a private equity firm. Within a year, Chang is asked to resign.

During his short stint at General Instrument, Chang does have one notable experience. At one point he’s approached by an entrepreneur trying to raise $50 million to start a chip design company, but shortly afterwards the entrepreneur comes back and tells him that he doesn’t need the investment after all: instead of spending the money building a fab, he can simply outsource all his manufacturing to other semiconductor companies. (The entrepreneur is Gordon Campbell, who would go on to found Chips and Technologies, one of the first fabless semiconductor companies.)

ITRI

While Chang was struggling at Texas Instruments, on the other side of the world Taiwan was making plans to develop its high-tech industries, and become more than a source of “plastic toys and low-cost running shoes”. In 1973 Taiwan founded the Industrial Technology Research Institute (ITRI) to help develop its industrial and technological capabilities. Taiwan’s industrial policy playbook was largely based on what had worked in Japan and Korea, and ITRI was based on similar institutes that had been founded in those two countries.

One of the technology-based industries ITRI aimed to develop was semiconductor manufacturing, and in 1976 Taiwan licensed an obsolete 7-micron semiconductor manufacturing process from RCA, who was exiting the business. Before long ITRI had built a full-scale semiconductor production line using the new technology.

Realizing that they would need outside expertise if they wanted the new semiconductor industry to succeed, Taiwanese leadership had been gently courting Morris Chang since the 1970s. During his time at TI Chang was invited to visit ITRI, and he authored a report on how Taiwan could develop its electronics industry. When ITRI was looking for technology to license, they asked Chang if TI was interested; though Chang declined (TI didn’t license any of its technology at the time), he helped them with their acquisition process, and confirmed that RCA’s CMOS-based technology was a good buy. And Chang was actually asked to become head of ITRI in 1982, before he even left TI, but he was unable to come to a workable arrangement with his boss and declined. After Chang leaves General Instrument, he’s once again asked to become head of ITRI, and this time he agrees.

The purpose of ITRI is to help develop Taiwan’s technical industries, but when Chang arrives he finds it far too focused on academic-style research, and much too disconnected from the needs of industry. He comes up with a three-point plan to change this. First, ITRI should change its funding model so that half of its funding comes from industry. This will ensure that ITRI is researching things that are actually useful for private industry, and not just what’s interesting to the researchers. Second, to help diffuse the technology it develops, ITRI should start to spin off companies, staffed by ITRI employees. And third, the worst-performing ITRI employees should be put on probation; at the time no ITRI employee had ever been fired, and there was no mechanism in place to even give a warning. Chang hopes this will improve employee motivation and performance.

Chang’s reform program is almost a complete failure. His idea to get industry funding is rejected, and he’s only able to maintain his employee probation program for one year. He’s able to spin off several companies, but none of them seem like immediate, obvious successes. Chang once again finds himself at odds with an organization he’s tasked with leading. His colleagues dislike him because of various failures of etiquette (such as asking questions during presentations, something considered normal in America but rude in Taiwan), and ITRI employees are angry at the threats to their job security the company spinoffs and probation policies represent. Eventually, Chang’s supporters in Taiwan’s government move on or retire; overwhelmed by opposition, Chang decides to leave ITRI in 1988.

During his tenure at ITRI, Chang notes that he only managed to accomplish one big thing, something which occupies almost half his time while he’s there: founding Taiwan Semiconductor Manufacturing Company.

Founding TSMC

On his second day at ITRI, Chang gets his predecessor’s to-do list. The first thing on the list is to deal with requests from three different integrated circuit companies to build semiconductor fabs. The requests have put ITRI in a bind: funding the construction of three separate fabs will cost more than $100 million, an enormous amount of money for ITRI to spend. On the other hand, ITRI’s whole reason for existence is to help develop industrial technology, and ITRI has spent years trying to develop its semiconductor manufacturing capabilities. How can they now turn around and reject requests that will help expand it? Leadership feels stuck, but Chang sees a way forward.

At the time, most semiconductor manufacturers both designed chips and manufactured them. A company might rent out its extra fab capacity if it happens to have some, but it's hard for other companies to rely on this. Using another company’s fabrication capacity typically requires very close relationships with it; it's not the sort of thing that someone can just walk in off the street and buy. And when using another company’s capacity, someone who might be a competitor, there’s always the risk that they’ll appropriate your chip designs, or prioritize their own production over yours. Indeed, Texas Instruments had been sued by companies arguing that TI had unfairly prioritized manufacturing chips for its own internal use; TI had only prevailed in that lawsuit because Chang disobeyed his boss and refused to do this.

So Chang knows that it’s hard to simply “buy” someone’s else's fab capacity. But he also knows that it’s getting harder and harder to build your own fabs; the cost of them keeps going up and up. Because of his knowledge of the learning curve, Chang also knows that it's very important for fabs to have as large an output as possible to achieve low production costs, and that current Taiwanese fabs are much too small.

Chang has already seen entrepreneurs trying to start semiconductor companies that won’t have fabs at all, and will simply rent the capacity they need. And he remembers the very high profit margins Texas Instruments achieved when it manufactured custom-designed products for other customers rather than standardized ones. He’s also seen what can be achieved by a fab using vendor-supplied manufacturing equipment; building a fab no longer means having to build all the manufacturing equipment yourself.

ITRI doesn’t have much in the way of semiconductor manufacturing capability. Thanks to licensing RCA’s technology, it’s built a real integrated circuit manufacturing line, but the technology was outdated when it was licensed, and while ITRI has pushed it forward (going from 7 microns to 5 microns to 4.5 microns), the rest of the industry has advanced faster, and ITRI only fallen further behind the cutting edge. It has no ability to design or market its own chips. ITRI’s only real advantage is that it has managed to achieve very high process yields, something Chang has repeatedly seen with southeast asian semiconductor fabs.

Based on all these factors, Chang proposes that instead of building a separate fab for each firm requesting one, ITRI should build a single “common fab” to be used by all three. This operation will be a corporate spin-off from ITRI, staffed by existing ITRI personnel. And it won’t just make chips for these three companies: the new company will offer “foundry” semiconductor fabrication services; anyone who wants to have a chip built by them can. It won’t do any design work, or compete with any of its customers in any way. All the chips it makes will be designed by others.

Chang hopes that this new business can eventually operate at the cutting edge of semiconductor technology, but as conceived its essentially a customer service play: its “product” will be reliable contract semiconductor fabrication services that right now don’t exist. Customers can get their chips fabricated without needing to develop special relationships, or worrying that they might get booted to the back of the line. It won’t offer cutting-edge technology, but not all chips need the latest and greatest process node. Chang thinks that with the rising cost of building a semiconductor fab, such services will only get more attractive, and that reliable contract manufacturing will allow for the creation of many more “fabless” companies that don’t have their own fabs at all.

Chang’s plan faces opposition, but he eventually convinces Taiwan’s leadership that funding this new venture, to be called Taiwan Semiconductor Manufacturing Company, or TSMC, is worthwhile. However, there’s no way to fund such a venture without private investors contributing, and private investors won’t contribute if the project looks too risky. To create confidence in the project, Chang needs to find an existing semiconductor manufacturer to partner with, who will contribute a significant portion of the funding.

Chang asks all the major semiconductor manufacturers, including Intel, Texas Instruments, Motorola, AMD, Panasonic, and Sony if they want to invest. Though everyone knows and respects Chang, they all turn him down. The only company that’s interested is Phillips, a second-rate semiconductor manufacturer from the Netherlands. Phillips agrees not because the idea of TSMC seems especially compelling, but because they want to take advantage of Taiwan’s low-cost labor and expand their factory operations there, and investing in TSMC seems like a good way to win favor with Taiwan’s government. Ultimately Phillips invests $40 million in the new venture, private investors contribute $35 million, and the government invests $70 million.¹

To run the company, Chang needs someone with close connections to the European and American semiconductor industries. While TSMC’s early customers will be Taiwanese, Chang knows that long term it will only be successful if it can sell to the larger American and European markets. Chang thinks he himself is best suited for this, but because he’s running ITRI Chang instead hires Jim Dykes, the former director of GE’s semiconductor business, which GE is closing down. Dykes joins in early 1987, shortly after TSMC’s first factory (a repurposed production line in an ITRI building) comes online.

TSMC starts operations

As expected, TSMC’s first customers are Taiwanese semiconductor companies, but in late 1987 the company gets an unexpected visit from Intel. Intel wants to use TSMC to produce some of Intel’s older, less advanced products so it can free up capacity to produce more leading edge chips. Intel first gives TSMC a trial run to see if TSMC’s yields will meet Intel’s requirements; when they do, Intel becomes TSMC’s first American customer, but not before requiring TSMC to establish a rigorous production control system, including quality control, preventative maintenance, and statistical process control.

Landing Intel as a customer sends a strong signal of TSMC’s competence, and Intel is soon followed by many other major semiconductor manufacturers wanting to outsource portions of their production: Motorola, Texas Instruments, Phillips, and so on. In some cases, not only are TSMC’s costs lower than the company doing the outsourcing, but its yields are higher.

By 1988, TSMC is cash-flow positive, and (after Chang successfully navigates board opposition) the company breaks ground on its second factory. This factory is what really begins to drive TSMC’s success: while the first factory was a repurposed ITRI building, the second factory is a new, greenfield development. Not only will it operate at a much larger scale (30,000 wafers a month vs 13,000 in the first factory), lowering per-unit production costs, but it marks the beginning of TSMC’s efforts to push its manufacturing technology forward. The new factory is the first semiconductor factory in the world to adopt the SMIF (Standard Mechanical Interface) wafer-carrying pod The predecessor to today’s FOUPs, SMIFs act as micro-cleanrooms for wafers, reducing their exposure to environmental contaminants and decreasing defects. TSMC also takes a risk with its lithography machines, the beating heart of a semiconductor fab; while most semiconductor manufacturers use machines from Nikon or Canon, TSMC opts for higher-performing, but relatively untested machines from a new company called ASML (in part this is because ASML was spun off from Philips, TSMC’s major investor.)

SMIF pod, via Wikipedia.

(At some point during these events, Chang takes over as head of TSMC, but the book is weirdly silent on the specifics of that transition. Chang talks about Dykes’ successors but not what happens after he himself leaves ITRI.)

TSMC’s bets pay off, and the company grows rapidly; over a period of 10 years its revenues grow at an average rate of 49% annually, and by 1995 it has surpassed $1 billion in revenue. By 2000, TSMC is operating five factories, and producing more than 3 million wafers a year. And while much of its production is for other large semiconductor manufacturers like Intel, Chang’s prediction about the rise of “fabless” semiconductor manufacturers is born out. By 2000 60% of TSMC’s revenues are from fabless companies like Qualcomm, Broadcom, and Nvidia.

Conclusion

The first volume of the autobiography covers from Chang’s birth in 1931 to around 1964. The periods described above, covering most of Chang’s career at Texas Instruments and his tenure at TSMC up to around 2000, make up around half of volume two.

Unfortunately, I found the remainder of volume two surprisingly uninteresting. It consists mostly of discussion of high-level corporate operations: mergers, joint ventures, ownership agreements (there’s a lot on Chang dealing with Philips’ option to buy out TSMC), Chang’s thoughts on boards of directors and investment bankers, and so on. Conversely, there’s very little on what I was most interested in: the nuts and bolts on how TSMC continued to push manufacturing technology forward, adopting EUV, and so on. There’s a few interesting bits (such as how the deal for TSMC to make chips for the iPhone came about, or Chang and Nvidia CEO Jensen Huang resolving a business dispute over a pizza dinner), but they’re the exception.

What strikes me most about reading Chang’s story is that, despite his immense talent, how often he ends up failing (or at least ends up in situations where he can’t succeed). Much has been made about how foolish Texas Instruments was to pass over Chang, and how if they hadn’t maybe TSMC could have been founded in America, but the reality is more complex and more interesting. Chang not only gets passed over for leadership at Texas Instruments (due to his perceived failures in the consumer products group and his ongoing tensions with executive leadership), but he fails at his next two jobs as well. He’s asked to leave by the CEO of General Instruments, and he’s effectively forced to resign as the head of ITRI after mostly failing to reform it and angering everyone in the process. After he leaves ITRI (his third resignation in five years) Chang wonders if, per the Peter principle, he’s risen to the level of his incompetence.

And it seems like his one big accomplishment at ITRI, founding TSMC, wasn’t exactly a sure thing. Almost no existing semiconductor companies thought Chang’s plan was worth funding; the one that does, Philips, is more interested in currying favor with the Taiwanese government than it is with the merits of the business. It’s very easy to imagine TSMC not getting off the ground at all if things go slightly differently.

Chang’s immense success with TSMC looks obvious only in hindsight; nobody at the time, including the Taiwanese government, other semiconductor manufacturers, and Chang himself, saw it that way.

1. Footnote: TSMC also gets another $57 million in the form of a loan, from (presumably) the government.