TSMC: The Quiet Titan
How Taiwan's Semiconductor Giant Shapes the Global Technology Landscape
Authors note: Thanks to everyone who liked and commented on my post about Taiwan’s semiconductor industry. Because of popular demand, I’ve written the following to delve deeper into TSMC's history. While this post first appeared on
’s (excellent) substack, I’ve added an addendum on TSMC’s shocking announcement about its success in Arizona.Now that Nvidia is worth over $3 trillion and has become the world’s most valuable company, we can officially say that semiconductors have gone mainstream. While its stock price is fantastic - rocketing over 3,000% in the last five years - what’s perhaps most surprising is how few people Nvidia has needed to get there. While Microsoft and Google, numbers two and three on the most valuable company list, boast over 200,000 and 160,000 employees, respectively, Nvidia has 30,000, meaning that each one “contributes” over $100 million - nearly 7x as much as Microsoft and Google.1
Nvidia can achieve this type of value-per-employee because it has, literally and figuratively, offshored its manufacturing process. Large plants would mean lots of employees and, consequently, would bring down its value per employee. Instead, it uses a variety of companies, most notably Taiwan Semiconductor Manufacturing Company, or TSMC as it’s better known, to do its literal dirty work. TSMC’s inimitable manufacturing process has allowed Nvidia and others, such as ARM, to realize their lofty plans in a reliable, cost-efficient manner, enabling everything from lane assist to iPhones. While Nvidia is undoubtedly the star of the semiconductor show, TSMC, as the behind-the-scenes fixer, is not making out badly: at $800 billion, it is in the top ten most valuable companies in the world, and its stock is up over 300% in the last five years.2
While Nvidia may have a higher stock market valuation, TSMC’s unique attributes make it more important from a geopolitical perspective. First, as its name suggests, TSMC is based in Taiwan and, therefore, at the center of the tension between China and the US. Second, while designing semiconductors may be the highest value activity, as we all learned in the pandemic, all those plans are useless if they cannot be manufactured. TSMC’s dominant market share in manufacturing the most cutting-edge semiconductors means that any country that wants a robust artificial intelligence industry will have to rely on it. And every country wants a strong AI industry.
What’s most incredible about all of this is that, in 2024, a manufacturing process can be so difficult to copy. Since Henry Ford, companies have recognized that manufacturing processes can be a source of competitive advantage, with Michael Porter notably codifying it as such in his famous work Competitive Strategy in the 1980s.3 But, previously, companies (and indeed countries) were often able to imitate results. The USSR couldn’t quite replicate Ford’s assembly line, but it was able to come close.4 And, GM has never been able to achieve the results Toyota can with its Just-in-Time system, but attempting it has increased efficiency.5
Things have been different with semiconductors, however. The USSR was never able to make very good semiconductors. And, despite the best efforts of challengers, no one has ever been able to come near TSMC at its own game. TSMC earns just over 60% of the world’s semiconductor manufacturing revenue, up from just under 50% five years ago.6 Samsung, Nvidia’s second supplier, comes in at a distant 11%. But, when it comes to the most high-end chips, its market share is nearer 90%. 7
Understanding how it got there, though, requires some further explanation…
The Project of a Nation
As I explained in detail in my previous post, TSMC was an outgrowth of Taiwan’s government’s interest in semiconductors. While I will leave the larger history of semiconductors and Taiwan to that essay, it’s important to emphasize that Taiwan was a national project. In and of itself, that doesn’t make TSMC special, however. In the latter half of the twentieth century, many Asian countries sought to escape poverty and various isms (feudalism and socialism, most notably) by harnessing growing technologies and leaning into markets.8 However, they did not do so as the West expected; they formed extensive industrial policies aimed at creating economic growth through technological modernization. The most famous of these is Japan’s MITI, the effectiveness of which has been a topic of vigorous debate for economic historians for decades.9
However, Taiwan was late to the party when it came to semiconductors. By the 1960s, when Taiwanese leader KT Li was looking to create a semiconductor industry, the US was dominant in designing and manufacturing semiconductors. What little business existed at the more commoditized end of the spectrum was owned by Japan. Nor was Taiwan the only country looking to enter: Korea had similar plans to create an industrial policy that would establish them as players in the semiconductor industry.10 Put simply, competition was steep and increasing, and the market was still relatively nascent. Moving into semiconductors would not have “resonated” with a strategy consultant.
But, in a story made for Malcolm Gladwell’s David and Goliath, Taiwan succeeded by being both lucky and good.
Taiwan was lucky in that its interest in semiconductors occurred just as a breakthrough in electrical engineering. Before the 1970s, relatively few transistors could fit on a single chip (Fairchild was proud of putting four transistors on a chip in 1960!) due to the tremendous complexity of integrating chips and circuits.11 Carver Mead and Lynn Conway developed a “Very Large Scale Integration” (VLSI) system that enabled many more transistors to be placed on a chip by separating the design and manufacturing of chips, enabling the use of computer-assisted designer (CAD) models.12 In essence, they developed a means of standardizing chip design so that each chip did not have to be custom-made.
A Brilliant Founder
Though Taiwan may have benefitted from auspicious timing, its strategy was absolutely spot on. First, it made one of the greatest hires of the 20th century: Morris Chang. Chang and TSMC are inextricably bound, like Jobs and Apple, and consequently, any history of TSMC is effectively a history of Morris Chang. But, unlike Jobs (and many other 20th-century business giants), he lacked the dramatic flair for marketing that we associate with colorful leaders today, succeeding more on efficiency and less on ‘brand.’
That is not to say that his life lacked color, however. Though something of a Taiwanese hero, Chang was born in mainland China in 1931, growing middle class at a time when the Chinese middle class was still relatively poor. The civil and Sino-Japanese wars led his family to move frequently, likely preventing him from ever feeling he had roots in China. At 18, he left for Harvard, but, in a decision that warms every MIT’er’s heart, he transferred to MIT after a year “in order to get a good job.” He stayed at MIT to earn a master's, failing to obtain a PhD, and left to work at Sylvania Semiconductors briefly before moving on to Texas Instruments (TI), where his career would take off. 13
As he described in his own words, Chang was “on the fast track” by 1961, only three years into his tenure at TI.14 He quickly climbed the ranks and earned perks, including a fully funded Stanford PhD. Unlike others in the pantheon of twentieth-century technology, however, his gift was not inventing new technology nearly so much as his ability to manage production processes. Not that he didn’t understand the science - his initial success in TI revolved around solving a complex technical problem - but that his true gift was efficiency.
Though the ability to make people and machines move with harmonious efficiency is always prized in business, this is especially true of semiconductors. Semiconductors are unbelievably challenging to manufacture because any slight flaw renders them useless. Success requires an entirely pristine, sterile environment, a highly skilled workforce able to work around the clock, and the ability to cut and weld objects with a tolerance measured in nanometers. And all the while, the market is constantly changing, forcing manufacturers, called foundries, to adopt new production processes. Consequently, someone who can wring increased efficiency from the process can expect a long and successful career in the industry.
Chang rode this skill to the top of TI, becoming General Manager of integrated circuits and successfully building several factories, most notably in Canada. But, alas, Chang’s career at TI stalled out in the late 1970s when TI passed him over for the CEO role, causing him to leave TI and work briefly as a COO at General Instruments. Uninspired by the work there, he left, wondering if retirement was all that was left for him.15
Li had other plans. Though trained as a physicist, Li has been immortalized in Taiwan as the leader of Taiwan’s economic miracle. It’s unclear how well Li and Chang knew each other before Li approached Chang to come to Taiwan- it appears Chang did take a business trip there with TI years earlier - but I’m unable to substantiate whether the two met then.16 Regardless, in a testimony to Li’s research skill and long-term vision, he recognized that Chang was just the man needed to build a semiconductor industry in Taiwan. As Chang himself explained:17
So I knew how competitive it was, and how difficult it would be to carve out a niche for a new Taiwan company. So that was the second thread of thought. The third thread of thought was I paused to try to examine what we have got in Taiwan. And my conclusion was that very little, you know. What strengths have we got? The conclusion was very little. We had no strength in research and development, or very little anyway. We had no strength in circuit design, product design, IC product design. We had little strength in sales and marketing, and we had almost no strength in intellectual property. The only possible strength that Taiwan had, and even that was a potential one, not an obvious one, was semiconductor manufacturing, wafer manufacturing. And so what kind of company would you create to fit that strength and avoid all the other weaknesses? There was pure-play foundry.
In essence, Chang recognized the implications of the Mead-Conway revolution and that no one else in the world was taking advantage of it. The idea of TSMC was born!
In addition to its superb choice of talent, Taiwan made one further strategically excellent decision: it invested heavily. Unlike Western companies, which had to rely on venture capitalists, public markets, or acquirers, Taiwan’s government invested heavily in companies, specifically in TSMC. Though outsiders were asked for funding, only Philips was willing to give TSMC any money. The rest came from Taiwan and wealthy individuals in Taiwan whom the government cajoled to invest.18
A Classic Growth Story
In its first years, TSMC struggled, as would-be investors, including Motorola, had predicted. Its only initial business was for small Taiwanese electronics companies, mainly based on a license for RCA that Taiwan had acquired. Starting in 1987, when the company was founded, TSMC saw losses through 1990 but then slowly began to turn a profit. By the early 1990s, the US was undergoing the “fabless revolution” as companies like ARM and Nvidia were born, and companies like Broadcom and Qualcomm began to focus on designs, not manufacturing. In doing so, they vindicated Chang’s vision and rewarded Taiwan’s bet.19
During this time, US semiconductor companies went from scoffing at the pure-foundry idea, as they had done in the mid-1980s, to building their business models around it. Doing so offered several advantages. First, consistent with the rise of dot-coms and Silicon Valley VC’s love of software, they could focus on the easily scaleable aspect of their business. Second, having a foundry in a country like Taiwan with a low cost of living meant TSMC could have manufacturing costs that US companies would never be able to imitate. Outsourcing manufacturing to TSMC was much more palatable than competing in the US or managing an offshore empire. Finally, in a nod to Fordism, TSMC’s specialization (and Chang’s unique managerial gifts) meant that TSCM could produce semiconductors with an accuracy no other firm could compete with.
By 2000, the nearly 70-year-old Chang finally began to see his success translated into revenue. Though unprofitable in 1990, by 2000, TSMC was bringing in nearly $5B in revenue a year, and turning a handsome profit.20 In the 2000s, Broadcom and Qualcomm closed their own factories, and others began to wind down their manufacturing, leading TSMC to increase its revenue growth steadily. But nothing could have prepared them for the windfall that came next.
As the graph above shows, TSMC did well before 2010, but revenue shot up like a firework after that. In 2010, Chang wrote shareholders that TSMC had spent $355M in R&D, nearly double its R&D spending in 2009. Its 28-nanometer technology was ready and would lead the industry.
These investments could not have come at a better time.21 First, the iPhone, launched in 2007, would be accompanied by the iPad, launched in 2010, and both would come to rely heavily on TSMC. In 2014, TSMC would become Apple’s leading supplier, producing its A8 chip. By 2022, Apple would be TSMC’s single largest customer, accounting for nearly one-fourth of its revenue. After TSMC agreed to favorable terms, Apple ultimately bought its entire 3nm production line, ensuring no competitor could use the chips for at least a year. And this is just Apple. Not only did Apple’s technology require a lot of TSMC chips, but their devices also enabled an entire IoT world that required chips and processors, pulling TSMC’s revenue growth up.
Second, and even more critical of late, the rise of machine learning and artificial intelligence dramatically expanded TSMC’s business. As noted by MIT Computer Scientist Neil Thompson, many of the statistical theories underpinning AI have been around for decades; they were held back by a lack of computational ability, often shortened to “compute” and a lack of data.22 Data was growing, thanks in no small part to social media. But, this left compute as the limiting factor.
Here, Nvidia earned its money by solving this problem somewhat serendipitously through video games. Jensen Huang, the rock star CEO of Nvidia, began Nvidia looking to make GPUs, or semiconductors for graphics cards, focusing on 3D graphics. To get the most out of those graphics cards, he developed software in 2006 called “CUDA” that allowed the cards to be programmed. Together, this created the environment needed for AI to take off. As Berkeley Computer Scientist and Deep Learning legend Andrew Ng explained:23
I remember at Stanford when my students were telling me, “Hey Andrew, there’s this thing called CUDA, not that easy to program, but it’s letting people use GPUs for something different. Could we build a server to use GPUs and see if they could scale-up deep learning?”And, one of my students at the time, Ian Goodfellow, who was my undergrad, helped me build a GPU server in his dorm room. And that server wound up being what we used for our first deep learning experiments to train neural networks.
We started to see 10x or even 100x speedups training neural networks on GPUs because we could do a thousand or 10,000 things in parallel, rather than one step after another.
That’s a total game changer for what you can do with neural networks.
With AI researchers dependent on Nvidia and Nvidia dependent on TSMC, TSMC was set. In 2010, TSMC had revenue of just under $10B, but by 2023, it had over $70B. Perhaps even more promising, investors were willing to pay 33x earnings for the company, implying that they see the $70B behemoth as a huge growth opportunity.
The Road Ahead
While technology may be moving in TSMC’s favor, geopolitics is not. As Chris Miller details in his excellent Chip War, both the US and China recognize their dependence on Taiwan, and TSMC specifically, as they seek to lead in AI. If that wasn’t enough, the tremendous supply chain problems the US experienced in the pandemic has made the consequences of getting locked out of technology all too real, making both countries very keen to ensure they have access to TSMC.
In light of this, the Biden Administration passed the CHIPS Act in 2022 to subsidize semiconductor companies' production in the US. Under this act, TSMC agreed to create a new factory in Arizona to receive $6.6B in funds. Though some American companies received more—Intel, for example, received $11B—this still means that TSMC received almost a sixth of the Act’s $39B budget.24
While onshoring semiconductor manufacturing here will undoubtedly make Americans feel better, it is still being determined how much this will change the geopolitical dynamic. Though it may be manufacturing onshore, TSMC remains a Taiwanese company, and China continues to see Taiwan as its own. Though having a plant in the US would safeguard production, it’s unclear which chips the plant will produce or how many. Though he relinquished leadership to long-time TSMC executive Mark Liu, Chang has described US efforts to recreate Taiwan’s semiconductor industry in the US as “futile.”25
One suspects the recent case of ByteDance looms large in the minds of TSMC executives. Though “onshore,” the US has decided to allow TikTok to operate in the US only if China parent ByteDance sells it to a US company. Given the US’s long-running interest in TSMC, it’s hard not to see the US setting itself up for a similar maneuver with the CHIPS Act, were China to get control of TSMC.
That would be another chapter in the story, however. For now, TSMC reigns supreme in the world of semiconductor manufacturing. And what a valuable world it is.
Addendum: The Next Chapter
The next chapter arrived early! On September 6th, TSMC announced that its Phoenix location is seeing ‘similar’ productivity levels to its facilities in Taiwan. As I noted above and in my previous post, this announcement is surprising given the difficulty of replicating the precision semiconductors required.
The entire significance of this isn’t visible yet. But it’s clearly a victory for TSMC, the US, and the Biden Administration. While America’s best-known semiconductor company, Intel, continues its descent despite receiving nearly $20B in aid from the CHIPS Act, TSMC’s success provides a tangible success that would have been unlikely without government assistance. And, while it’s unclear exactly what ‘similar’ means (i.e., how close is ‘similar,’ and for which chips exactly?), increasing onshore production decreases the geopolitical risk Taiwan’s success creates, at least to some extent.
As noted above, this only partially solves TSMC’s problem. While this step likely avoids a world where TSMC faces Huawei-level challenges from the federal government, Byte-Dance, too, has had success in the US. In fact, that was its downfall. For now, one can imagine that TSMC is pleased that it can diversify its production facilities, especially given that taxpayers took the risk.
Nguyen, Britney. “Nvidia’s Stock Surge Means Its Employees Are Worth More Than $100 Million — Each.” Quartz, June 10, 2024. https://qz.com/nvidia-employees-worth-over-100-million-each-3 trillion-1851530069#:~:text=The%20generative%20artificial%20intelligence%20boom%20has%20propelled%20chipmaker,its%20employees%20is%20now%20worth%20over%20%24100%20million.
Daly, Lyle. “The Largest Companies by Market Cap in 2024.” The Motley Fool, July 2, 2024. https://www.fool.com/research/largest-companies-by-market-cap/#:~:text=Microsoft%20is%20the%20largest%20company%20in%20the%20world%2C,%28%242.15%20trillion%29%2C%20and%20Saudi%20Arabian%20Oil%20%28%241.87%20trillion%29.
Porter, Michael E. Competitive Strategy. Free Press, 1980.
See Freeman, Joshua B. Behemoth: A History of the Factory and the Making of the Modern World. WW Norton & Company, 2018, pages 213-225.
Nakamura, Masao, Sadao Sakakibara, and Roger Schroeder. "Adoption of Just-in-Time Manufacturing Methods at US-and Japanese-Owned Plants: Some Empirical Evidence." IEEE Transactions on Engineering Management 45, no. 3 (1998): 230-240.
Statista. “Semiconductor Foundries Revenue Share Worldwide 2019-2024, by Quarter,” July 2, 2024. https://www.statista.com/statistics/867223/worldwide-semiconductor-foundries-by-market-share/.
Seeking Alpha. “TSMC: By Far The Strongest Chip Company.” June 18th, 2024. https://seekingalpha.com/article/4699870-tsmc-by-far-the-strongest-chip-company
See Studwell, Joe. How Asia Works: Success and Failure in the World's Most Dynamic Region. Open Road+ Grove/Atlantic, 201l
Langlois, Richard N. The Corporation and The Twentieth Century: The History of American Business Enterprise. Princeton University Press, 2023, pages 427-428.
Miller, Chris. Chip War: The Fight for The World's Most Critical Technology. Simon and Schuster, 2022, 129-133.
Bauer, Luc Olivier, Wilder, E. Marshall. The Microchip Revolution: A Brief History. Amazon. 2020.
Chang, Morris. “Oral History of Morris Chang.” Interview by Alan Patterson, Interview. Uploaded by Computer History Museum. Computer History Museum, 2007. https://archive.computerhistory.org/resources/access/text/2013/05/102658129-05-01-acc.pdf.
Ibid.
Ibid.
Miller, Chip War, page 165.
Miller, Chip War, page 164
Chang, “Oral History.”
Miller, Chip War, page 167.
Gratton, Peter. “What Are "Fabless" Chipmakers?” Investopedia, January 31, 2024. https://www.investopedia.com/ask/answers/050615/what-are-fabless-chip-makers-and-why-are-they-important-semiconductor-market.asp.
Ibid.
TSMC. “TSMC 2010 Business Overview,” 2011. https://investor.tsmc.com/sites/ir/annual-report/2010/2010_Business_Overview_E.pdf
Leiserson, Charles E., Neil C. Thompson, Joel S. Emer, Bradley C. Kuszmaul, Butler W. Lampson, Daniel Sanchez, and Tao B. Schardl. "There’s Plenty of Room at the Top: What Will Drive Computer Performance after Moore’s Law?." Science 368, no. 6495 (2020).
Sequoia Capital. “Nvidia: An Overnight Sucess Story 30 Years in the Making,” November 30, 2023. https://www.sequoiacap.com/podcast/crucible-moments-nvidia/
Ngo, Madeleine, and Don Clark. “TSMC Will Receive $6.6 Billion to Bolster U.S. Chip Manufacturing.” New York Times, April 8th, 2024. https://www.nytimes.com/2024/04/08/us/politics/tsmc-taiwan-chips-grants.html
Ibid.
Yeah,just of recent, a new plant has been established in Arizona USA,TSMC is leading this time.