TSMC Deep Dive
Welcome to the forty-second Pari Passu newsletter.
Today, we are going to learn more about the world’s largest chip contract manufacturer, Taiwan Semiconductor Manufacturing Company (TSMC).
Introduction
In this edition of Pari Passu, we will be diving into Taiwan Semiconductor Manufacturing Company (TSMC) and assessing what made the company the technological, economic, and industrial powerhouse it is today. Founded in 1987 by Morris Chang, TSMC is the world's largest independent manufacturer of semiconductor chips, a leader in advanced process technologies, and a key player in the global technology supply chain. In this summary, we will cover TSMC's current performance as well as the company’s history and its possible headwinds and tailwinds for the coming years. This summary is dense, but each section can stand alone if you would like to divide up the reading.
Overview – What is a Semiconductor?
Let's start by reviewing the underpinnings of the semiconductor industry's value and importance. The key concepts we will be covering here include what semiconductors are, how they are produced, who produces them, and why they are so important.
Semiconductors are materials with electrical conductivity between that of conductors and insulators, making them ideal for controlling the flow of electrical current in electronic devices. Their ability to act as a conductor and an insulator is due to the band gap in the energy levels of the electrons in the material. The most common semiconductor material is silicon, although other materials like gallium arsenide are also used.
The process is semiconductor manufacturing is extremely complex, so here is a relatively simplified version.
The production of semiconductors begins with the purification of silicon through the Czochralski process, which involves growing a single crystal of silicon from molten silicon to create high-purity silicon
The purified silicon is then melted, slowly cooled to form a single crystal ingot, and sliced into thin wafers using a diamond saw
The wafers are polished, cleaned to remove surface impurities or defects, and prepared for fabrication
Photolithography (light) is then used to etch patterns onto the wafer surface by coating the wafer with a light-sensitive photoresist (special material), exposing it to ultraviolet light through a patterned mask, and developing and removing the exposed areas of the photoresist
The wafer then undergoes etching and deposition processes to create electronic components by removing unwanted material and adding new materials
Doping involves adding impurities, such as boron or phosphorus, to the semiconductor material to change its electrical properties
After fabrication, the wafer is tested to ensure the electronic components function correctly
Finally, the individual chips are cut from the wafer and packaged into their final form
The manufacturing process is extremely capital-intensive and consists of very high start-up costs, posing grave challenges for new entrants. To put it into context, only 50 chip manufacturing machines are made per year by the leading producer, Advanced Semiconductor Materials Lithography (ASML).
The cost of one machine is between $200 and $300 million
It takes seven Boeing 747s to ship the parts for one machine, which is the size of a school bus when assembled and operates at a temperature 40x hotter than the surface of the sun
Each machine must be operated by a highly skilled ASML employee
There are three primary semiconductor manufacturing businesses that are known for their ability to manufacture semiconductor chips with high yields: TSMC, Samsung, and Intel. These three companies accounted for 84% of ASML’s business in 2021.
Yields define the proportion of manufactured chips that are operational and can be installed in another device. Yields for semiconductor manufacturing are remarkably low. Even for the most established technology, yields sit at around 90%. For the latest, cutting-edge technology, yields are only 70%.
TSMC is a “pure play” foundry company, which means that they strictly manufacture semiconductor chips that are designed by chip design companies such as Nvidia, Intel, and Qualcomm.
Samsung has both a foundry side and a design side, so they primarily manufacture their own semiconductor chips.
TSMC and Samsung are currently the two powerhouses leading the semiconductor industry.
Intel is similar to Samsung with regards to their business model of having a foundry side and a design side. Intel used to strictly manufacture their own semiconductor chips but now their foundry business is open to manufacturing third-party designs.
Although these chips are extremely capital intensive and the machines used to produce them are expensive, they are a crucial part of our everyday life. They are the reason our smartphones, computers, and cars function and the barrier to unlocking functionality and speed in these machines.
Without these chips, the modern technological innovation boom that we have experienced would not exist. The constant back-and-forth game between software and semiconductor chips is what keeps driving innovation. As software gets better, chip designers and manufacturers improve their processes and vice versa.
Overview – Performance
Now, let’s dive into TSMC’s financial and operational performance.
Finance – TSMC’s current market cap is $444 billion, which makes it the 14th most valuable company in the world.
FY22 revenue came in at $73.7 billion, nearly a 30% jump from 2021. The increase from $57.3 billion to $73.7 billion was caused by the uptick in Apple iPhone orders and new developments in AI technology.
Cost of goods sold amounted to $29.8 billion and operating expenses totaled $7.4 billion. These expenses grew at 7.5% and 20.6% YoY, respectively.
Operating margin for TSMC was 49.6%, which is way ahead of its competitors, due to relentless manufacturing efficiency.
TSMC’s net income came in at just over $33 billion, signifying a 53.7% YoY growth rate.
Operations – TSMC operates 23 manufacturing facilities (fabs), primarily in Taiwan.
History – Pre-Founding
The founding of TSMC is intrinsically linked to the story of its founder, Morris Chang. Without him, TSMC and the larger semiconductor industry would not be as advanced as it is today.
Morris Chang was born in Ningbo, China in 1931. His childhood was marked by constant movement around mainland China and Hong Kong due to the Second Sino-Japanese War and World War II.
In 1949, at the age of 18, Chang moved to Boston to attend Harvard University. However, he soon realized that Harvard did not offer an undergraduate mechanical engineering degree, so he transferred to the Massachusetts Institute of Technology (MIT), where he completed both his undergraduate and master’s degrees.
Chang emphasizes how critical moving to the United States and studying at MIT was in furthering his intellectual curiosity and helping him become “the father of the Taiwanese semiconductor industry.” His first step towards obtaining that title was working for Sylvania Electric Products, which produced semiconductors.
Fun fact: Ford Motor Company offered Chang a position, but Sylvania’s starting salary was $1 more than that offered by Ford. Ford refused to match the salary and Chang decided on Sylvania.
At Sylvania, Chang studied Electrons and Holes in Semiconductors with Applications to Transistor Electronics, which was a textbook written by William Shockley who developed the original transistor.
Chang attributes his foundational understanding of device physics and semiconductors to this textbook.
After three years at Sylvania, Chang transferred to Texas Instruments (TI). At the time, TI did not have the consumer product line that they do today, but they were becoming a rising star in the semiconductor industry.
Chang rose through the ranks at Texas Instruments (TI) and became the manager of a technical division in 1961. Recognizing his talent, TI sponsored Chang to pursue a PhD at Stanford, and he successfully graduated in 1964. Upon his return to TI, Chang was able to improve semiconductor yields by 20%. This achievement impressed his superiors and helped him obtain his promotion to General Manager of the Integrated Circuitry (IC) division in 1967.
Chang now wanted to grab a hold of the senior management’s attention and did so by improving TI’s IC pricing model. He found that by using learning curve pricing (reducing the price every quarter even without market indication), TI could gain large quantities of market share.
Over the coming years, TI became the world’s largest IC business in the world
Chang was subsequently promoted to Vice President of Texas Instruments
Shortly after his promotion, Chang was transferred to TI’s consumer division. After being passed up for the position of CEO (he was second in line), Chang resigned in the late 1970s. He was then faced with two job prospects:
Become the COO of General Instruments (GI), a mid-tier electronics manufacturer specializing in semiconductors, or become a venture capitalist
Chang went to GI and quit after one year
History – Founding
In the mid-1980s, Taiwan was primarily a manufacturing nation with limited strengths in research and development, circuit design, IC product design, and minimal intellectual property.
Typical gross margin of a Taiwanese company was 4-5%
Almost all profits were arbitraged away
Annual household income per capita in Taiwan was less than $2,500
Although Taiwan was a poor manufacturing country, its manufacturing capabilities were solid and efficient. The Taiwanese government realized this and began establishing a path to technological innovation by striving to become the best semiconductor manufacturing country in the world. Furthermore, by becoming a key supplier to American technology companies, Taiwan’s fabs (and thus the country as a whole) would receive valuable protection from the American military in the event of a Chinese invasion.
The Industrial Technology Research Institute (ITRI) was established by the Taiwanese government
Kwoh Ting Li, one of the organizers of ITRI, recruited Chang to run the institute
Against the opinions of his peers, Chang moved to Taiwan to lead ITRI in 1985
ITRI’s first major move was starting United Microelectronics Corporation (UMC) with RCA semiconductors. UMC did chip design and manufacturing, which initiated the start of the modern wave of Taiwanese chip manufacturing.
UMC had solid performance despite producing chips that were two generations behind the leading innovators of TI, Motorola, and Intel
UMC built a fab to begin manufacturing semiconductors for third-parties
Chang’s quasi-success with UMC led Kwoh Ting Li to approach him again, but this time he asked Chang to build a Taiwan-based semiconductor company that would lead the world. In three days, Chang came up with a unique business model, known as a “pure play” foundry (manufacturing only).
At the time, however, there was no market for a “pure play” foundry, since most semiconductor companies had both a design side and a manufacturing side. Chang recognized the lack of market, but he remained optimistic regarding the growth of fabless semiconductor companies (research and design only). His optimism was caused by two visions:
IC designers who were eager to start their own design companies would leave their current positions to create startups
Those startups would use TSMC, rather than Intel or TI as their chip manufacturer, because of TSMC’s manufacturing expertise (which would also engender lower costs and greater efficiency)
Chang received approval from the Taiwanese government for 50% of the startup cost and was tasked with finding investors for the other 50%. His primary focus during this phase was to speak with the leaders in the industry, but they all shut him down. However, Chang eventually found his way to Phillips (the Dutch healthcare, lighting, and consumer products conglomerate) and convinced them to put up 28% of the startup capital. The initial capital for TSMC’s fab was $220 million, with the following split:
$110 million (50%) from the Taiwanese government
$61.6 million (28%) from Phillips
$48.4 million (22%) from local businesses
Fun fact: Chang got no equity during the initial startup phase. The only reason he is worth $3 billion today is because he purchased all of his own shares in TSMC
In 1987, TSMC’s first fab officially opened and the TSMC flywheel began spinning.
History – The TSMC Flywheel
The market that TSMC joined was not stable, because most companies still operated and manufactured in their own fabs. This meant that they would only use TSMC when they reached capacity, but once capacity leveled again, they did not use TSMC anymore.
As the 1990s came around, the markets began to favor TSMC. IC designers were leaving their jobs to create fabless semiconductor design companies, just as Chang had predicted, without a place to see other designs come to life. Nvidia and Qualcomm, which have market caps of approximately $1 trillion and $120 billion, respectively, were both founded during this time.
This marked the start of the TSMC flywheel. What this flywheel business model did was simple, but it was crucial in helping grow TSMC from a small startup in 1987 to a $5.2 billion revenue-generating business at the turn of the century. The flywheel works like this:
A fabless design company places a manufacturing order with TSMC
TSMC uses this revenue to advance its manufacturing efficiency, thus producing higher-quality chips
Higher quality chips generate more revenue for the fabless design company, which then invests in more research and development, and places larger orders with TSMC
This cycle continuously repeated itself throughout the 1990s. It generated the cash flow capable of sustaining the billions in capex spending that were required to scale TSMC.
TSMC’s revenue grew from $204 million in 1992 to $5.3 billion in 2000 (2,456% growth)
Net PP&E grew 3,047% to $7.7 billion over the course of the 1990s
Capex spending went from $83 million in 1992 to $3.3 billion in 2000, with TSMC still generating positive cash flow
History – The Golden Opportunity
In 2005, after years of impressive growth metrics, Chang decided to step down as CEO and relinquished the position to Rick Tsai. From 2005 to 2009, the company managed to maintain its pre-crisis profitability. Diluted EPS remained stable at $0.11 and TSMC’s period of record-setting growth seemed to have ended.
Then, at 74 years old, Chang returned as CEO of TSMC, claiming that he found “the golden opportunity.” This opportunity was the smartphone.
Apple had just released the iPhone 3GS in June 2009
Apple’s iPhones captured 14.4% of the market share for mobile devices
Apple shipped 25 million smartphones throughout the course of 2009
Chang recognized the growing popularity of smartphones and the trend towards using Advanced RISC Machine (ARM) processors.
ARM processors are significantly faster than the Samsung processors that were being used at the time
The companies designing ARM processors were Qualcomm, Broadcom, MediaTek, and Apple. All of these companies were fabless, a ripe opportunity for TSMC
Throughout 2009 and 2010, these companies approached Chang and began utilizing TSMC’s fabs to create their ARM processors. The use of ARM revolutionized the smartphone industry because it paired the power of a personal computer (PC) and the size of a mobile phone. Demand for the iPhone grew exponentially and subsequently so did the demand for TSMC-produced ARM processors. Chang’s flywheel in action.
A $9 billion deal between Apple and TSMC was struck in 2014 for ARM processors
TSMC's total revenue grew from $9 billion in 2009 to $26 billion in 2015
Diluted EPS climbed from $0.11 in 2009 to $0.35 in 2015
This was one of the most profitable periods in TSMC history due to the sheer quantity of ARM processors it was manufacturing
Intel and Samsung weren’t producing ARM processors at the time, so TSMC captured all of the market share
Chang retired from his role as chairman in 2018 at 86 years old, following the immense success that he brought to TSMC by establishing them as Apple’s primary chip producer. Since then, Mark Liu (Chairman) and C. C. Wei (CEO) have continued to align TSMC with Apple’s interests, continuing to ride the wave that this “golden opportunity” brought TSMC.
Contemplating the Future – Opportunities
TSMC is rolling out their newest chips with 3 nm architecture and demand from the consumer technology industry is growing fast.
Apple recently purchased TSMC’s entire capacity of 3 nm chips, which will be used in their A17 Bionic processor (the chip in the iPhone 15 Pro and iPhone 15 Pro Max)
The deal was valued at around $17 billion
3 nm technology delivers a 35% power efficiency improvement and is 15% faster than 4 nm
Samsung and TSMC are the leaders in 3 nm technology
TSMC has also broken ground in the US by starting construction of a fab in Arizona, bolstering its presence in the US market.
The investment for the Arizona fab was estimated to be around $12 billion
Construction began in 2021 and is expected to be completed in 2024 or 2025
The Arizona fab will use 5 nm technology. It is not TSMC’s latest technology, but it will provide 20,000-25,000 chips per month at full capacity
Geopolitically, TSMC is strengthening its relationships with the US government and key customers, including Apple, who was responsible for 23% of TSMC’s 2022 revenue of $72 billion
Expansion of manufacturing footprint globally builds supply chain resistance and reduces dependence on Taiwanese fabs
TSMC has plans to continue expansion into the US, which will offer more opportunities to provide semiconductor chips for various American industries that remain untapped. However, the Arizona fab has already been delayed one year, citing a lack of available labor and expertise.
The high-performance computing (HPC) market has been growing because of new artificial intelligence applications, Internet of Things devices, cloud computing, robotics, digital healthcare, and autonomous cars.
TSMC generated 41% of its total revenue in Q1 2022 from HPC markets
The global HPC market is estimated to grow at a compound annual growth rate (CAGR) of 6.1% through 2025
Additionally, our modern world’s reliance on TSMC cannot be overstated. 90% of the most advanced chips (those used in phones, computers, and data centers) can only be produced by TSMC. If TSMC stopped producing, the impact on manufacturing and technology would be more significant than that of the Great Depression. We would no longer be able to manufacture the technology that makes our lives efficient (phones, computers, cars) and comfortable (white goods, medical equipment).
Contemplating the Future – Challenges
The most concerning issue for TSMC over the next decade is reaching the maximum amount of transistors that can be placed on one chip. There are only so many transistors you can fit on a chip, so once that limit is reached, it is unsure whether TSMC’s growth will flatline, or if another innovative chip production method will be discovered.
Moore’s Law is a principle that states that the number of transistors on a chip doubles roughly every two years, named after Gordon Moore, the co-founder of Fairchild Semiconductor and Intel who made this observation in 1965. Consensus is fragmented now; the CEO of Nvidia claimed that Moore’s Law was “dead,” but the current Intel CEO and Head of Marketing for TSMC have the opposite view.
Today, the maximum transistor density for a chip is 300 million transistors per square mm, which is found on TSMC’s 3 nm chip
TSMC is currently working on a 1 nm process which could come to fruition by 2030
Once the 1 nm chip is rolled out, it will become very challenging to come up with processes capable of increasing transistor density without destroying the chip
TSMC has been battling with higher costs associated with the manufacturing process in recent years. They are transferring these costs to their consumers, which is pushing market share towards Samsung and Intel.
TSMC increased prices for their 7 nm and 5 nm process technologies by 10%, while the 16 nm process technology price grew by 20% during 2022
TSMC has stated that they plan to increase prices for a majority of their fabrication processes by 6% starting in January 2023
From a geopolitical standpoint, TSMC is directly in the crosshairs. Being located in Taiwan and having the global influence that they do, they are at risk of being pressured by governments into conducting business outside of the manner in which they choose. TSMC has made it evidently clear that they side more with the US but being so close to mainland China (and the creeping presence of China in Taiwan) forces C. C. Wei to consider playing both sides of the political spectrum as a safety precaution.
TSMC has a fab in China and (soon will have one in) the US
Semiconductor Manufacturing International Corporation (SMIC), China’s in-house semiconductor manufacturer, has poached nearly 200 TSMC employees
TSMC has filed countless patent infringement lawsuits against the SMIC over the past two decades
TSMC will have to exercise caution during this period of high geopolitical tension. The US government has been both an enabler and inhibitor to TSMC’s success, including forbidding the company from manufacturing for Huawei in 2020 (TSMC’s second-largest account at the time) and providing TSMC a slice of the $52 billion in subsidies earmarked for research and development under the 2022 CHIPS and Sciences Act. CHIPS and Sciences also forbids the export of advanced chips used for AI applications to China. Biden has this leverage, because the AI chip supply chain is dominated by Western countries and Japan. Fabless semiconductor companies (Nvidia, Qualcomm, and Intel) need the US on board to actually source the components for the chips they design and sell and TSMC manufactures.
Aligning its interests solely with the United States eliminates its ability to generate revenue from China (10% of total revenue for FY22). In fact, China has spent more importing chips than it has importing oil since 2012. When the Taiwanese government orchestrated its fab-enabled military protection strategy, it was undisputable that the US military would dominate in any conflict. This question no longer holds the same clarity. The key dilemma TSMC must face now is whether the next decade will be focused on leveraging their unprecedented levels of capital expenditure to capture market share, or if it will be on resisting the mounting pressures of geopolitical tensions.
Our primary sources for this summary include Ben Gilbert and David Rosenthal’s Acquired podcast and Joe Weisenthal and Tracy Alloway’s Odd Lots Bloomberg podcast. If you want to learn more about TSMC, I recommend you check those out!
Interested in our reading list? Check it out here.
Looking to partner with PP? Complete the form here.
Do you have any feedback? DM us on Instagram, we will get back to you!