A Deep Dive into TSMC: Part I

Estimated Reading Time: 10 minutes

TL;DR

TSMC is one of the world’s most iconic companies. Founded by Dr. Morris Chang in 1987, TSMC has ascended to become the world’s largest and most influential semiconductor company. This week’s subscriber post is the first part of our two issue deep dive into TSMC. We will discuss the history of TSMC (and its founder) over the last several decades and will end with a tour of the current capabilities of the company. Our goal in this week’s issue will be to answer the following question:

How has TSMC achieved global dominance of the semiconductor foundry market?

As a short answer, TSMC invented a unique business model, had the benefit of good timing and luck, invested heavily and consistently in R&D, moved aggressively to protect itself in courts, and worked hard to secure critical contracts from powerful players like Apple. The combination of these factors has created powerful structural advantages that are unmatched elsewhere in the world.

A Brief History of TSMC

A history of TSMC has to start with a history of its founder. Morris Chang (Zhang Zhangmou) was born in Ningbo on the Chinese mainland in 1931. Chang’s family was middle class, and he grew up with a backdrop of “war, poverty, injustice” (source.) His family had to flee air raids in Guangzhou and lived through Japanese occupation of Hong Kong. Chang moved to the US in 1949 to attend college at Harvard, where he was the only Chinese student among the school’s roughly thousand freshmen. “My reaction entering Harvard was sheer ecstasy, almost disbelief,” Chang said. “What a country! The United States was at its peak in its moral leadership, in its political leadership in terms of democracy—and it was the richest country in the world” (source). As an aside, our goal today as Americans should be, I think, to recapture this global moral and political leadership. It’s a tall order, but not an impossible one!

Chang felt that as a Chinese-American, he would need to enter a technical discipline in order to find success. Chang decided to transfer to MIT as a sophomore in order to further his technical training, and graduated with his Bachelor’s degree in 1952. He obtained a Master’s degree in 1953 (also from MIT) and started his PhD afterwards, but failed his qualifying exam and had to leave without obtaining his degree to join Sylvania semiconductors. While at Sylvania, he learned about Germanium transistors and studied the foundations of device physics during his evening hours. After 3 years, he felt Sylvania was slow moving and not innovative and decided to jump ship to join Texas Instruments, which had become a rising star in the semiconductor industry. He rose quickly up the ranks and by 1961 had become manager of an engineering division. Impressed by his performance, Texas Instruments management decided to sponsor Chang to get a PhD. He would be paid a full salary by the company and would also have his tuition covered. Chang went to Stanford for his PhD. This time, he aced his qualifying exams and graduated in the spring of 1964.

Chang returned to Texas Instruments and steadily rose up in the ranks, becoming general manager of the integrated circuit division by 1967. Part of the reason for his success was his mastery of the technical fundamentals of device physics. He dramatically improved the yields of Texas Instrument’s manufacturing process from 2-3% to 25-30% by reasoning from first principles about the right temperature/pressure/etc needed to achieve good yields. Chang’s successes caught the eye of senior management. Chang was promoted to run Texas Instrument’s full semiconductor division for 6 years, and then later to run its consumer division. Chang unfortunately ran into headwinds at the consumer division and ended up being put out to pasture in a side role as a senior vice president. He agonized over the decision, but decided to depart Texas Instruments since he sensed no prospects for further advancement (source).

Chang spent a year at the [General Instrument Corporation] as COO. He discovered that General Instrument was less focused on R&D and more on mergers and acquisitions. He also separated from his then-wife Christine (source). In 1985, Chang was invited by Premier Sun Yun-Suan to become president of the industrial technology research institute (ITRI) in Taiwan. Drawn by the idea of running his own institution (and perhaps spurred by a midlife crisis after separating from his wife), Chang accepted despite his lack of familiarity with Taiwan. Chang’s mandate was to transfer research results into economic benefits for the Taiwanese industry. Chang instituted performance reviews at ITRI and pushed for more industrial contracts, which drew considerable pushback from his Taiwanese employees at the time.

Minister K.T Li pushed Chang to grow the Taiwanese semiconductor industry and asked him to put together a business plan for a Taiwanese semiconductor company (source). Chang thought hard and came up with the idea of founding a pure-play foundry, TSMC. It’s worth noting that at the time, Chang was 56 years old! Chang landed on the idea of a pure play foundry as a tactic to handle Taiwan’s weak IP position; the majority of IP disputes centered around circuit designs, so focusing on process innovations allowed TSMC to operate in a clear patent space. TSMC was founded with starting capital from the Taiwanese government and from Philips Electronics. Interestingly, Chang himself received no stock for founding the company! (Source.)

Chang's decision to build a pure-play foundry faced a lot of skepticism, with industry opinion that a pure play foundry would go nowhere interesting. The first few years of TSMC were indeed challenging with a struggle to get orders. TSMC’s early executives were hired from American universities and companies and helped build out TSMC's core talent pool. American semiconductor companies were locked in a tight competition with Japanese firms and routed business to TSMC as a way to gain an edge. Intel gave the first large order to TSMC and passed along technical information about 200 processes to help bootstrap TSMC manufacturing (source). The eventual rise of the fabless semiconductor industry gave TSMC a much needed lift. As the comic below shows, TSMC developed a mutually beneficial relationship with fabless companies such as Nvidia that helped both sides achieve considerable financial success.

A comic gifted to Morris Chang by Nvidia’s Jensen Huang commemorating the crucial role Morris Chang and TSMC played in Nvidia’s success. (Source.)

In its early years much of TSMC’s technology IP was licensed from IBM, so TSMC was viewed as a second rate technical company by Silicon valley. Chang however aimed for TSMC to become a first rate technology player. TSMC was able to achieve this goal in the early 2000s due to its far-sighted investments in copper interconnects and immersion lithography. TSMC’s copper interconnect technology was developed in house in 2003 and was productionized before IBM’s competing technology had even exited the lab. In the wake of its failure, IBM ended up exiting the foundry business entirely. TSMC also seized a crucial advantage by developing immersion lithography in partnership with ASML and beat out Japanese players Canon and Nikon which had bet on dry lithography. The combination of these two technical successes established TSMC as a premier technological player in the global industry.

By 2004, TSMC had captured half the world’s foundry business and was in a strong position. Morris Chang decided to step back and retired in 2005 to spend more time with his family. Samsung however seized the opportunity to intensify its competition with TSMC, and used its cell phone manufacturing and other product lines to drive business to its semiconductor foundry division. In 2009 after the financial crisis, profits fell dramatically and TSMC was forced to lay off employees. TSMC’s new CEO Cai Lixing instituted performance cuts for poorly performing employees. These cuts were poorly executed and built resentment among TSMC employees. In tandem, yield rates from factories failed to improve, and customers started to cancel orders. Morris Chang decided to return to TSMC and removed Cai Lixing. Chang cancelled all previous layoffs and dramatically boosted the company’s R&D budget to $1 billion. Chang brought back retired technical leader Chiang Shang-Yi, who’d led research on copper interconnect technology, to lead the transition to the 28 nanometer process node. TSMC successfully executed on the transition to 28 nanometers and achieved higher yields than Samsung, giving it a crucial edge in the competition (source).

In 2010, TSMC landed Apple as a key customer, securing an agreement from Apple to order an entire generation of chips from TSMC. Apple was in tight competition with Samsung in the smartphone market but was simultaneously forced to rely on Samsung as a supplier. Apple was eager to remove an awkward dependence on Samsung, especially since Apple and Samsung were embroiled in a messy patent dispute. Apple pushed TSMC to help it design new designs independent of Samsung IP to meet manufacturing needs. TSMC rose to the challenge by sending its best engineers to Cupertino to work closely with Apple engineers. TSMC also dramatically scaled up its manufacturing capabilities at this time to meet Apple’s demands. In 2014 Apple awarded its entire A8 chip series manufacturing contract to TSMC and sent TSMC’s share prices soaring.

Samsung struck back by hiring key technical talent away from TSMC (Liang Mong-Song in particular, Chiang Shang-Yi’s student) and beat TSMC to the first 14 nanometer FinFET node. Samsung’s technical successes lured away Apple which decided to go with Samsung Foundries for its A9 chip series. Qualcomm also followed suit. TSMC started a 24 hour around the clock R&D effort, with 3 shifts of workers, to catch up to Samsung. TSMC in parallel took Samsung to court for patent violations, alleging that Liang Mong-Song had stolen crucial IP from TSMC. Tied up in lawsuits, Samsung flubbed the A9 order, and TSMC surged back to the lead for Apple's A10 chip series in 2016. TSMC internally credits its extreme research pushes for dethroning both Samsung and Intel for world semiconductor leadership (source).

Reassured by these successes, Morris Chang retired for the second time in June 2018.

Sources

[1] https://www.semi.org/en/Oral-History-Interview-Morris-Chang
[2] https://spectrum.ieee.org/at-work/tech-careers/morris-chang-foundry-father
[3] https://docs.google.com/document/d/1RMr9lzlgrReruoosnoHLIZ0hHbqJT5Kq7hZmcG6_uqg/mobilebasic

TSMC Today

Let’s take a look at some of TSMC’s current financials. TSMC has an impressive market cap of about $600 billion and achieved annual revenue of about $38 billion USD in 2019. The figure below provides a snapshot of TSMC’s financial numbers over the last few years. Note the steadily increasing revenue and income, but also the tandem increases in R&D expenditures to match. TSMC’s discipline in doubling down in R&D investment has been a core driver of its success over the last several years.

TSMC revenues, net income, market capitalization, assets/stockholder’s equity, capital expenditures, and R&D expenditures. Numbers are in Taiwanese dollars with a conversion rate of roughly $28NTD = $1 USD. TSMC has scaled both revenue and expenditure across the board (source).

TSMC’s steady capital expenditures have built the world’s most advanced semiconductor manufacturing foundries. As the figure below shows, TSMC maintains a constellation of cutting edge foundries spread across Taiwan with the ability to manufacture 12 million 12-inch wafers (as of 2019). These foundries however place a heavy environmental toll on Taiwan, with large water and power needs. These pressures, alongside geopolitical pressure from the US, will likely push Taiwan to broaden its manufacturing base to other countries in the years to come.

TSMC has the world’s most sophisticated collection of semiconductor foundries with a capacity to manufacture 12 million 12-inch wafers annually in 2019 (source).

TSMC has a wide range of product lines. Its logic portfolio (for CPUs) typically monopolizes the most advanced process nodes, but TSMC also maintains product lines across other chip types by using older foundry process nodes as the diagram below shows.

TSMC maintains a broad range of different foundry offerings for different chip types. The most cutting edge foundries are usually reserved for processors (think Apple M1 on the TSMC 5nm node for example). Older foundries are often repurposed for other chip types (source).

Discussion

Despite its dominant position in the semiconductor industry, TSMC faces formidable challenges. On the business end, TSMC faces major poaching efforts from SMIC which has lured top talent away from TSMC, hiring Liang Mong-Song as co-CEO in particular (source). American pressure on SMIC has driven the company to court talent from TSMC aggressively as the CCP attempts to break free of America’s restrictions. Taiwan faces a challenging balancing act as it attempts to walk a tightrope between American and CCP interests.

Our newsletter of course is written firmly from the American camp. What can we do to better our semiconductor manufacturing position? Continuing to place pressure on Chinese manufacturing efforts while shoring up American semiconductor manufacturing capabilities is probably a good bet. The US government should offer both tax incentives and diplomatic pressure on TSMC to shift more manufacturing capacity to the US from Taiwan. The construction of new domestic foundries will help the US ecosystem build up semiconductor know-how and talent. Offering expedited visas to Taiwanese engineers (paired with high salaries), could also help the US close its semiconductor talent gap.

American companies should study TSMC more broadly as an example of exemplary industrial achievement. Morris Chang is an inspiring entrepreneur who we should count in the same ranks as Bill Gates or Elon Musk. TSMC has applied hard work, business savvy, and scientific genius to achieve its dominant position. Future American foundries should learn from TSMC’s example, much as Chang learned from Texas Instruments.

In next week’s subscriber post, we will continue our deep dive into TSMC from the technical side. In particular, we will seek to understand TSMC’s roadmap for the future of the semiconductor industry.

Feedback and Comments

Thank you for reading our subscriber-only newsletter! We’re still figuring out the rhythm for these posts, so if you have feedback on changes you’d like to see, please send them over to bharath@deepforestsci.com! If you’d like to see more financial analysis, or more technical analysis, or deeper dives into a particular industry let me know and I’ll see what we can do.

About

Deep Into the Forest is a newsletter by Deep Forest Sciences, Inc. We’re a deep tech R&D company specializing in the use of AI for deep tech development. We do technical consulting and joint development partnerships with deep tech firms. Get in touch with us at partnerships@deepforestsci.com! We’re always welcome to new ideas!

Credits

Author: Bharath Ramsundar, Ph.D.
Editor: Sandya Subramanian