Summary of Dr. Morris Chang’s Autobiography (Volume 2)
First book review!
I finished reading the second volume of Dr. Morris Chang’s autobiography on Sunday. As the founder of TSMC, his writing is direct and emotional, making it both engaging and insightful. Since the book is available only in Chinese, I decided to provide a summary. Hopefully I don’t plagiarize too much of Dr. Chang’s work.
The Genesis of the Foundry Business Model
Challenges in Chip Manufacturing
Building fabs is expensive and time-consuming. Managing them requires skills vastly different from designing systems or chips.
Integrated Device Manufacturers (IDMs) like Intel or General Instruments often prioritize their own needs. They resist customer input on their processes, leading to substandard service and timeliness.
Awkward Dynamics in the Industry
Chip designers must share technical details with IDMs, essentially giving away trade secrets to direct competitors.
Early Lessons in Yield and Workforce Management
In 1981, Texas Instruments’ Miho memory chip fab in Japan achieved yields of 40–50%, double that of its Houston fab.
Key factors included Japan’s low employee turnover (2% vs. 10–25% in Houston) and the availability of technically educated shift leaders and equipment engineers. In Houston at that point they could no longer recuirt shift or line leaders with technical education, in Japan they could, sometimes with university engineering degrees. Japan also has more competent equipment engineers.
Equipment engineers are responsible for operating machines 24/7, they collect data and do some M&R. Even when they are off-duty they stay ready in case something happens. It is very costly to replicate that in the US.
“Copy Exactly!”
Intel’s Craig Barrett introduced this concept to replicate production lines with exact specifications, leaving no flexibility for chip designers' adjustments.
Dr. Chang’s Insights on Customization
From 1967 to 1970, Dr. Chang developed Medium Scale Integration (MSI) at TI, leveraging the company’s strengths. MSI are customized chips, margin is better than standard chips. He further pushed for better yield, quality, and reliabily to strength TI’s pricing power.
Taiwan’s Strengths in 1985
While Taiwan lagged in chip design, process technology and marketing, Dr. Chang recognized it could compete through quality and time-to-market.
Texas Instruments
Outdated Expertise and Strategic Missteps
TI’s Mark Shepherd(Chairman), an expert in the 1950s, but failed to adapt to the industry's rapid evolution. By the 1970s, his outdated approach led to critical strategic errors. In the 1950s, Texas Instruments (TI) built its own manufacturing equipment due to the sector's immaturity. However, in the 1970s, Shepherd maintained the same mechanical engineering team despite significant industry changes. During 1972-78 Dr. Chang repeatedly challenged Shepherd over this inflexibility. Another reason that their Japan fab had a higher yield is because photoresist production line was made by Japanese equipment manufacturer.
Consumer Electronics Misadventures
The initial attraction was clear: in the early 1970s, calculators sold for over $200, with component costs (integrated circuits, displays, keypads) around $20-$30. TI believed they could capitalize on this margin. They repeated this miscalculation with watches and home computers.
Consumer electronics and semiconductors operate on fundamentally different business models:
Consumer Electronics: Retailer-focused, emphasizing consumer needs and aesthetics
Semiconductors: System company-oriented, prioritizing quality and cost-performance
FThis strategy put TI in direct competition with its own customers. In some cases, like with Canon, TI was the sole supplier. J. Fred Bucy(President and CEO) attempted to prioritize the consumer electronics division's production capacity, but Dr. Chang resisted. The internal dynamics became increasingly dysfunctional. TI's IC division was now required to satisfy the consumer electronics division before serving external customers. In contrast, companies like Intel could focus solely on optimizing microprocessors.
Questionable Expansion
In the mid-1970s, Bucy built a foundry in Lubbock, Texas, driven more by personal nostalgia (his hometown) than strategic considerations. The location made it challenging to attract talent.
Notable Trivia:
By 1974, TI was already using internal email.
The company occasionally implemented "mass vacations" where the entire staff would take approximately two weeks off.
Business relationship with other companies
Phillips
In 1985, Phillips invested in TSMC when other major tech companies(Intel, TI, Motorola, Panasonic, Sony, AMD, Thomson, DEC) declined. Their decision was decided by two key factors:
1985 was a challenging year for semiconductors, with negative 16% growth. Phillips had other lines of business to soften the impact.
Phillips already had existing assembly lines in Taiwan and was looking to expand into lighting and passive component production.
Investment Outcome:
Initial stake: 27.5%
Full exit: August 2008
Total return: 134.6 times (26.34% Compound Annual Growth Rate)
TSMC continued to grow 19.1 times post-Phillips' exit (20.17% CAGR)
Intel
Intel placed significant orders during TSMC’s second year, helping establish production standards like statistical process control, quality control and preventive maintenance.
Nvidia
In 1997, Nvidia’s Jensen Huang reached out directly to Dr. Chang after facing neglect from TSMC’s San Jose office.
In 1998, TSMC sent managers to assist Nvidia with resource planning. In 2013, Dr. Chang even offered Huang the CEO position at TSMC. At that time TSMC's market cap was 10x of Nvidia.
Qualcomm
Qualcomm moved from IBM’s fabs to TSMC due to IBM’s arrogance, a problem TSMC avoided through its customer-first approach.
Apple
Apple in 2010 began exploring a partnership with TSMC.
Its relationship with Samsung was inherently complicated: Samsung simultaneously supplied critical components to Apple (displays, memory, chips) while competing directly with Apple through its own device lineup. Since 2007, Dr. Chang anticipated Apple would eventually partner with TSMC.
Apple first approached Intel, but Intel was unwilling to sacrifice margin and service was subpar.
Goldman’s MBO Proposal in 2007
Proposed by Hsueh-Jen Sung, who was also key to TSMC's Taiwan and Nasdaq IPOs.
However TSMC’s capital expenditure and R&D expenses are too high and it will have difficulties in financing as a private company. There is also a potential loss of control.
After one month of consideration, he rejected the proposal.