AI / LLMChina
← Sticker Wall
JOURNEY · DETAIL
TSMC: Founded by Morris Chang in 1987, pioneered the pure-play foundry model, and became the king of global AI computing chips through partnerships with Apple and NVIDIA.
Founded: Morris Chang · Taiwan Semiconductor Manufacturing Company, Limited
JOURNEY
Key Fields
FIELD STAMPSIndustryAI / LLM
RegionChina
ScaleGiant
ChannelOther
Origin
Morris Chang entered the semiconductor industry in the 1950s, holding positions at Texas Instruments and General Instrument. At age 55, he was recruited back to Taiwan to serve as president of the Industrial Technology Research Institute (ITRI). He predicted that semiconductor design and manufacturing would inevitably decouple. Recognizing that Taiwan lacked brands and markets but could focus on manufacturing, he founded TSMC in the Hsinchu Science Park in 1987. He pioneered the pure-play foundry business model, addressing the manufacturing needs of fabless semiconductor companies worldwide.
Milestones
1985
Founding Period Turning Point
In 1985, Morris Chang returned to Taiwan from the U.S. to lead ITRI. In 1986, TSMC was planned as a joint venture between the Executive Yuan's Development Fund and Philips. The company was officially established in 1987 with a capital of approximately NT$1.38 billion. At the time, the semiconductor industry was dominated by the IDM model, and TSMC's pure-play foundry approach was met with skepticism. Leveraging his 26 years of experience at Texas Instruments and his vision for industry specialization, Chang persisted. The company started with technology transfers from ITRI and licensing from Philips, spanning from 1985 to 1987.
1994
IPO and Expansion Growth
TSMC went public on the Taiwan Stock Exchange in 1994, followed by an ADR listing on the New York Stock Exchange in 1997, becoming the first Taiwanese semiconductor company to list in the U.S. With the explosion of PC and communication chip demand in the 1990s and the rise of fabless companies like Qualcomm, Broadcom, and NVIDIA, TSMC's foundry orders grew rapidly. Revenue exceeded NT$70 billion in 1997, cementing its position as the global leader in professional foundry services, spanning from 1994 to 1997.
1999
Technology Catch-up Failure
In 1999, TSMC secured a major order for NVIDIA's graphics chips, but the dot-com bubble burst in 2000 led to a sharp downturn in the semiconductor industry, causing TSMC's revenue to decline for the first time in 2001. In 2003, while competing for the 0.13-micron copper process, TSMC lost the lead to the IBM and UMC alliance, a rare instance where Morris Chang admitted a misjudgment in technology strategy. In 2005, Chang handed over the reins to Rick Tsai. The 2008 global financial crisis caused capacity utilization to plummet, and Tsai's layoffs triggered internal backlash, plunging the company into a period of painful transformation, spanning from 1999 to 2008.
2009
Chang's Return and 28nm Inflection Point
Following the financial crisis, Morris Chang returned as CEO in 2009, decisively increasing capital expenditure to $5.9 billion against the trend and focusing resources on the 28nm process. In 2011, the 28nm process entered mass production, and TSMC secured major orders from Qualcomm and Apple due to superior yields compared to competitors. In 2014, Apple's iPhone 6 featured the A8 chip manufactured by TSMC. That same year, revenue exceeded NT$762.8 billion, officially surpassing UMC to become the absolute leader in global foundry, spanning from 2009 to 2016.
2016
Advanced Process Arms Race PMF
In 2016, TSMC became the exclusive foundry for Apple's A10 chip, with the 10nm process entering mass production in 2017. The 7nm process launched in 2018, securing the first major GPU foundry order from NVIDIA. However, after Chang retired in 2018, handing over to Mark Liu and C.C. Wei, the company faced a photoresist contamination incident in 2019, resulting in the scrapping of tens of thousands of wafers and a loss of approximately $550 million. Through rapid process improvements and client trust, 2019 annual revenue exceeded NT$1.06 trillion, making it one of the world's most valuable semiconductor companies, spanning from 2016 to 2019.
2020
Geopolitics and 3nm Turning Point
In 2020, under U.S. pressure, TSMC was pushed to build a plant in Arizona, forcing the company to initiate overseas expansion. The 3nm process was launched in 2021, and in December 2022, the company held a tool-in ceremony in Arizona, announcing an increased investment of $40 billion. 2023 revenue reached NT$2.16 trillion. In 2024, AI chip demand exploded, with NVIDIA's H100 and subsequent chips entirely outsourced to TSMC. Annual revenue reached NT$2.89 trillion, and market cap briefly exceeded $1 trillion, spanning from 2020 to 2024.
2025
Consolidation of AI Hegemony Growth
In 2025, at age 94, Morris Chang gave a 20,000-word in-depth interview, revealing how he secured key orders from Apple and NVIDIA. In 2026, TSMC raised its full-year revenue growth forecast to 40% and increased capital expenditure to $64 billion, with global expansion of 3nm capacity. Chairman C.C. Wei stated that AI demand is very strong and that while the company does not seek short-term price hikes, it aims to achieve high profit margins similar to memory manufacturers. The company has become the core supplier for global AI computing infrastructure, spanning from 2025 to 2026.
Turning Points
- In 2009, Morris Chang returned as CEO, counter-cyclically increasing capital expenditure from ~$3 billion to $5.9 billion, successfully betting on the 28nm process and creating an insurmountable gap with UMC and GlobalFoundries.
- In 2014, winning back Apple's A8 processor orders from Samsung proved that advanced process yield is more important than low pricing.
- In 2018, being the first to mass-produce the 7nm process led NVIDIA to shift all AI training GPU orders to TSMC, initiating its AI foundry hegemony.
- In 2022, the start of construction on the Arizona plant forced TSMC to transition from a local Taiwanese company to critical global geopolitical infrastructure.
- In 2024, the AI chip demand explosion made NVIDIA, AMD, Apple, and Qualcomm entirely dependent on TSMC, leading to a supply shortage in advanced process capacity.
Failures & Pitfalls
- In 2003, the 0.13-micron process was beaten by the IBM and UMC alliance; Chang admitted a misjudgment in technology strategy, leading to the loss of high-end client orders.
- During the 2008 financial crisis, capacity utilization plummeted; the layoffs led by Rick Tsai caused a collapse in internal morale, forcing Morris Chang to return to save the company.
- The 2019 photoresist contamination incident resulted in the loss of tens of thousands of wafers and approximately $550 million, exposing vulnerabilities in single-supplier risk management.
- Post-2020, the cost of setting up plants in the U.S. has been far higher than in Taiwan; the Arizona plant has faced repeated setbacks in progress and cost control, leading to long-term questions regarding return on investment.
关键成功要素
- Pioneered the pure-play foundry model in 1987, decoupling design from manufacturing and enabling the fabless semiconductor industry.
- Counter-cyclical investment in 28nm during the 2009 financial crisis traded capital expenditure for technological leadership, building long-term barriers during an industry trough.
- Proved yield and advanced process capabilities with Apple's A8 orders in 2014, after which Apple consistently contributed over 20% of annual revenue.
- Exclusive foundry for NVIDIA's AI GPUs since 2018, becoming the only large-scale mass-production channel for AI computing chips.
- Maintained R&D spending at 8% to 10% of revenue for years, with 2026 capital expenditure reaching $64 billion, far exceeding competitors.
Lessons
- Business model innovation does not require inventing technology from scratch; perfecting a specific link in the supply chain can define an entire industry.
- Technological leadership requires counter-cyclical investment over several years; heavy bets during financial crises are more effective than expansion during peaks.
- Client concentration is both a risk and a lever; major orders from Apple and NVIDIA transformed TSMC's foundry business from a low-margin service into a high-barrier monopoly.
- Ignoring geopolitical risks leads to forced high-cost choices; profit margins for overseas plants are significantly lower than those in Taiwan since the 2022 U.S. expansion.
- Governance crises can emerge rapidly after a founder steps down; the 2008 layoff incident proved that succession must be supported by a comprehensive mechanism.
Core Data
- 2024 Full-Year Revenue：NT$2.89 trillion (Public data, not independently verified)
- 2026 Capital Expenditure：$64 billion (Public data, not independently verified)
- 2026 Revenue Growth Guidance：40% (Public data, not independently verified)
- 2023 Full-Year Revenue：NT$2.16 trillion (Public data, not independently verified)
- 2019 Photoresist Contamination Loss：$550 million (Public data, not independently verified)
- Market Capitalization：Exceeded $1 trillion in 2024 (Public data, not independently verified)
- 2025 Morris Chang's Age：94 (Public data)
Competitors / Peers
TSMC's biggest competitor is Samsung Electronics of South Korea, which continues to compete for orders using low prices and an IDM model, though its yields significantly lag behind. Intel has attempted to enter the foundry market, with CEO Pat Gelsinger pushing for a revival in 2026, but there remains a massive gap in process technology and client trust. UMC and GlobalFoundries retain some market share in mature processes above 28nm but have abandoned advanced processes below 7nm. SMIC primarily serves China's domestic substitution needs and cannot challenge high-end AI chip foundry in the short term. TSMC's true core advantage is the flywheel effect composed of yield, capacity, and client trust, which competitors find difficult to replicate within five years, even with heavy investment.