Why TSMC Became Indispensable

In 1986, Taiwan's Executive Yuan minister Li Kwoh-ting handed Morris Chang what amounted to a blank check. Build a semiconductor industry for Taiwan. Chang, then 55, had spent 25 years at Texas Instruments rising to senior vice president and had been passed over for the CEO job. He took the offer, became president of the Industrial Technology Research Institute in Hsinchu, and went looking for a partner.

Texas Instruments said no. Intel said no. Only Philips signed — putting up $58 million, transferring production technology, and licensing intellectual property in exchange for 27.6 percent of a company that didn't yet exist. The Taiwanese government's National Development Fund put up another 48.3 percent of the startup capital. The remaining 20 percent was raised from eight private companies — Formosa Plastics, CAPCO, Taiyuan Textile, others — owned by some of the island's wealthiest families, who were directly "asked" by the government to invest. The Kuomintang's investment arm contributed the final 4.1 percent. On 21 February 1987, Taiwan Semiconductor Manufacturing Company opened in Hsinchu Science Park. Its first CEO was James E. Dykes, sent over by Philips. After a year, Dykes left and Chang took the chief executive job himself.

What Chang did next reshaped the global electronics industry. Every existing chip company in 1987 — Intel, Motorola, Texas Instruments, NEC — designed and manufactured its own silicon. Fabs cost hundreds of millions to build and a few billion to keep at the leading edge, so only the largest integrated firms could play. Chang's bet was that if a foundry would manufacture chips for anyone, refusing to design or sell its own products, then engineers with good ideas wouldn't need a billion dollars and a fab to start a company. They'd just need a customer slot at TSMC. That model — the pure-play foundry — created the entire fabless industry. Without it, there is no Nvidia, no Qualcomm, no Broadcom, no MediaTek, no Marvell, no AMD as the company exists today, no Apple Silicon. They all design chips that TSMC builds.

The fabless customers eventually outgrew the integrated rivals. By 2017, TSMC's market capitalization passed Intel's for the first time, hitting $168.4 billion against Intel's $165.7 billion. By June 2020 TSMC briefly cracked the world's ten most valuable companies at $410 billion. In 2025, the company posted US$122.42 billion in revenue, US$62.22 billion in operating income, and US$55.13 billion in net income, with 83,825 employees as of 2024. It now controls roughly 70 percent of the global foundry market, and a far higher share — over 90 percent — of the most advanced logic nodes.

The technology lead is the lock-in. TSMC was first to volume-produce 7-nanometer chips, first to 5-nanometer with the 2020 Apple A14 and M1, the first foundry to commercialize ASML's extreme ultraviolet lithography in N7+, and began 3-nanometer volume production in late 2022 — yielding a 70 percent density improvement over 5nm. By late 2025 the company moved to its N2 node, switching from FinFET to gate-all-around nanosheet transistors, the first major architectural shift since 2011. Each new node now requires a single $370 million ASML EUV scanner per machine, and TSMC takes the bulk of ASML's output. Samsung and Intel have struggled to match the yield. Apple's A-series, M-series, and Nvidia's H100 and Blackwell AI accelerators all run exclusively on TSMC silicon.

Concentration created the chokepoint. Roughly 90 percent of TSMC's wafer capacity sits on a single island that the People's Republic of China claims as its territory and has refused to renounce force over. Taiwan's exports of integrated circuits hit $184 billion in 2022 — nearly 25 percent of the country's GDP — and TSMC alone accounts for about 30 percent of the Taiwan Stock Exchange's main index. After the 2020 chip shortage and rising tension over the Taiwan Strait, TSMC began an unprecedented overseas build-out: a $7 billion JASM joint venture with Sony in Kumamoto, Japan, with Denso joining for a 10 percent stake — Fab 23 began commercial production in December 2024 with 476 billion yen subsidized by Japan's METI. A €10+ billion European fab in Dresden was approved in August 2023 with €5 billion from the German government and Bosch, Infineon, and NXP each holding 10 percent of the resulting joint venture, ESMC.

The Arizona project is the largest. Phoenix Fab 21 was approved in November 2020 at $12 billion. By December 2022 TSMC tripled the planned investment, by April 2024 the Biden administration committed up to $6.6 billion in direct CHIPS Act funding plus up to $5 billion in loans, and in November 2024 the Commerce Department finalized the binding award — the first major CHIPS Act completion — covering $65 billion across three fabs. TSMC agreed to deploy its A16 manufacturing in Arizona and to forgo stock buybacks for five years under an "upside sharing agreement." Then in July 2025, C.C. Wei announced an additional $100 billion expansion toward a six-facility "gigafab" cluster.

But Arizona has been hard. Chang himself said the cost of building a fab in the United States runs four to five times what an identical Taiwan fab costs, pushing the cost of US-made chips at least 50 percent higher. Volume advanced production at the second Arizona fab won't begin until 2027 or 2028, and chips packaged in Arizona still get shipped back to Taiwan for advanced packaging. Morris Chang retired as chairman in 2018 after 31 years; Mark Liu held the chair until C.C. Wei took both CEO and chairman roles on 4 June 2024. The lock-in TSMC built over four decades cannot be replicated quickly anywhere else, which is exactly the problem Washington, Tokyo, and Berlin are now spending tens of billions of dollars trying to solve.