Lesson 6 — Taiwan's Invisible Power
How Do Chips Actually Work?
Learning Material
1 pagesLesson 6 — Taiwan's Invisible Power
Understanding the Complex: How Do Chips Actually Work?
In 1985, the Taiwanese government asked Morris Chang, then a 54-year-old veteran of Texas Instruments and General Instruments, to come home and help build a semiconductor industry. Chang had lived in the United States for 37 years. He had risen to vice president at TI and was, by any measure, a successful American executive. He also had an idea that most people thought was either naive or absurd.
The chip industry at the time was organized around integrated device manufacturers (IDMs): companies like Intel, TI, and NEC that designed their own chips and made them in their own factories. The factory — the "fab" — was considered an essential competitive advantage. Giving up control of manufacturing meant giving up the ability to optimize the design-manufacturing interface, to protect proprietary processes, to ensure quality.
Chang's idea was to build a company that did only the manufacturing. No chip design of its own, no branded products — just a neutral, world-class manufacturer that would make chips designed by others, faster and cheaper than they could do it themselves. A "pure-play foundry," in the industry jargon.
He founded TSMC — Taiwan Semiconductor Manufacturing Company — in 1987. Thirty-five years later, it is the most important manufacturing company in the world.
Why the foundry model won
The foundry model turned out to be a profound structural innovation, not just for TSMC but for the entire industry.
By outsourcing manufacturing, chip designers could focus entirely on design — and over time, design became the primary value driver. Companies like Nvidia, Qualcomm, AMD, and Apple became "fabless": they employ no factory workers, own no fabs, and never touch a silicon wafer. Their value is purely intellectual — chip designs that TSMC then renders in silicon.
For TSMC, the arrangement was equally powerful. Because it served many different customers rather than competing with them, it had no incentive to reverse-engineer anyone's designs or show one customer's proprietary information to another. It became, in effect, the Switzerland of semiconductors — a neutral party trusted by rivals who would never allow each other into their facilities.
And as TSMC grew, it achieved something that fabless companies and IDMs couldn't: scale. By concentrating volume production at a single site, it could spread the cost of each new generation of manufacturing equipment — which now costs hundreds of millions per machine — across enormous output. The economics became self-reinforcing. More customers meant more revenue. More revenue funded better equipment. Better equipment attracted more customers.
The supply chain that makes it all possible
A chip that powers an iPhone passes through at least three continents before reaching a consumer. The chain runs roughly like this:
- Raw silicon dioxide is mined and processed, often in various locations globally
- Ultra-pure silicon wafers are manufactured, with Japan being a major supplier (Shin-Etsu Chemical, Sumco)
- ASML machines, built in the Netherlands, are shipped to TSMC's fabs in Hsinchu and Tainan, Taiwan
- Chemical precursors — photoresists, etchants, deposition gases — come from hundreds of specialized companies worldwide, many in Germany, Japan, and the US
- TSMC processes the wafers through hundreds of steps over several weeks
- Completed wafers are cut into chips, tested, and assembled into packages — often in Malaysia, Philippines, or China
- Packaged chips are shipped to contract manufacturers (Foxconn, Pegatron) in China, where they're assembled into finished devices
- Devices are distributed globally
This chain is astonishing for its complexity, its precision, and its brittleness. Every link depends on specialized knowledge and equipment that has taken decades to develop and cannot be quickly replicated.
The 2021 chip shortage: a stress test
When the COVID-19 pandemic disrupted demand forecasts in early 2020, automakers cancelled chip orders, expecting sales to drop. When demand recovered faster than expected, they went back to reorder — and found that TSMC and other fabs had already allocated their capacity to consumer electronics, which had surged as people worked and schooled from home.
Fabs cannot quickly "add capacity." Building a new fab takes three to five years and costs $10–20 billion. TSMC had capacity, but it was committed to other customers. The automotive industry, which had optimized its supply chain for "just in time" delivery with minimal buffer stocks, had no cushion.
Ford lost an estimated 2.5billioninprofitin2021fromtheshortage.GMlost2 billion. The automobile industry collectively lost hundreds of billions in revenue — because the supply of cheap microcontrollers (chips costing less than a dollar each) couldn't keep up.
The shortage exposed what strategists had been warning about for years: the global economy runs on a single-point-of-failure supply chain for its most critical industrial input.
The geopolitics: from trade war to chip war
Taiwan's central role in chip production intersects with one of the defining geopolitical tensions of our era: the relationship between the United States and China, and the unresolved question of Taiwan's political status.
The United States responded to China's advances in semiconductor production — and to its military applications of advanced chips — with escalating export controls. In October 2022, the Biden administration imposed sweeping restrictions on chip sales to China, limiting not just the most advanced chips but also the equipment and US-person services needed to make them. In 2023, the Netherlands restricted ASML from shipping older lithography machines to China as well.
China has responded by massively increasing its domestic semiconductor investment. The country has spent hundreds of billions of dollars attempting to build a self-sufficient chip industry. As of 2026, China's leading chipmaker, SMIC, is producing chips at roughly 7nm — several generations behind the 2-3nm frontier — using equipment it has sourced domestically or from non-restricted suppliers.
The gap illustrates the depth of the technological moat that TSMC and the broader Taiwan-centric supply chain represents. Manufacturing at the frontier requires not just investment but accumulated expertise, supply chains, and institutional knowledge that take decades to build.
There are no simple conclusions here. Whether Taiwan's current status is stable, whether export controls are effective policy, and how the geopolitical situation should be managed involves genuine disagreements among experts, governments, and citizens — disagreements that reflect differing values about national sovereignty, free trade, and the management of technological competition.
Next lesson: Who Does What? Why? Who Pays? — the ecosystem of companies, governments, and research institutions that make the semiconductor industry run.
Reading time: approx. 10–11 minutes