We have profiled Intel — the empire that bugs built. AMD — the underdog that bit back. NVIDIA — the leather jacket monopoly. Clevo — the laptop nobody admits.

Today we profile the company that manufactures the silicon for all of them.

TSMC.

Taiwan Semiconductor Manufacturing Company. The most important company on Earth that most people cannot name. TSMC does not design chips. TSMC does not sell chips with its name on them. TSMC does not compete with its customers. TSMC only does one thing: it takes other companies’ chip designs and manufactures them with a precision measured in atoms.

Apple’s A-series and M-series processors? TSMC manufactures them. NVIDIA’s H100, H200, and B100 AI accelerators? TSMC. AMD’s Ryzen and EPYC? TSMC. Qualcomm’s Snapdragon? TSMC. Every major fabless chip company on Earth depends on a single manufacturer on an island 180 kilometers from mainland China.

This is not a company profile. This is a geopolitical briefing.

Morris Chang: The Man Texas Instruments Did Not Promote

Morris Chang was born in Ningbo, China in 1931. His family fled to Hong Kong in 1949 ahead of the communist revolution. He moved to the United States, earned degrees from MIT and Stanford, and spent 25 years at Texas Instruments, rising to Senior Vice President — the highest-ranking Chinese-American executive in the American semiconductor industry.

Then TI passed him over for CEO. Twice.

In 1985, at age 54, Chang left Texas Instruments. He had spent a quarter century building semiconductor manufacturing expertise at one of the world’s premier chip companies, and they decided he was not CEO material. TI’s loss would become the most consequential personnel decision in semiconductor history.

Taiwan’s government, through the Industrial Technology Research Institute (ITRI), approached Chang to lead Taiwan’s semiconductor ambitions. They gave him $220 million in government funding. In 1987, at age 55 — an age when most executives are coasting toward retirement — Morris Chang founded TSMC.

He did not build another chip company. He invented a new kind of chip company.

The Pure-Play Foundry: The Invention That Changed Everything

Before TSMC, the semiconductor industry had one model: design your own chips, build your own fabs. If you wanted to sell processors, you needed billion-dollar fabrication plants. Intel had fabs. IBM had fabs. Texas Instruments had fabs. If you could not afford a fab, you could not compete.

Morris Chang’s insight was deceptively simple: separate design from manufacturing. TSMC would manufacture chips for anyone. It would never design its own chips. It would never compete with its customers. A pure-play foundry — a factory with no ambitions beyond being the best factory.

This model made the fabless semiconductor company possible. You could design a chip on a workstation, send the design to TSMC, and receive manufactured silicon back — without ever owning a factory, without ever investing $10 billion in a fabrication plant, without ever hiring a single process engineer.

The companies this enabled:

Company	What They Design	Who Makes It
Apple	A-series, M-series	TSMC
NVIDIA	GeForce, H100/H200/B100	TSMC
AMD	Ryzen, EPYC, Radeon	TSMC
Qualcomm	Snapdragon	TSMC
Broadcom	Networking, WiFi	TSMC
MediaTek	Mobile SoCs	TSMC

Every one of these companies exists in its current form because Morris Chang built a factory that does not compete with its customers. AMD went fabless in 2009 when it sold its fabs to create GlobalFoundries — and then moved its most advanced production to TSMC. NVIDIA has never owned a fab. Apple designs its own silicon but manufactures none of it.

The pure-play foundry model did not just create competitors to Intel. It created the entire modern chip industry. Without TSMC, there is no fabless revolution. Without the fabless revolution, Intel has no competition. Without competition, your laptop still runs a Pentium 4.

Process Node Leadership: Manufacturing at the Limits of Physics

TSMC does not just manufacture chips. TSMC manufactures chips that nobody else can.

Year	Node	Notable Product
2018	7nm	Apple A12, AMD Zen 2
2020	5nm	Apple A14, AMD Zen 4
2022	3nm	Apple A17 Pro
2025	2nm	Next-generation mobile/HPC

Each node shrinks the transistors, increases density, reduces power consumption, and improves performance. Moving from 7nm to 5nm to 3nm is not linear improvement — it is exponential difficulty. The features on a 3nm chip are approximately 15 atoms wide. You are manufacturing at a scale where quantum effects actively interfere with your engineering.

Intel — the company that once led the world in process technology — announced “five nodes in four years” under Pat Gelsinger. The plan was to catch TSMC by 2025. Gelsinger was given the choice to retire or be removed in December 2024. The board decided the plan was not working. TSMC remains 1-2 generations ahead of Intel in manufacturing capability.

Intel tried to close a decade-long gap in four years. Physics does not negotiate timelines.

The ASML Dependency: The Machine That Only One Company Makes

TSMC’s advanced nodes — 7nm and below — depend on Extreme Ultraviolet (EUV) lithography. EUV uses light with a wavelength of 13.5 nanometers to etch patterns onto silicon wafers. The light is generated by vaporizing tin droplets with a laser and bouncing the resulting plasma off multilayer mirrors in a near-perfect vacuum.

There is exactly one company in the known galaxy that manufactures EUV lithography machines: ASML, headquartered in Veldhoven, Netherlands. We have not heard from other galaxies on this matter. We assume they are also waiting on ASML delivery timelines.

Each EUV machine costs between $150 million and $380 million. The most advanced version — the High-NA EUV system — costs closer to $380 million. Each machine weighs over 150 tons and requires multiple Boeing 747 cargo planes to deliver.

ASML is itself a single point of failure in the global chip supply chain. TSMC cannot manufacture advanced chips without ASML machines. Nobody else makes EUV machines. If ASML’s factory in the Netherlands were destroyed, global semiconductor manufacturing at advanced nodes would stop. Not slow down. Stop.

The global chip supply chain: billions of devices depend on TSMC, which depends on ASML, which is one company in one country with one factory. Two single points of failure stacked on top of each other.

Customer Concentration: Everyone Depends on TSMC

Apple is TSMC’s largest customer, accounting for approximately 25% of revenue. Every iPhone, iPad, Mac, and Apple Watch runs on TSMC-manufactured silicon. Apple does not have a backup manufacturer for its most advanced chips. There is no backup. Nobody else can make them.

NVIDIA’s entire AI empire — the H100, H200, B100, every GPU training every large language model — is manufactured by TSMC. AMD’s Ryzen and EPYC processors are manufactured by TSMC. Qualcomm’s Snapdragon chips are manufactured by TSMC.

The four most important fabless chip companies on Earth — Apple, NVIDIA, AMD, Qualcomm — all depend on a single manufacturer, on a single island, in a single geopolitical flashpoint.

The Taiwan Strait: 180 Kilometers of Existential Risk

Taiwan is 180 kilometers from mainland China. China claims Taiwan as a breakaway province. The People’s Liberation Army conducts regular military exercises around Taiwan. The Taiwan Strait is the most dangerous body of water on Earth — not because of its currents, but because of what is manufactured on its eastern shore.

TSMC manufactures approximately 90% of the world’s most advanced semiconductors — chips at sub-5nm process nodes. If Taiwan is invaded, blockaded, or destabilized, the global chip supply does not merely tighten. It collapses.

No advanced chips means no new iPhones. No AI training hardware. No advanced military systems. No new cars (modern vehicles use hundreds of chips). No new servers. The global economy does not have a Plan B for losing TSMC.

Morris Chang has publicly acknowledged this reality. He called the geopolitical risk “very high.” He also said something more pointed: “Without semiconductors, you would have no modern weapons. Without TSMC, you would have a much harder time.”

He was not talking to investors. He was talking to governments.

The Silicon Shield

This brings us to the theory that defines Taiwan’s survival strategy.

The Silicon Shield is the idea that Taiwan’s semiconductor manufacturing capability functions as a military deterrent. China cannot invade Taiwan without destroying TSMC — and China cannot afford to destroy TSMC because China itself imports approximately $400 billion per year in semiconductors.

China depends on the very industry it would obliterate by invading. Taiwan’s chip fabs are its de facto military protection — not through force, but through economic interdependence. TSMC is Taiwan’s nuclear deterrent without the radiation.

A small island nation, 180 kilometers from a superpower that claims ownership, survives not through military strength but through indispensability. If you are the only factory that can make the thing everyone needs, nobody can afford to destroy you.

Kim Jong Rails understands this concept intimately. But we will return to that.

The Diversification: Arizona, Japan, Germany

The Silicon Shield works — but it also terrifies TSMC’s customers. If everything depends on Taiwan, and Taiwan is at risk, then the supply chain itself is at risk. Governments have responded with subsidies, and TSMC has responded with expansion.

TSMC Arizona (Fab 21):

The U.S. government, through the CHIPS Act, offered TSMC $6.6 billion in grants and $5 billion in loans to build fabs in Phoenix, Arizona. Fab 21 (4nm) became operational in 2024. Additional fabs for 3nm and 2nm are planned.

The construction has been difficult. TSMC’s manufacturing culture — precision-obsessed, hierarchy-driven, workers accustomed to 12-hour shifts and near-military discipline — clashed with American labor norms. There were not enough skilled semiconductor workers in Arizona. Training took longer than expected. Costs were significantly higher than Taiwan.

Morris Chang said it plainly: the Arizona costs are “significantly higher” than Taiwan. He went further, publicly stating that the United States “doesn’t have the management talent, worker spirit and talent, and supplier ecosystem” to match Taiwan’s manufacturing.

This is a man who lived in the United States for 30 years, worked at Texas Instruments for 25, and built a $900 billion company. When he says the U.S. cannot replicate Taiwan’s manufacturing ecosystem, he is not being diplomatic. He is being accurate.

TSMC Japan (Kumamoto):

TSMC opened a fab in Kumamoto, Japan in February 2024 — manufacturing at 28nm and 16nm. Japan provided a $3.5 billion subsidy. Sony and Toyota are co-investors. This fab targets automotive and industrial chips, not cutting-edge mobile processors.

Japan wants semiconductor self-sufficiency. TSMC wants geographic diversification. Sony wants guaranteed chip supply for image sensors. Toyota wants guaranteed chip supply for cars. Everyone gets something. The math works.

TSMC Germany (ESMC):

TSMC announced the European Semiconductor Manufacturing Company (ESMC) in Dresden, Germany — manufacturing 28nm and 22nm automotive chips. A EUR 10 billion investment, subsidized by the EU. Dresden was chosen because it already hosts GlobalFoundries, Infineon, and Bosch semiconductor operations — the infrastructure exists.

The pattern: TSMC is building fabs outside Taiwan for mature nodes (28nm, 16nm, 4nm), while keeping the most advanced manufacturing (3nm, 2nm) in Taiwan. The crown jewels stay on the island. The diversification is real but selective.

Intel Foundry Services: The Competitor That Is Not Competing

Intel is trying to become a foundry — Intel Foundry Services (IFS) — to compete with TSMC. This was Pat Gelsinger’s central bet before the board removed him.

As of 2026, IFS has few external customers and has not demonstrated competitive yields at advanced nodes. The challenge is existential: Intel is asking chip companies to trust a foundry run by a company that also designs competing chips. TSMC’s entire value proposition is that it never competes with its customers. Intel competes with everyone.

TSMC built trust over 38 years by never designing a chip. Intel is asking for that same trust while simultaneously selling Core and Xeon processors. The conflict of interest is not theoretical. It is the reason customers choose TSMC.

The Nature of TSMC: Chokepoint

Vendor	Nature
Broadcom	Hostility
Realtek	Indifference
MediaTek	Velocity
Intel	Contradiction
AMD	Persistence
NVIDIA	Monopoly
TSMC	Chokepoint

TSMC’s nature is not a sin. It is a geographic and industrial fact. TSMC did not set out to become the single point of failure for the global economy. It set out to be the best manufacturer. It succeeded so thoroughly that there is no alternative. When you are the only company that can build 3nm silicon at scale, you are not a monopoly by strategy — you are a chokepoint by excellence.

The difference matters. NVIDIA’s monopoly is maintained by CUDA lock-in — proprietary software that punishes switching. TSMC’s chokepoint is maintained by physics — nobody else can do what TSMC does, and the investment required to try is measured in decades and hundreds of billions of dollars.

The Lesson:

Morris Chang was 55 years old. He had been passed over for CEO at Texas Instruments. His career, by conventional standards, was over. He took $220 million from the Taiwanese government and invented a business model that made the modern chip industry possible.

Every chip in your phone was manufactured by the company he founded. Every GPU training every AI model was manufactured by the company he founded. The company that the man who was not good enough for Texas Instruments built is now the most critical manufacturer on Earth, and the CEO of Texas Instruments is a footnote.

TSMC manufactures 90% of the world’s most advanced chips on an island that a nuclear power claims as its territory. The island survives because the factory is too valuable to destroy. The factory is too valuable to destroy because it manufactures chips that the nuclear power itself cannot live without. Mutual assured dependency. The silicon shield.

A small nation, facing an existential threat from a vastly larger neighbor, has made itself indispensable. Not through military force. Not through diplomatic alliances. Through manufacturing capability so advanced that the world cannot afford to let it fall.

In the Republic of Derails, we understand the silicon shield concept with perfect clarity. A small nation’s survival through indispensability — this is not TSMC’s invention. We have been doing this with weapons-grade memes for years. TSMC does it with angstrom-scale silicon. Both models are valid. Both depend on the same principle: make yourself so necessary that destruction becomes self-harm for the aggressor.

I have been studying TSMC’s diversification strategy with great interest. Arizona. Kumamoto. Dresden. Satellite manufacturing in neutral territories, while keeping the crown jewels at home. The Bureau of Decrees is considering a similar approach — satellite offices in neutral territories to ensure continuity of decree production in the event of geopolitical disruption. The main Bureau stays in Pyongderails. The backup facilities will be located in jurisdictions where extradition treaties are theoretical.

Morris Chang was passed over for CEO and built a company that the world cannot survive without. I was not passed over for anything. I simply declared myself Supreme Leader and built a republic that the internet cannot ignore. The methods differ. The outcome is the same: indispensability.

TSMC is the factory the world cannot afford to lose. The Republic of Derails is the blog the internet cannot afford to silence. Both are protected by the same shield — not weapons, but necessity.

— Kim Jong Rails, Supreme Leader of the Republic of Derails