1. A Founder at 56
In 1985, the government of Taiwan placed a phone call to Texas.
The man who picked up was Morris Chang (張忠謀), 54. MIT master's in mechanical engineering, Stanford PhD in electrical engineering. A man who had spent 25 years at Texas Instruments, rising to senior vice president, and who had never received the CEO seat he had wanted. The Taiwanese government's offer was simple. Take over the Industrial Technology Research Institute (ITRI) and pull Taiwan's backward semiconductor industry up to world class.
Chang studied the offer for two years. And at 56, he founded an entirely new company. In 1987 — TSMC, Taiwan Semiconductor Manufacturing Company.
His idea inverted the common sense of the industry at the time. Established chip firms like Intel, TI, and Motorola were vertically integrated: they designed, manufactured, and sold their own chips. Chang asked a completely different question.
"What if there were a company that only manufactured chips — for firms that want to design chips but cannot afford to own a factory?"
The pure-play foundry. It designs no chips of its own. It sells no chips of its own. It only makes other people's chips. Why was this revolutionary? Because the model allowed NVIDIA, AMD, Qualcomm, and MediaTek to concentrate entirely on design without owning multi-billion-dollar fabs. The fact that NVIDIA today has a market capitalization in the trillions would be impossible without TSMC.
And one decisive principle sat underneath the model. We do not compete with our customers. That is fundamentally different from Samsung, which manufactures its own chips (Exynos) while simultaneously running a foundry for other firms. TSMC makes chips for Apple, for NVIDIA, and for AMD, but it does not become a competitor to any of the three. That trust architecture is the core of TSMC's competitive advantage. Customers hand their most precious blueprints to TSMC because they are certain TSMC will not copy those blueprints and build chips of its own.
The capital structure was significant too. The Taiwanese government's National Development Fund took 48 percent, the Dutch firm Philips took about 28 percent, and the rest was held by private investors. The state planted the seed, the private sector grew it. It was the Taiwanese version of the "designed indispensability" Lee Kuan Yew had executed in Singapore. The formula we saw in Chapter 3 reappears here.
Morris Chang formally retired in 2018. A 31-year run, from 56 to 87. The structure he designed became, in that interval, the core of global semiconductor geopolitics. And the survival strategy of a small island country.
2. The Most Important Company in the World
As of 2025, look at TSMC's position in numbers.
Global foundry market share: 67 percent. The gap to second-place Samsung at 8 percent is 59 percentage points. Its share of the leading-edge process (3 nanometers and below) is above 90 percent. 2024 annual revenue: roughly $88.2 billion. Its market capitalization oscillates between $800 billion and $1 trillion.
The customer list makes the company's meaning sharper. The number-one customer is Apple (about 25 percent of revenue). Every chip that goes into an iPhone, a MacBook, or an iPad comes out of TSMC. Number two is NVIDIA (about 12 percent). The H100, H200, and B200 GPUs — the very heart of the AI revolution — are manufactured exclusively by TSMC. Number three is AMD, number four MediaTek, number five Qualcomm, number six Broadcom, number seven Intel. The top ten customers together account for 91 percent of revenue.
State the meaning bluntly. TSMC is the manufacturing subcontractor of all of Silicon Valley. But unlike the connotation of the word "subcontractor," it is TSMC that is the less replaceable side of this relationship. Apple could find another foundry. But there is no alternative on earth with TSMC-level yields and process technology.
The technology roadmap locks the monopoly in further. TSMC is currently in volume production at 3 nanometers, and will begin volume production of 2 nanometers in the fourth quarter of 2025. The 2-nanometer generation is a revolutionary transition — from the existing FinFET structure to nanosheet transistors. Performance rises 10 to 15 percent. Power consumption falls 25 to 30 percent. Beyond that, A16 (a 1.6-nanometer-class process) is scheduled for volume production by the end of 2026. That process, which applies backside power delivery (SPR), is dedicated to AI data centers and saves 20 percent of power at the same performance level.
Samsung is struggling with yield at 3 nanometers. China's SMIC is still at 7 nanometers. TSMC is two or three generations ahead, and at the same time preparing the next generation. The gap is not simply a matter of the size of the investment. It is the product of manufacturing know-how accumulated over decades — the micro-tuning of equipment settings, the tacit knowledge of defect patterns, the intuition of process engineers. In Morris Chang's phrase: learning is local.
The physical base of that overwhelming position is the scale of capital expenditure. TSMC's 2025 CapEx is projected at $38 billion to $42 billion. No other company on earth invests that much in semiconductor manufacturing. It has risen almost threefold from $14.9 billion in 2019 in six years.
Look at the scale of a single fab. Fab 18, located at the Southern Taiwan Science Park in Tainan. Two hours south of Taipei by high-speed rail, an expanse of flat land on the outskirts of Taiwan's second-largest city. From outside it looks surprisingly ordinary. Light gray walls, repetitive rectangular volumes, parking lots that seem to run on forever. For the place where the most precise manufacturing on earth is carried out, its outward face is almost aggressively expressionless.
The inside is the most precise space on the planet.
Each phase of cleanroom floor area is 58,000 square meters. Eight phases are in operation. A ninth is under construction. Cleanliness class ISO 1–3 — airborne particles are controlled down to 100 nanometers. For comparison, operating rooms run at ISO 5–7. The Fab 18 cleanroom is a thousand times cleaner than an operating room.
To enter, you wear a head-to-toe garment called a bunny suit. You pass through an air shower, put on anti-static slippers, wear goggles and a mask. This is a place where circuits are etched at 3 nanometers — roughly one twenty-thousandth the thickness of a human hair (about 70,000 nanometers). A single flake of skin, a single sneeze, can turn a multi-million-dollar wafer into scrap.
Under the yellow light of the cleanroom, 300-millimeter silicon wafers ride on the arms of automated robots. Hundreds of chips are etched on a single wafer. Each chip integrates billions of transistors. The process runs through hundreds of steps — oxidation, photolithography, etching, deposition, ion implantation, cleaning — and from the moment a wafer is first loaded until a finished chip comes out, it takes about two to three months. Twenty-four hours a day, 365 days a year, the line does not stop.
Total investment: NT$1.86 trillion (more than $58 billion). Construction alone mobilized 23,500 workers. Direct high-tech employment is 11,300. One factory is one city.
On December 29, 2022, TSMC held a ceremony at Fab 18 to mark the start of 3-nanometer volume production. Morris Chang attended in person. Taiwan's economic minister, senior figures from the semiconductor industry. The 91-year-old founder stepped to the microphone.
"Today is Taiwan's day."
There was pride in that line, and there was weight. In the factory of a company he had started 35 years earlier at the age of 56, the finest structures humanity can build were rolling off the line. The chips made in this plant drive the iPhone, run AI servers, and move self-driving cars. And they defend the country called Taiwan.
3. The Logic of the Shield
The term "silicon shield" was first used in 2001 by the analyst Craig Addison. The logic is simple and powerful.
China invades Taiwan. Semiconductor production in Taiwan, TSMC included, halts. The global semiconductor supply chain collapses. Bloomberg estimates $5 trillion in economic damage in the first year alone. The United States, Japan, and Europe have no choice but to intervene. Therefore — Taiwan's semiconductor hegemony is a military deterrent.
A shield made not of nuclear weapons but of chips. Not the threat of destroying the other side, but the economic version of mutually assured destruction: if you destroy me, you collapse too. If TSMC stops, NVIDIA's AI chips disappear, Apple's iPhones disappear, AMD's server chips disappear. The AI revolution itself grinds to a halt.
There is evidence that this shield has worked so far. The fact that China has not invaded Taiwan. But this cannot be ascribed to the silicon shield alone. The U.S. Seventh Fleet, U.S.–China economic interdependence, the real limits of People's Liberation Army amphibious capability, the pressure of international opinion — all of these are elements of deterrence. The silicon shield may not be the only bulwark. But it is hard to deny that it is the most powerful economic element of deterrence. The fact that American economic calculation changes the moment TSMC stops is a kind of deterrence no fleet can provide.
But the shield has cracks.
The first crack is American strategic ambiguity. The United States has never clearly pledged to defend Taiwan. Former President Biden said on several occasions that "we would defend Taiwan," and each time the White House immediately corrected him with "no change in policy." Under Trump's second term, the ambiguity has deepened. I don't comment on that. I don't want to have to be in that position. In the State Department's 2025 factsheet, the language on unilateral changes to Taiwan's status was softened from oppose to does not support.
The ambiguity is deliberate. A clear defense guarantee could provoke a Taiwanese declaration of independence; a clear promise of non-intervention could invite a Chinese invasion. But from TSMC's standpoint, the silicon shield only gives the United States a reason to intervene. It does not guarantee intervention.
The second crack is China's push toward semiconductor self-sufficiency. On August 29, 2023, on the very day U.S. Commerce Secretary Gina Raimondo was visiting Beijing, Huawei released its new smartphone, the Mate 60 Pro, without any advance notice.
It took several days for the teardown analyses to come out. Inside was something the United States had believed impossible. A 7-nanometer chip made by SMIC — the Kirin 9000S. Without EUV equipment, using only DUV, the Chinese had reached 7 nanometers. The Western assumption that China cannot make advanced semiconductors had been shaken.
On the Chinese internet, a meme spread mocking Raimondo as "the unofficial brand ambassador of the Mate 60 Pro." At the very moment an American commerce secretary was in Beijing talking about strengthening sanctions, Huawei was replying: we are still alive.
There is important context. SMIC's 7 nanometer has low yields and high cost. A two- or three-generation gap still separates it from TSMC's 3 nanometer and 2 nanometer. China's semiconductor self-sufficiency rate was roughly 23 percent as of 2023. That is far below the 70 percent goal set by Made in China 2025. But the trend is clear. Containment did not break Chinese determination toward self-sufficiency. It accelerated it.
For the silicon shield to be effective, TSMC must remain irreplaceable. If China begins to produce its own "good enough" chips, the shield corrodes. Not yet. But not forever not, either.
The third crack is the sharpest. The Arizona paradox.
4. Shield or Trap
The U.S. CHIPS Act (2022) is a law to bring semiconductor manufacturing back to American soil. TSMC is investing $165 billion in Arizona. Samsung is putting tens of billions into Texas. Intel is building fabs in Ohio. On the surface, these are measures to strengthen supply-chain security.
But seen from Taipei, the story changes.
Once the United States begins to make Taiwanese semiconductors on its own soil, its motive to intervene in a Taiwanese crisis weakens. Some scholars call this the Silicon Trap. Once the Arizona fabs are running and the Kumamoto fab is running, the West can secure semiconductors even if Taiwan is gone. Paradoxically, the measures taken to reinforce the silicon shield can be the very measures that corrode it.
No one understood this paradox better than Morris Chang.
In an interview after his 2022 APEC speech, he said:
"In semiconductors, globalization is dead. Free trade is no more. The only priority is national security."
A man who since his TI years had believed in the separation of design from manufacturing, and in borderless supply chains, saying only national security is the priority. The weight of that line. It was a lament. The lament of one who had watched the world he had spent a lifetime building come apart in the hands of geopolitics.
His skepticism of the Arizona fab was more concrete.
"Learning is local. The experience curve only works when you are in the same place."
The technician ecosystem, the supply-chain cluster, the tacit knowledge accumulated over decades in Taiwan — reproducing all of it in the Phoenix desert is either impossible or extremely expensive.
His skepticism was not wrong.
5. Culture Clash in the Desert
In 2021, when TSMC announced a semiconductor plant on the outskirts of Phoenix, Arizona, the American press cheered. The start of semiconductor independence. The original plan was $12 billion, a 5-nanometer process, operations beginning in 2024. But the moment construction started, a different kind of war broke out.
In mid-2023, TSMC formally announced production delays. A shortage of skilled technicians. More than 1,000 engineers had to be sent back to Taiwan for training. The target slipped to 2025 — a year late.
The real problem was not equipment or money. It was culture.
The TSMC-style work intensity that the Taiwanese engineers brought with them — night calls, direct rebukes from supervisors, immediate execution once a target was set — collided with American labor. American employees wanted to know the purpose before starting a job: why do I have to do this? Taiwanese employees, once a target was set, did not ask why and simply immersed themselves in the work. One side called the other arrogant. The other side called the first submissive.
A reporter's account left a description.
"The Taiwanese employees described their Phoenix colleagues as arrogant, laid-back, and more willing to push back against orders. The American employees said the Taiwanese engineers were bewildered when workers asked why they had to do something before doing it."
TSMC introduced weekly English and cultural classes. It issued guidelines to Taiwanese managers: do not rebuke employees in public; consult HR before threatening dismissal. It sent American employees to the Taiwanese headquarters to experience the TSMC way firsthand.
In the fourth quarter of 2024, the first phase of Fab 21 began volume production on a 4-nanometer process. Of the roughly 2,200 employees, 50 percent were from Taiwan. TSMC announced that the yield had reached parity with Taiwanese fabs.
But that was only the beginning. In March 2025, President Trump pressured TSMC for an additional $100 billion in U.S. investment. TSMC expanded its plan to a total of $165 billion and opened the possibility of up to 12 fabs.
Arizona was not the only one. TSMC's overseas fabs were stretching out in three directions at once.
Kumamoto: The Opening Ceremony on the Rice Fields
February 24, 2024. Kikuyō Town, Kumamoto Prefecture, Japan. On land that had until recently been rice fields, a plant was inaugurated. Morris Chang, Chairman Mark Liu, and CEO C.C. Wei all attended. Japan's minister of economy, trade, and industry, the executives of Sony Semiconductor, officials from Denso and Toyota filled the hall. Prime Minister Fumio Kishida sent a video message of congratulations. It was a symbolic scene of an era in which semiconductors had become a state affair.
JASM (Japan Advanced Semiconductor Manufacturing). A joint venture with TSMC at 86.5 percent, Sony Semiconductor at 6 percent, Denso at 6 percent, Toyota at 1.5 percent. The Japanese government put in ¥800 billion in subsidies (about $5.4 billion) — the largest single-firm semiconductor subsidy in Japanese history. The first fab runs 12-to-28-nanometer processes with a monthly capacity of 55,000 wafers. These are not leading edge but legacy processes. The fab specializes in automotive and industrial semiconductors.
Economy minister Ken Saitō said: "This is not just a factory. It is the beginning of the rebuilding of Japan's semiconductor industry."
The second fab broke ground at the end of 2024, targeting a 6-nanometer process with operations to begin in 2027. An additional ¥490 billion in subsidies was confirmed. Kumamoto was chosen for practical reasons — adjacency to Sony's existing image-sensor plant, the clear groundwater essential to semiconductor manufacturing (Kumamoto is famous for its water quality), and relatively low seismic activity.
Dresden: The European Foothold
In August 2024, ground was broken on the ESMC (European Semiconductor Manufacturing Company) plant in Dresden, Germany. A partnership with Infineon, NXP, and Bosch, scheduled for completion in 2027. It, too, specializes in automotive and industrial-grade process technology.
Arizona, Kumamoto, Dresden. TSMC's overseas expansion across three continents is the physical form of "friend-shoring" — the strategy of rebuilding the semiconductor supply chain inside allied countries, now being realized in concrete.
One heavy question rises here. When all this investment is complete in the 2030s, will Taiwan's silicon shield still exist?
There is a counter-argument. Kumamoto runs 12 to 28 nanometers. Dresden handles specialty automotive processes. Arizona started at 4 nanometers. The leading edge — 2 nanometer, A16 — remains concentrated in Taiwan. The overseas fabs handle older-generation technology, and the core of the leading edge stays in Hsinchu and Tainan. And semiconductor manufacturing is not a matter of equipment alone. As Morris Chang said, learning is local. Transplanting the decades of ecosystem accumulated in Taiwan into a desert, rice fields, and a forest is not the same thing as building a plant.
The current conclusion is this. The silicon shield is still working. But there is a structural risk that its force will weaken after the 2030s. The lifespan of a shield is not infinite.
In Chapter 1 we saw that the monopoly of the Hansa was not eternal. Just as the technological innovation of Dutch merchants went around Lübeck's monopoly, Chinese self-sufficiency and American reshoring can go around TSMC's monopoly. The core proposition of Part I rings again here: indispensability is not forever.
6. The Day They Cut Off Huawei
Sunday evening, September 14, 2020.
A statement came out of TSMC's press office. "TSMC will stop shipments to a specific customer starting September 15, 2020." Everyone knew the specific customer was Huawei.
Huawei had been TSMC's number-two customer, accounting for 14 percent of total revenue. The Kirin chips designed by Huawei's subsidiary HiSilicon were manufactured by TSMC. Tens of billions of dollars in annual contracts were cut off overnight.
Here is the background. In May 2020, the Trump administration tightened the Huawei regulations again. Under the expanded Foreign Direct Product Rule (FDPR), any foreign firm that used American equipment or software to supply Huawei would need U.S. government approval. Since most of TSMC's equipment is American (Applied Materials, Lam Research, KLA) or contains American technology (even ASML's EUV uses American light-source technology), the very act of making chips for Huawei became illegal.
That moment was the most dramatic decision in Taiwan's in-between strategy.
Can Taiwan hold a neutral line between the United States and China? TSMC answered the question by action. It cut off semiconductor supply to China's largest telecom-equipment company. The Taiwanese government supported the move. It gave up tens of billions of dollars in annual revenue and folded itself into the U.S. sanctions regime.
No official statement came out of Beijing. But the implication was clear. The most important chip company in the world had chosen the American side.
Employees inside TSMC were divided. Some felt we have been dragged onto the battlefield of geopolitics. Some resigned themselves: there was no choice; 90 percent of our equipment is built on American technology.
A company that had been built on the principle of not competing with our customers now stood in front of another principle: U.S. national security. When the two principles collided, geopolitics beat the market.
In 2024, a more intricate incident occurred. A Chinese customer of TSMC was caught secretly routing TSMC-made chips into Huawei's AI accelerators. TSMC immediately suspended dealings with the customer and reported it to the U.S. government. TSMC had, in effect, become an enforcement agent of the U.S. export-control regime.
The lesson for Korea is direct. Samsung Electronics and SK hynix can be forced into the same choice at any time. As we saw in the Prologue, in October 2022 Samsung's Xi'an fab had already come close to that moment. The U.S. export-control rules had tried to make Korean firms' Chinese fabs illegal, and Korea bought breathing room through a reprieve. But in August 2025 that reprieve was converted into an annual licensing regime. There is no permanent exemption.
Taiwan has already made its choice. Clearly, irreversibly, it has chosen the American side. Korea has not yet arrived at its moment of final choice. But nothing guarantees that moment will not come.
7. The Politics of the Tightrope — Structural Change in Cross-Strait Relations
Cutting off Huawei was not a single event. It was part of a current. To understand how Taiwan's in-between strategy has evolved, you have to look at the structural change in cross-strait relations.
The Tsai Ing-wen years (2016–2024). The most explicitly pro-American and anti-Beijing leader in Taiwanese history. In a 2021 Foreign Affairs essay, she formally framed semiconductors as Taiwan's shield.
"The semiconductor industry enables Taiwan to protect itself and others from the supply-chain disruptions of authoritarian regimes."
Tsai fully supported TSMC's Arizona investment, joined in the Huawei sanctions, and expanded the defense budget from 2.1 percent of GDP to 2.5 percent. And she made the most symbolic decision of all — the reintroduction of one-year mandatory military service for men, to begin in 2024. It was a signal that security now took priority over economic opportunity in cross-strait relations.
The Lai Ching-te years (2024–present). A more explicit supporter of Taiwanese independence than Tsai. Beijing has labeled him a "separatist." From the moment of his inauguration, cross-strait relations have grown more tense. Lai's task is to complete the export-control and investment-screening mechanisms the Tsai administration left unfinished while supporting TSMC's global expansion.
The trajectory of the defense budget tells the story in numbers. Roughly 1.7 percent of GDP in 2020 → 2.1 percent in 2022 → 2.5 percent in 2024 → targeted 3.3 percent in 2026. The NATO standard (2 percent) was crossed starting in 2023. But the absolute size of China's defense budget is many dozens of times larger. Matching the balance in the Taiwan Strait with conventional forces alone is impossible. That is why the silicon shield must function not as a simple economic strategy but as an alternative to military power.
The numbers on cross-strait trade are shifting too. Taiwan's export share to mainland China and Hong Kong fell from 43.9 percent in 2020 (the peak) to 32.9 percent by early 2024 — the lowest in 21 years. This is structural change more than a political act. A slowdown in Chinese growth, a surge in demand from the United States, Japan, and Europe for Taiwanese semiconductors during the AI boom, and Foxconn's dispersion into India and Vietnam all compounded. Still, roughly one-third of exports still go to China and Hong Kong. It is still the single largest export market.
China uses that dependence as leverage. The Economic Cooperation Framework Agreement (ECFA), signed in 2010, had given cross-strait trade preferential treatment. In January 2024, China ended ECFA preferences on 12 chemical products, and it announced that it was considering ending preferences on agricultural, fishery, machinery, and textile products as well. Every time Taiwan strengthens its pro-American line, Beijing tightens the economic screws. It is a signaling mechanism.
This is the cost of the in-between strategy. In Chapter 2 we saw that Swiss neutrality and Finnish patience both had their price. Taiwan's price is double. It pays for choosing the American side with Chinese economic pressure, and it pays for transferring technology to the United States with the risk of a weakened shield. Taiwan is not paying the cost of neutrality. It is paying the cost of bandwagoning.
8. A Country of Two Faces — The Foxconn Lesson and Taiwan's Displaced
There is one more shadow on Taiwan's in-between strategy.
Consider Terry Gou, the founder of Foxconn (鴻海精密工業). Foxconn built its first Chinese plant in Shenzhen in 1988, and today produces more than 70 percent of Apple's iPhones in China. It is the largest electronics contract manufacturer in the world. In 2023, when Terry Gou announced a run for the presidency of Taiwan, Chinese authorities launched simultaneous tax investigations against Foxconn's Chinese plants. Gou withdrew from the race in November.
The message was clear. A Taiwanese company that has invested in China cannot be free of Chinese political pressure. Even the world's largest contract manufacturer bowed to tax audits. This is a textbook case of how economic dependence converts into political leverage. Korea had the same experience during the THAAD retaliation of 2016–2017. The withdrawal of Lotte Mart from China, the unofficial "K-ban" (限韓令) — both countries know in the skin that economic dependence is vulnerability.
But Taiwan's deeper problem is internal.
The stronger TSMC becomes, the greater the imbalance inside Taiwanese society. Semiconductors account for roughly 20 to 25 percent of GDP and 40 percent of exports. TSMC alone accounts for about 8.9 percent of GDP. But that enormous wealth is concentrated in a very narrow place. ICT workers are only 11 percent of the entire labor force.
When the median annual income in Hsinchu's ICT sector is about ₩44 million, Taiwan's overall median income is about ₩17.7 million. A 2.5-fold gap. The widening gap between a semiconductor engineer and a convenience-store clerk is the reality of Taiwanese society.
The house-price-to-annual-income ratio in Taipei is 15.5. It has nearly doubled from 8.03 in 1993. While the semiconductor boom pushes real-estate prices up, young people outside the semiconductor sector face a reality in which buying a home is impossible. Youth unemployment for those aged 20 to 24 is about 12 percent — four times the overall rate.
The irony is that a shortage of semiconductor workers is equally severe. As of 2025, there are 34,000 unfilled positions in the semiconductor industry. Taiwan's annual births fell from 210,000 in 2014 to 135,000 in 2024, and the number of STEM graduates fell by 15 percent over the same period. The industry wants people, and the people either cannot enter the industry or are not being born.
This is the two faces of Taiwan. The capital of world semiconductors, and at the same time a country where the young cannot buy a home. The 77,000 TSMC employees and their supplier network are the beneficiaries of the AI age. Workers in every other industry — manufacturing, services, agriculture — are hit by the headwind of a strong Taiwan dollar eroding export competitiveness and high prices eroding purchasing power.
In the apartments around Southern Taiwan Science Park in Tainan, TSMC employees wait every dawn for the commuter bus. Their salaries are more than double the Taiwanese average. But so is their anxiety. Conscription is back — a year of mandatory service for men was reintroduced in 2024. Talent outflow — as the Arizona fab comes online, more experienced engineers head overseas. And the most fundamental contradiction of all: they know the chips they make are the shield defending Taiwan, and at the same time they are being asked to build the Arizona fab that can weaken that shield.
An engineer told a reporter:
"We make the most important chips in the world, and we cannot buy a house. If a war comes, we will be the first target."
In Book 1 we saw the formula in which technological innovation concentrates capital and concentrated capital produces social instability. In Taiwan that formula is running at the scale of an entire country. The technological innovation called TSMC is concentrating capital in the semiconductor industry, and that concentration is producing the social instability of a dual economy. The silicon shield may stop invaders from outside. It cannot stop cracks from inside.
Korea in the Mirror
Taiwan and Korea. Both East Asian, both semiconductor powers, both standing between the United States and China. But the face in the mirror is not the same.
Difference 1: Concentration vs. Distribution.
Taiwan has bet the fate of the country on a single company, TSMC. Semiconductors account for 40 percent of exports. Korea is different. In 2024, semiconductors were about 21 percent of exports; the rest was distributed across automobiles, ships, batteries, K-content, and bio. Total exports of $683.8 billion were the highest on record. The all-in strategy of Taiwan versus the portfolio strategy of Korea. In the short run, Taiwan's concentration produces stronger leverage; in the long run, Korea's distribution provides stronger buffering.
Difference 2: The Nature of the Monopoly.
TSMC's leading-edge foundry is irreplaceable. A 90 percent share. But Korea's memory — DRAM at 70.5 percent (Samsung plus SK hynix) and NAND at 52.6 percent — has alternatives in Micron of the United States and Kioxia of Japan. Foundry means logic chips — the brains of AI, smartphones, and servers. Memory is the storage of those brains. Both are essential, but the intensity of "irreplaceability" is different.
There is one exception, however. SK hynix's HBM (high-bandwidth memory). In this component, essential to NVIDIA's AI accelerators, SK hynix's share is over 50 percent. Including Samsung, Korea's two firms together hold more than 60 percent of the global HBM market — what TSMC has in leading-edge foundry, Korea has on the narrower front of HBM. SK hynix's 49 percent operating margin in 2024 surpassed Samsung Memory's and showed that the biggest beneficiary of the AI boom was the HBM monopoly. HBM is the most plausible candidate for a Korean silicon shield. A small shield, but a sharp one.
Difference 3: Samsung's Dilemma.
Taiwan has one TSMC. Korea has two firms, Samsung and SK hynix. But Samsung competes with TSMC in foundry and at the same time competes with SK hynix in memory. A war on two fronts. In foundry, yield problems at 3 nanometers drove customers like Qualcomm and NVIDIA over to TSMC, and Samsung's market share dropped to 8 percent. A 59-percentage-point gap behind TSMC's 67 percent — from TSMC's point of view, Samsung Foundry is not so much a competitor as a distant pursuer. In memory, Samsung ceded the HBM lead to SK hynix.
Meanwhile SK hynix is producing results with the clear strategy of "HBM focus." The closest analogue in Korea to Taiwan's TSMC role is not Samsung as a whole but SK hynix's HBM business.
Difference 4: Geopolitical Position.
China claims sovereignty over Taiwan, and the military threat is direct. Korea has the security variable of North Korea, but a scenario in which China annexes Korea by force is not realistic. Korea's between is less physically dangerous than Taiwan's between, but the intensity of economic pressure is similar.
What the Mirror Reflects.
Taiwan shows Korea two things. One is a possibility. Technological monopoly can be converted into the geopolitical leverage of a small country. HBM shows the way. The other is a warning. The vulnerability of a strategy that bets everything on a single industry. When TSMC shakes, the Taiwanese economy shakes as a whole. When the semiconductor cycle turns down in Korea, batteries, K-content, and shipbuilding buffer the blow.
But Korea's portfolio is not perfect either. Semiconductors, autos, and ships all share the structural vulnerability of having China as their largest export market. The strength of distribution can flip into the weakness of dependence.
Perhaps Korea's in-between strategy has to be different from Taiwan's. Not a single shield, but a dense net woven from many threads. Taiwan holds the world hostage with a single blade. Korea has to weave a net of many threads so that no side can cut it through.
The density of that net is indispensability itself.
On to the next mirror. This time, a country that raises unicorns on a battlefield — Israel.