1. The Heart of Icheon
Autumn 2024. Bubal-eup, Icheon, Gyeonggi Province.
Inside SK hynix's M16 fab on the Icheon campus, the HBM3E 12-layer stacking production line runs around the clock. The process: stacking twelve layers of DRAM die vertically, penetrating each layer with TSV (through-silicon via), bonding layers with microbumps, and inserting a non-conductive film between each layer for heat dissipation — repeated, without interruption. The finished HBM3E stack is approximately 775 micrometers thick. Inside a structure thinner than ten human hairs laid side by side, 36 gigabytes of data move at 1.18 terabytes per second.
This component goes into NVIDIA's H200 GPU. Six HBM3E units per H200. Hyperscale data centers worldwide deploy hundreds of thousands of H200s a year, and every one of those GPUs carries SK hynix's HBM. The critical component of the AI revolution comes out of a cleanroom in Icheon.
Jensen Huang, NVIDIA's CEO, publicly called SK hynix "the core partner of AI infrastructure." That was not diplomatic language. As of 2024, SK hynix holds 62 percent of the HBM market. Micron is second at 21 percent. Samsung Electronics is third at 17 percent.
If SK hynix stops supplying, NVIDIA's AI accelerator production stalls immediately. In Chapter 4 we saw Taiwan's TSMC function as the silicon shield. Does Korea have a shield? HBM is the closest answer to that question.
But this shield has an expiration date. Once Micron scales up HBM3E production in earnest and Samsung resolves its thermal issues, the 62 percent share will not hold at its current level. The sharpness of a shield comes from a technology gap, and technology gaps dissolve with time.
2. Semiconductors — Two Faces of a Giant, and the Empty Middle
The numbers first. In 2024, Korea's semiconductor exports reached $141.9 billion — approximately ₩190 trillion. Growth of 43.9 percent year on year. An all-time high. Semiconductors alone accounted for 20.8 percent of total exports of $683.8 billion. In 2025, the figure rose again to $173.4 billion, breaking the record once more. Semiconductors are the heart of Korea's export economy.
That heart has two chambers: Samsung Electronics and SK hynix.
Samsung Electronics' semiconductor division (DS Division) recorded ₩131 trillion in revenue in 2024 — its highest ever. It recaptured the top spot in global semiconductor revenue, surpassing Intel again with a 10.5 percent market share. In DRAM, Samsung remains the world's number one. In NAND flash, it leads at 32.8 percent. Combined, Samsung and SK hynix together hold 70.5 percent of the global DRAM market and 52.6 percent of NAND.
But within that same Samsung, there are fractures.
In foundry — manufacturing chips on behalf of other companies — Samsung's share stood at 8.1 percent in the fourth quarter of 2024. The gap behind TSMC's 67.1 percent is 59 percentage points. Samsung was the first in the world to introduce a 3-nanometer GAA (gate-all-around) process in 2022, but yield problems recurred. Large customers including Qualcomm and NVIDIA moved to TSMC.
On the 2-nanometer process, Samsung reported achieving yields of 55 to 60 percent. TSMC, at the same point in time, had already reached 65 percent. That 65 percent is not a simple number. Once that threshold is crossed, major customers lock in their schedules and fix their orders. While Samsung remained below that line, Qualcomm and NVIDIA did not wait. The problem is less one of technology itself than one of customer trust. Repeated yield instability and schedule delays have solidified a "second-tier" image.
In HBM, Samsung has also fallen behind. In 2023 and 2024, Samsung's HBM3E failed NVIDIA's qualification tests due to thermal and power-consumption problems. An HBM market share of 17 percent. Against SK hynix's 62 percent, the gap is difficult to believe between two Korean companies. Samsung is fighting a two-front war — competing against TSMC in foundry while simultaneously losing ground to SK hynix in memory.
SK hynix, by contrast, achieved record results with the clear strategy of "HBM focus." In 2024: revenue of ₩66.19 trillion, operating profit of ₩23.47 trillion — both the highest figures in the company's history. Year-on-year revenue growth of 102 percent. HBM accounted for more than 40 percent of DRAM revenue.
In Chapter 7 we saw ASML hold a globally dominant position in EUV lithography equipment. SK hynix has built the closest approximation to that monopoly structure — in HBM.
The next chapter is being prepared in Yongin. SK hynix's Yongin semiconductor cluster: investment in the first fab alone stands at ₩31 trillion — approximately $21.5 billion. Construction broke ground in March 2025, with completion targeted for 2027. Chey Tae-won, chairman of SK Group, declared it will become the world's largest mega-fab complex, incorporating four fabs and more than 50 partner companies. Long-term investment is projected at more than ₩600 trillion.
Samsung, too, is pouring more than ₩30 trillion into the P5 fab in Pyeongtaek, and has announced plans to invest a total of ₩450 trillion — approximately $310 billion — in AI and semiconductors over five years. The government-led Gyeonggi semiconductor mega-cluster will invest ₩622 trillion through 2047. In Chapter 9 we saw the UAE, through MGX, target $100 billion in AI infrastructure investment. The logic is the same — the judgment that securing a stake in future technological dominance now is the only option, because it cannot be purchased later. Korea's ₩600 trillion and the UAE's $100 billion differ in scale and background, but the calculation beneath them is one.
Yet above this colossal investment sits an empty middle. The vulnerability of the materials, parts, and equipment ecosystem — sobujang (소부장).
Korea's domestic self-sufficiency rate for semiconductor equipment stands at roughly 20 percent. The remaining 80 percent depends on imports. In Chapter 7 we saw that a single ASML EUV machine costs $220 million — approximately ₩300 billion — and that there is exactly one company on earth capable of building it. Every advanced fab operated by Samsung and SK hynix depends on that machine from a small Dutch town.
EUV is not the only constraint. For etch equipment: Applied Materials and Lam Research (United States). For inspection equipment: KLA (United States). For metrology and process equipment: Tokyo Electron (Japan). The supply lines for critical equipment are concentrated in the United States and Japan.
The bottleneck in materials runs just as deep. In July 2019, when Japan tightened export restrictions on three critical semiconductor materials — photoresist, hydrogen fluoride (etching gas), and fluorinated polyimide — the entire Korean semiconductor industry was shaken. Japan controls 87 to 91 percent of the global photoresist market.
In Chapter 10 we saw Japan choose to weaponize materials — and Korea was the direct target of that weaponization. In the years that followed, the Korean government and industry invested in localizing materials production, and the domestic self-sufficiency rate for hydrogen fluoride rose substantially. But for the most advanced materials — EUV-grade extreme ultraviolet photoresist in particular — dependence on Japanese suppliers continues. Just as ASML's moat, as we saw in Chapter 7, was thirty years of accumulated investment, the materials moat is built from decades of know-how. It is not a barrier that can be cleared by a localization announcement alone.
The structure, stated plainly, is this. Korea is the world number one in finished memory chips, but for the equipment and materials needed to make those chips, it is subordinate to the United States, Japan, and the Netherlands. The ASML equipment examined earlier contains American technology, meaning the United States exercises an indirect veto through the Netherlands. Korea sits at the end of that double bottleneck — bound to ASML for equipment, bound to American technology behind ASML, and bound to Japan for materials.
Applying the five conditions of indispensability from Chapter 3 to Korean semiconductors: technological irreplaceability is strong in HBM and weak in foundry. The position as a core node in the supply chain is solid in memory. But Korea's semiconductor shield is sharp — yet that sharpness is concentrated at a single point called HBM, and the tools that make the shield are someone else's.
3. Batteries — The Descending Wing
The trajectory of the Korean battery industry points in the opposite direction from semiconductors. In 2020: LG Energy Solution (LGES) at 24 percent, Samsung SDI at 6 percent, SK Innovation at 6 percent — the three Korean firms combined held 36 percent of the global EV battery market. More than one-third of the world's electric vehicle battery supply was Korean.
Five years later, in 2025, that number had fallen to 15 percent. Less than half. Over the same period, the combined share of Chinese firms surged to 69 percent. CATL at 39.2 percent, BYD at 16.4 percent — those two companies alone dwarf the Korean three combined.
The decline has three layers.
While Korean companies went all-in on NCM (nickel-cobalt-manganese) batteries, China locked up the low-cost market with LFP (lithium iron phosphate) technology. In the Chinese domestic market, LFP adoption reached 79.4 percent. The ripple effects extended to Europe. In the European market where Korea had held 70 percent in 2021, China crossed 49.7 percent by 2024, surpassing Korea's 45 percent. EV demand slowdowns compounded the blow to Korean firms directly. Factory utilization rates fell to 47.6 percent at LG Energy Solution and 43.6 percent at SK On. CATL maintained utilization above 90 percent — the contrast is stark.
BYD is a particular threat. A company that has completed vertical integration from battery cells to semiconductors, electric motors, and finished vehicle assembly. In 2024, BYD sold more than 4.27 million new-energy vehicles — surpassing Tesla on an annual basis for the first time. In Chapter 8 we saw Indonesia's nickel export ban as a resource-ladder strategy. BYD is the one looking down from the top of that ladder — having captured the entire value chain from battery raw materials to finished vehicles.
Yet Korean batteries still have weapons with which to fight.
LG Energy Solution's factory in Wroclaw, Poland is Europe's largest battery plant — annual capacity of 86 gigawatt-hours, 9,500 employees. It supplies cells to Volkswagen, Audi, Porsche, and Mercedes-Benz.
In the United States, a Korean-made EV cluster has taken shape in Georgia. At its center: Hyundai Motor Group's Metaplant America — $7.6 billion in investment, annual production capacity of 500,000 vehicles — surrounded by an SK On battery plant, an LG Energy Solution joint-venture factory, and five affiliates including Hyundai Mobis, forming a single ecosystem.
The next paradigm shift is also in preparation. Samsung SDI has announced its "Super Gap" technology targeting energy density above 900 watt-hours per liter, with volume production declared for 2027. In Chapter 10 we saw Japan's Toyota aiming at the same 2027 target. The transition from current lithium-ion to all-solid-state batteries can change the board itself — eliminating fire risk, raising energy density by 20 to 50 percent, enabling ten-minute charging. Whoever takes the lead in that transition can bypass CATL's dominance. Failure means the current decline becomes permanent.
4. The Rules War
August 16, 2022. President Biden signed the Inflation Reduction Act (IRA). Korea's three battery companies initially welcomed it. The judgment was that the enormous subsidies for U.S. domestic production would work in their favor by excluding Chinese companies. But within the fine print was an unfamiliar term — FEOC (Foreign Entity of Concern). Vehicles containing critical minerals or battery components supplied by companies under Chinese government control would be excluded from subsidies.
The entire supply chains of the Korean battery companies were shaken. More than 80 percent of lithium refining, 70 percent of cobalt refining, and 91 percent of anode material production are Chinese. Complying with the FEOC requirement means redesigning the supply chain from the ground up.
POSCO Future M's construction of a precursor plant in Gwangyang, South Jeolla Province — producing cathode materials without Chinese intermediate inputs — is the product of that pressure. The government also backed the supply chain transition with ₩9.7 trillion — approximately $7.1 billion — in policy finance.
The IRA was a supply-chain separation order wearing the skin of a subsidy law. The core is simple: exclude China. The structure: $35 per kilowatt-hour in tax credits for battery cell production in the United States, $10 for modules — but forfeited entirely if Chinese materials are included. UBS estimates that if Korea's three battery companies run their U.S. plants at full capacity, they could receive approximately $8 billion — about ₩11 trillion — in annual tax credits. But the condition for that money is a redesigned supply chain.
The same rules war is being fought in semiconductors. Samsung Electronics' Xi'an fab accounts for approximately 40 percent of Samsung's total NAND production. SK hynix's Wuxi fab produces about 40 percent of its total DRAM. In August 2025, when the U.S. government revoked Validated End User (VEU) status from Samsung and SK hynix, the future of those plants became uncertain.
After 120 days, a transition to an annual licensing regime provided some breathing room — but the structure had changed. Every year, Samsung and SK hynix must apply to the U.S. government, specifying the type and quantity of equipment, and await approval. There is no "permanent exemption." In the Prologue we saw the moment Samsung's Xi'an fab stood at the edge of the export-control cliff in October 2022. That tension was not resolved. It became an annual event.
The Trump administration pushed the rules war one step further. Legislation signed in July 2025 ended the consumer EV tax credit. This happened immediately after Hyundai announced $21 billion in U.S. investment — benefits cut right after the pledge was made. Promise, then withdrawal. In Chapter 4 we saw TSMC investing $165 billion in Arizona while remaining exposed to the volatility of American politics. Korean battery companies inhabit the same structure.
China writes its own rules too. Fully excluding China — which dominates the refining of battery critical minerals — drives costs sharply higher and destabilizes supply. Failing to exclude China triggers the U.S. FEOC requirement.
The IRA is simultaneously an opportunity and a shackle. FEOC is simultaneously a shield and a snare. The United States and China each write rules favorable to themselves, and Korea must comply with both sets of rules at the same time. This is the rules war. Korea holds weapons called chips and batteries — but the conditions for using those weapons are set by other countries.
5. The Crisis — Pursuers, Vacancies, and a Drying Pipeline
The crisis arrives from outside and from within.
China's CXMT (ChangXin Memory Technologies) recorded 2024 revenue of approximately $3 billion — three times its figure from two years prior. It has released DDR5 products and closed the technology gap with leading firms to one or two generations. In 2024, Korean prosecutors indicted ten former Samsung employees on charges of leaking 10-nanometer DRAM technology to CXMT. Samsung's technology may have accelerated CXMT's pursuit.
In Chapter 10 we saw Japan's semiconductor market share fall from 51 percent of the global market in 1988 to below 10 percent by 2019 — a collapse that took thirty years. And recovery required another thirty years. That is the lesson Rapidus embodies: losing semiconductors costs thirty years to recover.
In batteries, the pursuit has already become an overtake. CATL's 90 percent factory utilization rate is physical evidence of the cost advantage that scale creates. The Chinese government pays CATL and BYD each a minimum of approximately $2 billion in annual subsidies. Against the sub-50 percent utilization rates of the Korean three, this is not a technology gap. It is a structural gap.
But the crisis from within is more fundamental. There are not enough people.
Korea's total fertility rate is 0.72 — the lowest in the world. Seoul's is 0.64. How does this connect to semiconductors and batteries? A fertility rate is a leading indicator of the labor market twenty years out. A rate of 0.72 does not simply mean fewer children — it means the pool of engineers who will staff Samsung's Pyeongtaek and SK hynix's Yongin fabs in the 2040s is being determined right now. Running a single semiconductor fab requires thousands of highly skilled engineers. Operating EUV lithography equipment, diagnosing process defects, and pushing yields upward demands graduate-level physics and chemistry, plus years of field experience. Battery materials research and cell design are no different.
And the pipeline feeding that workforce is drying up.
Korea's aversion to science and engineering has become structural. As the medical school enrollment expansion debate illustrates, the highest-achieving students are streaming into medicine rather than engineering and science. The 2024 dispute over expanding medical school quotas triggered a residency-doctor strike that paralyzed the healthcare system for months.
Beneath that conflict lies a deeper problem. In a structure where a semiconductor process engineer's salary does not reach half of what a private-practice physician earns, it is rational for the brightest young people to choose the operating room over the cleanroom. Nationally, it is catastrophic.
In Chapter 4 we saw Taiwan's dual economy — a country where TSMC engineers say they cannot buy a home even while making the most important chips in the world. Taiwan's semiconductor workforce shortfall reaches 34,000 unfilled positions. Korea is losing population faster than Taiwan. Against Taiwan's total fertility rate of 0.87, Korea's 0.72 means Korea arrives at the workforce cliff first, and fastest.
The talent drain overseas is serious too. The migration of AI talent to the United States is accelerating. Google, Meta, and OpenAI are drawing Korean master's and doctoral AI researchers to Silicon Valley. Samsung Electronics and SK hynix need people who understand AI in order to build AI semiconductors. Those people are leaving Korea.
In Chapter 4 we saw TSMC relocate more than 1,000 engineers from Taiwan to Arizona to staff its new fab. In Korea, the state is not sending them — individuals are leaving on their own. Higher salaries, better research environments, more open cultures.
The long-term impact of a fertility rate of 0.72 has not yet arrived in full force. The shock hits the labor market in the 2040s. But the investment timeline for semiconductors and batteries is now. When the Yongin cluster is completed in 2027, when the Gyeonggi mega-cluster is finished in 2047 — where do the engineers to fill those fabs come from?
In Chapter 10 we saw Japan's figure of 76,020 solitary deaths. The population cliff is not only an industrial problem. It is a social one. Korea is following the same path as Japan at a pace compressed by twenty years.
In a structure where the four major chaebols — Samsung, SK, Hyundai, LG — account for 40.8 percent of GDP, the future of chips and batteries is effectively in the hands of a small number of chaebol decisions. Capital concentration enabled efficiency, but it also created single points of failure.
Samsung Foundry's yield problems are simultaneously a technology problem and an organizational one. Catching up with the manufacturing know-how TSMC has accumulated over decades — the micro-tuning of equipment settings, the tacit knowledge of defect patterns — requires time. During that time, customers do not wait. Samsung's selection as a supplier for NVIDIA's Rubin platform in HBM4 may be the beginning of a reversal. But the recovery in foundry remains uncertain.
6. Formula Checkpoint — The Present Coordinates of Two Wings
Applying the core formula, Korea's chips and batteries currently stand at the formula's second stage.
Technological innovation has occurred. SK hynix's HBM. LG Energy Solution's U.S. and European factory networks. Samsung SDI's all-solid-state battery research. Concentration of capital has also occurred — Samsung's ₩450 trillion over five years, SK hynix's ₩600 trillion in Yongin, the government's ₩622 trillion semiconductor mega-cluster. The chaebol structure made that concentration possible.
But signs of the formula's third stage — social instability — are visible. Battery materials companies are recording sharp revenue declines, and the expansion of overseas factories is generating concerns about domestic hollowing-out. The structure in which the gains from the semiconductor boom concentrate in a small number of large conglomerates resembles the dual economy of Taiwan examined in Chapter 4. Just as TSMC engineers said they could not buy a home even while making the world's most important chips, in Korea too there is a growing divergence between the semiconductor boom and lived economic experience.
Technological innovation to concentration of capital, concentration of capital to social tension. Can the formula's fourth stage — institutional redesign — follow? Expanding the tax credit under the K-Chips Act to 25 percent, and being the second country in the world to enact an AI Basic Act (인공지능 기본법) — these are the beginning of an institutional response. But in Chapter 10 we saw Japan spend thirty years failing at institutional redesign, blocked by the social inertia of lifetime employment and consensus culture. The fact that Korea's chaebol-centered structure enables fast decision-making is a strength — but the fact that the direction of innovation is left to the judgment of a few is simultaneously a vulnerability. That must be faced directly.
Seven mirrors have been examined. In Chapter 4 we saw Taiwan's TSMC hold the world with a single blade called the silicon shield. In Chapter 7 we saw ASML create a bottleneck through monopoly. In Chapter 8 we saw Indonesia use nickel as a resource ladder. In Chapter 10 we saw Japan lose thirty years before finally betting again on semiconductors.
Where is Korea's indispensability? One sharp blade called HBM, and batteries — still reforming their lines. But the tools that make the blade — equipment and materials — belong to others. And the people to hold the blade — engineers — are dwindling. Both wings are needed. With only one, there is no holding a position in the between.
Are chips and batteries Korea's shield or Korea's shackle? The precise answer is both, simultaneously. The higher the irreplaceability — as with SK hynix's HBM — the stronger the shield effect. The more replaceable — as with Samsung Foundry — the greater the shackle risk. In domains under pursuit, like batteries, the shield thins. And the hands to hold the shield are emptying.
What determines the strength of the two wings is not technology alone. Institutions, people, and the diplomatic balance between great powers — the next chapter examines the strategy of the in-between within that space.