Most people never think about where things come from. The phone in your pocket, the medication in your cabinet, the aircraft carrying you across continents - they exist because somewhere on Earth, a building full of machines and humans performs a process so specialized that it cannot be easily replicated elsewhere. In some cases, it cannot be replicated anywhere else at all.
The modern global economy runs on concentrated expertise. Decades of specialization, massive capital investment, regulatory approval, and accumulated know-how have produced a small number of factories that effectively operate as global monopolies in their domain. When geopolitical analysts talk about supply chain risk, they are often talking about these specific facilities - the choke points of civilization.
This is their story.
~60%Global advanced chip share
$30B+Annual capex (TSMC total)
3nmProcess node (N3 family)
6,000+Employees per major fab
If there is one factory on Earth that the entire technology sector depends on, it is a collection of buildings in the Tainan Science Park operated by Taiwan Semiconductor Manufacturing Company. TSMC's advanced fabrication facilities - particularly those producing chips on its N3 and N2 process nodes - manufacture the most complex objects ever made by humans.
The chips produced here power Apple's iPhones and MacBooks, Nvidia's AI accelerators, AMD processors, Qualcomm modems, and virtually every other cutting-edge semiconductor that touches modern life. A single wafer of 300mm silicon processed through Fab 18 contains hundreds of billions of transistors - features measured in atoms - produced by machines so precise they use extreme ultraviolet light with a wavelength of 13.5 nanometers to print circuit patterns.
The barriers to replication are almost incomprehensible. Building a comparable fab requires $20-30 billion in capital investment, 3-5 years of construction, and - most critically - decades of process know-how accumulated through millions of micro-adjustments to chemistry, temperatures, and equipment settings. TSMC has been refining these processes since 1987. The tacit knowledge embedded in its workforce and production systems cannot be transferred with blueprints alone.
Intel, Samsung, and now dozens of government-backed initiatives are trying to close the gap. None have fully succeeded. TSMC's process technology lead - the precise ability to print smaller, denser, more power-efficient circuits than anyone else - remains its most formidable competitive moat, and the world's most important manufacturing advantage.
Why it matters
A sustained disruption to TSMC's advanced fabs - through natural disaster, military conflict, or political crisis - would halt production of virtually every premium consumer device, AI server, and advanced military system on Earth within months. The U.S. CHIPS Act, EU Chips Act, and Japanese semiconductor initiatives all exist primarily because governments recognized this single point of failure.
The Concentration Problem
TSMC's situation is an extreme version of a pattern that repeats across dozens of critical industries. The economics of advanced manufacturing reward specialization so heavily that global supply chains have naturally consolidated around a very small number of facilities. This creates enormous efficiency - and enormous fragility.
98.7M ft²Building volume (largest by volume)
~400Commercial aircraft assembled/yr
30,000+Workers on-site at peak
1968Year opened (initially for 747)
The Boeing Everett Manufacturing Facility in Washington State is the largest building in the world by volume - a single structure so vast that it generates its own weather patterns, with clouds forming near the ceiling on humid days. Inside this colossal space, Boeing assembles the 747, 767, 777, and 787 Dreamliner - the widebody jets that form the backbone of long-haul commercial aviation globally.
Understanding what happens here requires appreciating the complexity of a modern commercial aircraft. A Boeing 777 contains roughly six million individual parts sourced from 900 suppliers in more than 100 countries. The final assembly process at Everett takes those components - fuselage sections, wings, engines, avionics, landing gear, interior systems - and integrates them into a flying machine capable of carrying 400 passengers across 9,000 miles at 35,000 feet with a failure rate measured in fractions of a percent per million flights.
The intellectual property embedded in these aircraft - the aerodynamic designs, the structural engineering, the systems integration knowledge - has been accumulated over more than 50 years of widebody aircraft production. Airbus's equivalent facility in Toulouse represents the only other comparable capability on Earth. Together, these two plants effectively supply all of the world's commercial widebody aircraft.
Why it matters
Global air travel depends on a replacement cycle for aging aircraft. With a duopoly between Boeing and Airbus - and each company's key capabilities concentrated in one or two major assembly facilities - any sustained disruption to Everett directly constrains the world's ability to maintain and expand its commercial aviation fleet. This affects not just passengers but global freight, since approximately half of all air cargo travels in the bellies of passenger aircraft.
"The factory of the future will have only two employees: a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment."
- Warren Bennis (reframing automation's endpoint)
100%EUV lithography market share
~$380MPrice per EUV machine
457,000Parts per EUV machine
~40EUV units shipped per year
TSMC cannot make advanced chips without machines it buys from one company: ASML, headquartered in Veldhoven, a small city in the Netherlands. ASML holds a complete monopoly on extreme ultraviolet (EUV) lithography machines - the equipment that physically prints the nanoscale circuit patterns onto silicon wafers. Without these machines, advanced semiconductor fabrication below approximately 7nm is effectively impossible with current technology.
Each ASML EUV machine is itself a marvel of manufacturing. It contains a laser that fires 50,000 pulses per second at tiny tin droplets, creating a plasma that emits extreme ultraviolet light. That light is collected by mirrors polished to atomic smoothness, guided through a vacuum, and focused onto a silicon wafer coated with photosensitive material. The entire system sits on a vibration-isolated platform and maintains positional accuracy measured in single-digit nanometers.
ASML's EUV machines take over a year to build, require several Boeing 747 cargo flights to ship (they arrive in pieces), and months to install and calibrate. The company ships roughly 40 per year. Every one is spoken for before it is built. The waiting list for a new EUV machine is measured in years, not months.
The geopolitical significance of this single factory in the Netherlands cannot be overstated. The United States government pressured ASML and the Dutch government to restrict EUV exports to China in 2019 - a measure widely credited with being one of the most effective single acts of technology policy in recent decades. China's entire advanced semiconductor industry is bottlenecked by its inability to obtain these machines.
Why it matters
ASML is the quintessential example of a company that controls a tool without which entire industries cannot function. No EUV machines means no advanced chips. No advanced chips means no next-generation AI hardware, smartphones, or defense electronics. One factory in the Netherlands is, in a very real sense, the gatekeeper of the global technology frontier.
Beyond Silicon: The Other Irreplaceable Plants
The semiconductor story gets the most attention, but it is far from the only domain where the world's supply has quietly concentrated in a handful of factories. Pharmaceuticals, specialty chemicals, aerospace components, and rare earth processing each have their own equivalents - facilities whose disruption would cascade through industries and economies in ways that take years to correct.
~50%Global GLP-1 supply capacity
$9B+Expansion investment (2023-2026)
1876Year Lilly founded in Indianapolis
10,000+Indianapolis-area employees
The Eli Lilly manufacturing complex in Indianapolis sits at the center of what may be the most dramatic pharmaceutical supply crunch of the 21st century. As demand for GLP-1 receptor agonists - sold under brand names like Mounjaro and Zepbound for diabetes and obesity - exploded from 2022 onward, Lilly's Indianapolis facilities became the global choke point for one of the most-prescribed drug classes in history.
GLP-1 drugs are biologics: complex molecules produced through biological processes, not simple chemical synthesis. The manufacturing process involves growing engineered cells in large bioreactors, purifying the resulting proteins through multiple filtration and chromatography steps, formulating the final product, and filling millions of injector pens with precision dosing. Each of these steps requires validated equipment, qualified operators, and regulatory-approved processes - none of which can be improvised quickly.
When demand surged, the shortage was not a matter of will or capital. Lilly committed billions to expanding capacity. But biologic manufacturing facilities take 3-5 years to build, qualify, and receive FDA approval. The result was a global shortage of drugs that clinical evidence suggests could meaningfully reduce rates of obesity, Type 2 diabetes, and cardiovascular disease in hundreds of millions of people.
The Indianapolis complex - including new facilities in Branchburg, New Jersey and international expansions - represents not just a pharmaceutical supply story but a window into why advanced manufacturing capacity cannot simply be conjured on demand, regardless of available capital.
Why it matters
Pharmaceutical manufacturing is subject to strict regulatory oversight that makes capacity impossible to scale quickly. A handful of specialized facilities control the global supply of drugs that millions depend on daily. The GLP-1 shortage demonstrated, with unusual public visibility, how concentrated pharmaceutical manufacturing creates healthcare supply risks that do not respond to normal market signals.
~85%China's share of rare earth processing
17Elements in rare earth group
$5B+Global rare earth market value
1950sWhen Baotou mining began
The industrial city of Baotou in Inner Mongolia hosts the world's largest concentration of rare earth processing capacity. While the name "rare earth" is a misnomer - these 17 elements are not particularly rare in the Earth's crust - their economic extraction and processing is extraordinarily concentrated. China controls not just mining but the separation and refining infrastructure that converts raw ore into the purified oxides and metals that industry requires.
Rare earth elements are embedded in virtually every piece of high-technology equipment on Earth. Neodymium and dysprosium go into the permanent magnets that power electric vehicle motors and wind turbine generators. Cerium is used in catalytic converters and optical polishing. Lanthanum goes into camera lenses and night-vision goggles. Europium and terbium are used in phosphors for displays and lighting. Yttrium is critical for superconductors and certain specialized alloys.
The processing technology required to separate these elements from ore and from each other is chemically complex, capital-intensive, and environmentally challenging - it produces substantial toxic waste. China built its dominance in this sector over decades by accepting lower environmental standards and subsidizing the industry strategically. Western nations largely ceded the market by the 1990s. When China restricted rare earth exports in 2010 in a territorial dispute with Japan, prices spiked by 500-700%, demonstrating precisely how much leverage this concentration creates.
Efforts to build alternative processing capacity in Australia, Canada, the United States, and Europe have made progress since 2020, but building parallel infrastructure to the Baotou processing complex - with its decades of refinement and scale - remains a multi-decade project.
Why it matters
Electric vehicles, wind turbines, defense electronics, and advanced manufacturing equipment all depend on rare earth elements processed primarily in China. Energy transition targets worldwide - including the net-zero commitments of virtually every major economy - require massively scaling production of these technologies. The concentration of processing capacity in Baotou is therefore a constraint not just on manufacturing but on the pace of global decarbonization.
~33%Widebody engine market share
£2.1M+Cost per large turbofan engine
25,000+Parts per engine
1906Year Rolls-Royce founded
At the Rolls-Royce facility in Derby, workers assemble some of the most complex mechanical objects on Earth: large civil turbofan engines of the Trent family, which power long-haul aircraft including the Airbus A350, Boeing 787, and Airbus A380. Along with GE Aviation in Cincinnati and Pratt & Whitney in East Hartford, Connecticut, Derby is one of only three significant sources of the large turbofan engines that power global aviation.
A modern turbofan operates at conditions that push the boundaries of materials science. Turbine blades rotate at 12,000 RPM in gas streams exceeding 1,700 degrees Celsius - above the melting point of the nickel superalloys they are made from. They survive only because of internal air cooling channels so fine they must be cast using ceramic cores that are later dissolved away, and thermal barrier coatings applied with plasma spray technology. The aerodynamic profiles are computed using millions of simulated airflow iterations and tested in wind tunnels for thousands of hours before entering service.
Derby's role as a design and assembly hub is supported by a supply chain of several hundred specialized suppliers, many of them small and highly specialized. The titanium fan blades come from one or two specialized forges. The ceramic thermal barrier coatings are applied by a handful of facilities. The fuel system components, the electronic engine control systems, the bearing housings - each comes from suppliers with years of qualification behind them.
Why it matters
Aviation's recovery from the COVID-19 pandemic created an engine shortage that had airlines retiring older aircraft while waiting years for replacement powerplants. With only three credible manufacturers of large turbofan engines, any disruption - in Derby, Cincinnati, or East Hartford - directly constrains global aviation's ability to grow or replace aging fleets. The three-way oligopoly in this sector means there is minimal buffer in the system.
The Common Thread: Time, Not Money
Across every factory profiled here, one pattern emerges with uncomfortable clarity. The constraint on these facilities is not capital. Governments and corporations are willing to spend enormous sums to build alternative capacity when they feel the strategic risk. The constraint is time - and more specifically, the time required to accumulate know-how that cannot be purchased or downloaded.
The TSMC problem is not that building a comparable fab costs $20 billion. It is that even after spending $20 billion and waiting five years, the facility will produce chips with a higher defect rate, lower yield, and lower performance than TSMC's equivalent. The gap is in yield engineering - the micro-adjustments to hundreds of process parameters that TSMC has refined through decades of trial, error, and accumulated institutional knowledge.
The same dynamic applies to ASML's EUV machines, to Rolls-Royce's engine manufacturing, to rare earth processing, to pharmaceutical biologics. In each case, tacit knowledge - the kind embedded in experienced workers, refined procedures, and calibrated equipment - is more valuable than any specific asset and far harder to replicate.
2010
China's rare earth export restrictionsDispute with Japan triggers export controls, prices spike 500%+. Global manufacturers scramble to find alternatives.
2011
Thailand floods disrupt hard drive supplyFlooding of Western Digital and Seagate factories causes global PC hard drive shortage lasting 18 months. Prices double.
2021
Global semiconductor shortageCOVID-era factory disruptions and demand surge create automotive and electronics chip shortages. Estimated $210B in lost automotive revenue.
2022
U.S. semiconductor export controlsCommerce Department restricts advanced chip technology exports to China, sharpening focus on TSMC and ASML as strategic assets.
2023
GLP-1 drug shortage peaksDemand for weight-loss and diabetes drugs overwhelms Lilly and Novo Nordisk manufacturing capacity. Hundreds of thousands unable to fill prescriptions.
2024
ASML EUV export restrictions tightenNetherlands joins U.S.-led restrictions on advanced lithography equipment exports. China unable to import EUV machines.
What Governments Are Doing - and Why It Takes So Long
The strategic risk of concentrated manufacturing has not gone unnoticed. The past five years have seen an unprecedented wave of industrial policy aimed at building redundant capacity for critical goods - semiconductors above all, but also batteries, rare earth processing, and pharmaceutical manufacturing.
The United States passed the CHIPS and Science Act in 2022, committing $52 billion to semiconductor manufacturing incentives and research. The European Union launched its own EU Chips Act targeting 20% of global chip production by 2030. Japan committed $13 billion to attract TSMC to build a fab in Kumamoto - which opened in 2024 but produces chips on older process nodes, not the cutting-edge technology at the center of most strategic concern. Taiwan itself has pushed TSMC to diversify, though the company and many analysts argue that the most advanced processes will remain in Taiwan for the foreseeable future due to the ecosystem concentration there.
These are not trivial investments. The TSMC Arizona fabs, planned to eventually produce chips on advanced nodes, represent more than $65 billion in committed capital. TSMC itself has called it among the most challenging projects in its history - not because of construction, but because replicating the workforce culture, the supplier ecosystem, and the operational discipline that makes Taiwan's fabs world-class takes time that cannot be compressed by spending more money.
| Sector |
Key facility / location |
Concentration risk |
Replacement timeline |
| Advanced semiconductors |
TSMC, Taiwan |
Critical |
10-15 years for full alternative |
| EUV lithography |
ASML, Netherlands |
Critical |
15+ years (monopoly) |
| Rare earth processing |
Baotou, China |
Critical |
10-20 years to diversify |
| Widebody aircraft |
Boeing Everett / Airbus Toulouse |
High |
Duopoly - 20+ years to enter |
| Large turbofan engines |
Rolls-Royce Derby / GE Cincinnati |
High |
15+ years for new entrant |
| GLP-1 biologic drugs |
Lilly / Novo Nordisk |
High |
5-8 years to build and qualify capacity |
| EV battery cells |
CATL (China), Panasonic (Japan) |
Medium |
Improving - 5-7 years |
Context
The replacement timelines above assume adequate capital, political will, and no regulatory obstacles. In practice, environmental permitting, supply chain development for specialized inputs, and workforce training often extend these timelines further. The "Critical" ratings indicate sectors where there is currently no credible alternative to the dominant facility or facilities even with unlimited capital.
The Invisible Architecture of Modern Life
There is something philosophically interesting about how civilization's most important manufacturing facilities are almost entirely unknown to the people who depend on them. Billions of people carry TSMC-fabricated chips in their pockets without knowing where they came from or what would happen if that supply were interrupted. Millions fly on Rolls-Royce engines without thinking about the engineers in Derby who designed the ceramic thermal barrier coatings keeping those blades intact at temperatures beyond their own melting points.
This invisibility is a product of success. When supply chains work - when chips appear in devices, when aircraft arrive on schedule, when medications are available at pharmacies - there is no reason for most people to think about the extraordinarily specialized facilities making it possible. The fragility only becomes visible in moments of disruption: the chip shortage of 2021, the GLP-1 drug shortage of 2023, the moment China restricted rare earth exports in 2010.
Each disruption has, to its credit, prompted a response. Governments have invested more in industrial policy than at any time since the post-war era. Companies have spent billions trying to diversify supply chains and build redundancy. The semiconductor industry in particular has seen a wave of new fab construction globally that will meaningfully expand capacity in non-Taiwan locations over the coming decade.
But the structural reality remains: the most advanced manufacturing processes in the world are extraordinarily concentrated, extraordinarily difficult to replicate, and extraordinarily slow to scale. The factories profiled here are not just buildings. They are the accumulated result of decades of scientific progress, billions in investment, and thousands of human-years of specialized expertise. They are, in a very real sense, irreplaceable - at least on any timescale that matters to the people living now and depending on what they make.
Understanding their existence, their constraints, and their fragility is not an academic exercise. It is a prerequisite for thinking clearly about the technology that surrounds us, the supply chains that deliver it, and the geopolitical forces increasingly organizing themselves around the question of who controls these critical nodes of global production.
Also Worth Knowing
Several other facilities deserve mention in any serious accounting of critical global manufacturing. Corning's specialty glass plants in Harrodsburg, Kentucky produce Gorilla Glass and the precision optical fiber that carries the internet - Corning commands roughly 40% of the global fiber market. Air Liquide's industrial gas facilities produce the ultra-pure nitrogen, oxygen, and specialty gases without which semiconductor fabrication is impossible. Novo Nordisk's facilities in Kalundborg, Denmark operate the world's largest insulin production complex, producing roughly half of all insulin consumed globally. SK Hynix and Micron control the overwhelming majority of DRAM and NAND flash memory - the storage and working memory in every computing device on Earth.
Each of these represents the same pattern: extreme specialization, massive capital investment, decades of accumulated know-how, and a global economy that has come to depend on continuity of supply from a very small number of facilities it could not quickly replace.
The factories that hold the world together are, by and large, the ones nobody is talking about - until the day they can't.