Posts
Part 1: The Ghost Cities | Part 2: Singapore's Farmland Empire | PART 3: SEMICONDUCTOR FORTRESS (The TSMC/Samsung Monopoly) | Part 4: Belt & Road | Part 5: Tax Haven Dual System | Part 6: Japan's Stealth Military | Part 7: South Korea's Chaebols | Part 8: Taiwan's Silicon Shield | Part 9: Rare Earth Monopoly | Part 10: The Reckoning
Part 3: Semiconductor Fortress
90% of Advanced Chips Made in Taiwan—That's Not a Supply Chain, That's a Geopolitical Weapon
The Monopoly: How We Got Here
In 1987, Morris Chang founded TSMC with a revolutionary business model: the "pure-play foundry." Instead of designing and manufacturing chips (like Intel), TSMC would only manufacture—producing chips designed by other companies (Apple, Nvidia, AMD, Qualcomm).
The thesis: Semiconductor manufacturing requires such massive capital investment (billions per fab, updated every 2-3 years) that most companies can't afford to both design and manufacture competitively. Specialize in manufacturing, achieve economies of scale, and become the best in the world.
It worked. Over 35 years, TSMC built an insurmountable lead:
The Technology Gap (2025):
- TSMC: Mass-producing 3nm chips, developing 2nm (expected 2025-2026), researching 1nm
- Samsung: Mass-producing 3nm, but yields are lower (more defects), customers prefer TSMC
- Intel: Just reaching 7nm equivalent (rebranded as "Intel 4"), 3-5 years behind TSMC
- China (SMIC): Stuck at 7nm (achieved via smuggled equipment), cannot access advanced lithography tools due to US/Dutch export controls
Smaller nanometers = more transistors per chip = more computing power and energy efficiency. The difference between 7nm and 3nm is roughly 2x performance improvement and 30-40% energy reduction. For AI, mobile devices, and datacenters, this gap is existential.
The Market Share (2024-2025):
- TSMC: 62% of global foundry market (all chip types), ~90% of advanced chips (7nm and below)
- Samsung: 12% of foundry market, ~8% of advanced chips
- Intel: 0% of foundry market (Intel Foundry Services launched 2021, no major customers yet)
- China (SMIC, Hua Hong): 6% of foundry market, ~0% of advanced chips
- Others: 20% (mostly legacy/older chips: 14nm and above)
ADVANCED CHIPS (7nm AND BELOW) PRODUCTION:
• TSMC (Taiwan): ~90% of global capacity
• Samsung (South Korea): ~8%
• Intel (United States): ~0%
• China (SMIC): ~0%
• Rest of world: ~2% (mostly R&D, not mass production)
TECHNOLOGY LEADERSHIP (NODE SIZE):
• TSMC: 3nm production, 2nm development
• Samsung: 3nm production (lower yields)
• Intel: 7nm production ("Intel 4")
• SMIC: 7nm (limited, sanctions-constrained)
• Technology gap: TSMC leads by 2-3 generations
CRITICAL CUSTOMERS (DEPENDENT ON TSMC):
• Apple: 100% of iPhone/Mac chips (A17, M3, M4)
• Nvidia: 100% of advanced GPUs (H100, B100 AI chips)
• AMD: 100% of Ryzen CPUs, Radeon GPUs
• Qualcomm: 100% of Snapdragon mobile chips
• Google: Tensor chips (Pixel phones)
• Amazon: Custom AWS datacenter chips
• Tesla: Full Self-Driving chips
DEFENSE DEPENDENCE:
US military weapons systems using TSMC chips:
• F-35 fighter jet: 100+ chips per aircraft
• Patriot missiles, Javelin missiles
• Aegis combat systems (Navy)
• Satellite systems, communications
• ESTIMATED: 60-80% of US defense chips from TSMC
CONCENTRATION RISK:
Single point of failure for global technology economy.
Why the Lead Is Insurmountable (In the Short Term)
Building competitive semiconductor manufacturing requires three components, each with 10+ year development timelines:
Component 1: Extreme Ultraviolet (EUV) Lithography Machines
To make chips below 7nm, you need EUV lithography—machines that use extreme ultraviolet light to etch microscopic circuit patterns onto silicon wafers.
The monopoly within the monopoly: Only one company in the world makes EUV machines: ASML (Netherlands). Each machine costs $150-350 million and takes 18 months to build. ASML produces ~60-80 machines per year globally.
TSMC owns approximately 70% of all EUV machines ever produced. Samsung owns ~20%. Intel is buying machines now but won't have enough to compete until 2027-2030.
China is blocked from buying EUV machines due to US/Dutch export controls (agreed 2019-2023). Without EUV, China cannot manufacture advanced chips—period. They're trying to develop domestic EUV technology but are estimated to be 10-15 years behind ASML.
Component 2: Process Technology and Know-How
Owning EUV machines isn't enough—you need to know how to use them. Semiconductor manufacturing is extraordinarily complex:
- 400-1,000 individual processing steps per chip
- Sub-nanometer precision (arranging atoms with near-perfect accuracy)
- Yield management (minimizing defects; a 70% yield is good, 90% is world-class)
- Process optimization (temperature, timing, chemical ratios, pressure—thousands of variables)
TSMC has 35+ years of accumulated knowledge. Their engineers have run billions of wafers through hundreds of process iterations. This knowledge isn't written down—it exists in engineers' expertise, institutional memory, and iterative refinement.
You can't buy this. You can't steal it (not entirely—you'd need to steal thousands of engineers). You have to build it over decades.
Component 3: Ecosystem and Supply Chain
Semiconductor manufacturing requires 500-1,000 specialized suppliers:
- Silicon wafer suppliers (ultra-pure crystalline silicon)
- Chemical suppliers (photoresists, etching gases, cleaning solutions—hundreds of chemicals)
- Equipment suppliers (deposition machines, etching tools, inspection systems)
- Packaging and testing facilities
- Logistics networks (cleanroom shipping, temperature control, contamination prevention)
This ecosystem took 40 years to build and is concentrated in Taiwan, South Korea, Japan, and the Netherlands. China is trying to replicate it domestically but faces two problems:
- Sanctions: US export controls block China from accessing critical equipment and materials
- Time: Even without sanctions, building a complete ecosystem takes 15-20 years
The Silicon Shield: Geopolitical Leverage Through Technological Monopoly
Taiwan is 110 miles from mainland China. Taiwan's population: 23 million. China's population: 1.4 billion. China's military: 2+ million active personnel. Taiwan's military: 170,000 active, 1.5 million reserves.
On paper, Taiwan cannot defend itself against Chinese invasion. But Taiwan has something better than a military deterrent: economic deterrent.
The Logic:
If China invades Taiwan, one of three things happens to TSMC fabs:
- Destroyed in fighting: Fabs are precision facilities requiring perfect conditions (vibration-free, contamination-free, stable power). Artillery, missiles, or bombing would render them inoperable for years.
- Sabotaged by Taiwan: Credible reports suggest Taiwan has plans to destroy TSMC fabs if invasion is imminent (deny China the prize). TSMC's founder Morris Chang has publicly acknowledged this scenario.
- Seized but unusable: Even if fabs survive intact, they require continuous supply of materials, components, and software updates from Western suppliers (ASML, US chemical companies, Japanese material suppliers). If the West imposes sanctions (which they would during a Taiwan invasion), fabs go offline within weeks.
In all three scenarios, global semiconductor supply collapses. The economic consequences:
- United States: AI development stops, defense manufacturing cripples, datacenter expansion halts—estimated $500B-1T loss first year
- China: Domestic tech industry (Huawei, Xiaomi, etc.) loses chip access, economic growth craters—estimated $300-500B loss first year
- Europe: Auto industry shuts down (modern cars need 1,000-3,000 chips), industrial base freezes—estimated $200-400B loss
- Global: Supply chain paralysis, consumer electronics stop, communications infrastructure can't expand—estimated $1-2T first year, $5-10T over five years
This is MAD (Mutually Assured Destruction) for the global economy. China invading Taiwan is economic suicide—not just for Taiwan or China, but for everyone.
The Shield Works (So Far):
China has not invaded Taiwan despite having military capability since the 1990s. Every Chinese military strategist understands the TSMC problem: invading Taiwan gives you a destroyed island and triggers global economic collapse, including China's own economy (which depends on semiconductor imports for manufacturing).
This is why TSMC is called the "Silicon Shield"—it protects Taiwan not through military deterrence but through economic interdependence so extreme that war becomes irrational.
SCENARIO 1: INVASION + FAB DESTRUCTION
• Fabs destroyed by fighting or sabotage
• Global advanced chip production: -90%
• Economic impact (Year 1): $1-2T USD global
• Recovery timeline: 5-10 years minimum
• China gains: Nothing (destroyed assets)
• Result: Pyrrhic victory, global depression
SCENARIO 2: INVASION + FABS CAPTURED INTACT
• China seizes fabs without damage
• Western sanctions immediate:
- ASML stops software updates (machines offline in weeks)
- Chemical suppliers stop shipments (production halts)
- Equipment maintenance ends (fabs degrade in months)
• Timeline to fab failure: 6-12 weeks
• Global impact: Same as Scenario 1 (delayed by weeks)
• China gains: Nothing (fabs become useless)
SCENARIO 3: NO INVASION (STATUS QUO)
• Taiwan maintains independence (de facto)
• TSMC continues operating
• Global economy stable
• China continues importing TSMC chips
• China gains: Continued economic growth
• US/Europe gain: Continued tech advancement
• Result: Everyone wins except Chinese nationalists
STRATEGIC CONCLUSION:
Invasion is economically irrational.
The Silicon Shield is the strongest deterrent Taiwan has—
stronger than its military, stronger than US security guarantees.
As long as TSMC's monopoly persists, Taiwan is safe.
The American Response: CHIPS Act and the Great Reshoring (That Isn't)
In August 2022, the US passed the CHIPS and Science Act—$52 billion in subsidies to bring semiconductor manufacturing back to America.
Major commitments secured:
- TSMC Arizona: $40 billion investment, two fabs (eventually three), producing 5nm and 3nm chips, operational 2025-2027
- Samsung Texas: $17 billion investment, advanced logic fab, operational 2024-2025
- Intel Ohio: $20 billion investment (potentially $100B over decade), two fabs, operational 2025-2027
- Micron New York: $20 billion, memory chip production (not logic chips), operational 2025+
On the surface, this looks like successful reshoring—$100+ billion in semiconductor manufacturing returning to the US. But look closer:
The Reality: Old Tech, Not Cutting Edge
TSMC Arizona Fab 1 will produce 5nm chips (2024-2025). But TSMC Taiwan is already mass-producing 3nm and developing 2nm. By the time Arizona comes online, it will be producing chips that are two generations behind Taiwan.
TSMC Arizona Fab 2 will eventually produce 3nm (2027-2028). But by then, Taiwan will be producing 2nm and developing 1nm (sub-nanometer). Arizona will always be 2-3 years behind.
Why? Because TSMC is keeping the cutting-edge in Taiwan. They'll move older technology to Arizona once it's no longer their competitive advantage. This is strategic—it captures US subsidies and provides insurance (if Taiwan fabs go offline, Arizona can ramp up older-generation production) while maintaining Taiwan's technological monopoly.
The Cost Differential: American Manufacturing Is Insanely Expensive
TSMC's Arizona fab costs approximately 3-4x more to build and operate than an equivalent Taiwan fab:
- Construction costs: Higher labor costs, stricter regulations, less experienced contractors
- Operating costs: US semiconductor engineers earn $150-300k; Taiwan engineers earn $50-100k (TSMC Taiwan engineers, not adjusted for cost of living)
- Water/power: Arizona water costs more than Taiwan (desert vs. island); power grid less reliable
- Supply chain: Most materials still ship from Asia; adds logistics costs and time delays
Even with $6.6 billion in CHIPS Act subsidies (awarded to TSMC Arizona in 2024), the fab is marginally profitable at best. TSMC is building Arizona for strategic reasons (geopolitical hedge, customer reassurance, subsidy capture), not because it's economically optimal.
The Workforce Problem: Not Enough Engineers
TSMC Arizona needs approximately 10,000-12,000 employees when fully operational. The US produces ~8,000 electrical/semiconductor engineering graduates per year (total, across all companies). TSMC is importing hundreds of engineers from Taiwan to Arizona—because the domestic workforce doesn't exist.
Intel and Samsung face the same problem. The US dismantled its semiconductor manufacturing base over 20-30 years. You can't rebuild workforce expertise in 5 years.
US GOVERNMENT INVESTMENT:
• CHIPS Act total: $52B in subsidies
• Manufacturing incentives: $39B
• R&D: $13B
• Committed to date (Jan 2025): ~$36B
PRIVATE SECTOR COMMITMENTS:
• TSMC Arizona: $40B (Fabs 1, 2, 3)
• Intel Ohio: $20B+ (expandable to $100B)
• Samsung Texas: $17B
• Micron New York: $20B
• Others (Wolfspeed, GlobalFoundries, etc.): $20B+
• TOTAL: ~$120B in announced investments
PRODUCTION CAPACITY (WHEN OPERATIONAL):
• TSMC Arizona: 5nm (2025), 3nm (2027-28)
- Capacity: 20K wafers/month per fab = 60K total
- Taiwan capacity: 140K wafers/month (3nm/2nm)
- Arizona = ~30% of Taiwan capacity (older nodes)
• Intel Ohio: 7nm equivalent ("Intel 4")
- Intel still 2-3 years behind TSMC technologically
• Samsung Texas: 4nm/3nm
- Lower yields than TSMC Taiwan
THE TECHNOLOGY GAP PERSISTS:
By 2028, projected leadership:
• TSMC Taiwan: 2nm production, 1nm development
• TSMC Arizona: 3nm production (2 generations behind)
• Intel Ohio: 5nm equivalent (3 generations behind)
• Samsung Texas: 3nm (2 generations behind)
ACTUAL OUTCOME:
US gains 20-30% of global advanced chip capacity (older tech).
TSMC Taiwan retains 60-70% (cutting-edge tech).
The monopoly persists—just with American insurance policy.
China's Desperate Attempt: $150 Billion and Counting
China recognizes semiconductor dependence as an existential vulnerability. Since 2014, China has invested an estimated $150-200 billion in domestic semiconductor development through the "Big Fund" (National Integrated Circuit Industry Investment Fund) and provincial subsidies.
What They've Achieved:
- SMIC (Semiconductor Manufacturing International Corporation): Can produce 7nm chips in limited quantities using smuggled/repurposed equipment
- 14nm and older nodes: Competitive production for legacy chips (not cutting-edge but useful for appliances, industrial equipment, older devices)
- Domestic supply chain: Partial self-sufficiency in materials, chemicals, older equipment
What They Can't Achieve (Yet):
- EUV lithography: Can't buy from ASML (sanctions), can't build domestically (10-15 years away)
- 5nm and below: Without EUV, physically impossible to manufacture at scale
- Yield rates: Even at 7nm, SMIC's yields are reportedly 50-60% (vs. TSMC's 90%+), making production economically marginal
The Sanctions Wall:
US export controls (2019-2023) and Dutch cooperation (2023-2024) have created a technology blockade:
- October 2022: US bans sale of advanced chips and chipmaking equipment to China
- January 2023: Netherlands (ASML's home) agrees to restrict EUV exports to China
- August 2023: US expands controls to include chip design software, cloud computing access for AI training
- December 2023: US pressures Japan and South Korea to limit equipment and material sales
These sanctions are devastating because semiconductor manufacturing is a global ecosystem—you can't do it alone. China can't build EUV machines without Dutch optics, can't refine photoresists without Japanese chemicals, can't design chips without US software (Cadence, Synopsys tools).
The Smuggling Response:
China is attempting to circumvent sanctions through:
- Shell companies: Buying equipment through third countries (Malaysia, Singapore, Vietnam) and rerouting to China
- Talent recruitment: Hiring TSMC/Samsung engineers (reportedly offering 3-5x salary + housing + benefits)
- Espionage: Industrial espionage targeting TSMC, Samsung, and US firms (documented cases by FBI/CIA)
- Reverse engineering: Buying advanced chips and attempting to reverse-engineer manufacturing processes
This can close some gaps but can't overcome the fundamental barrier: without EUV machines, China cannot manufacture cutting-edge chips. And building domestic EUV capability requires another decade minimum.
Samsung's Dilemma: Stuck in Second Place
Samsung is the world's second-largest advanced chipmaker—and distant second. Samsung's challenges:
Problem 1: Yield Rates
Samsung's 3nm yields are reportedly 60-70% vs. TSMC's 80-90%. This means Samsung produces more defective chips, raising costs and making customers hesitant. Apple, Nvidia, and AMD all chose TSMC over Samsung for their most advanced chips despite Samsung's aggressive pricing.
Problem 2: Technology Lag
Samsung is 6-12 months behind TSMC at every node. When TSMC launches 3nm, Samsung is finalizing 3nm. When TSMC moves to 2nm, Samsung is ramping 3nm. This perpetual lag means Samsung is always competing for yesterday's technology.
Problem 3: Customer Trust
Samsung's 2018-2020 production issues (yield problems at 7nm, delays at 5nm) damaged customer relationships. Once Apple/Nvidia/AMD shift production to TSMC, they're reluctant to return—switching foundries is expensive and risky.
The Strategic Paradox:
Samsung wants to compete with TSMC but also benefits from TSMC's monopoly. If TSMC's Taiwan fabs go offline (invasion, earthquake, etc.), Samsung becomes the world's only advanced chipmaker—instant monopoly pricing power.
This creates perverse incentives: Samsung invests enough to stay competitive (second place) but perhaps not enough to truly challenge TSMC (first place), because TSMC's geopolitical risk is Samsung's greatest strategic asset.
The Geopolitical Trap: America's Dependence on Taiwan
The US faces an extraordinary strategic vulnerability: American military supremacy depends on semiconductors manufactured by a company in a territory that the US does not control and may not be able to defend.
The Defense Dependence:
Modern US weapons systems are essentially flying/floating/rolling computers:
- F-35 fighter jet: 100+ advanced chips per aircraft (radar, avionics, weapons systems, communications)
- Aegis combat system (Navy): Requires continuous chip supply for maintenance and upgrades
- Patriot/THAAD missile systems: Advanced processors for guidance and tracking
- Satellite systems: Communication, GPS, reconnaissance—all require cutting-edge chips
- Unmanned systems: Drones, autonomous vehicles, robotic systems
Estimated 60-80% of chips used in these systems come from TSMC Taiwan. Intel doesn't make equivalent chips. Samsung could substitute for some, but not at current US defense procurement scale.
The War Scenario Problem:
If the US goes to war with China (most likely scenario: over Taiwan), the US immediately loses access to TSMC chips. This means:
- Cannot manufacture new weapons systems mid-conflict
- Cannot replace damaged/destroyed systems
- Cannot upgrade existing systems
- Cannot expand production (every war requires production surge)
Meanwhile, China faces the same problem (also depends on TSMC). So both sides fight with existing stockpiles, unable to produce replacements. Whoever runs out of missiles/planes/ships first loses.
This is why semiconductor independence is now considered a national security imperative—but achieving it requires 10-15 years and hundreds of billions in sustained investment.
DEFENSE SYSTEMS USING ADVANCED CHIPS:
• Aircraft: F-35, F-22, B-21 bomber, drones
• Naval: Aegis destroyers, Virginia-class subs, carriers
• Missiles: Patriot, THAAD, Tomahawk, Javelin, HIMARS
• Space: Satellites, GPS, reconnaissance systems
• Communications: Encrypted systems, data links
• Cyber: NSA systems, offensive cyber capabilities
CHIP SOURCING (ESTIMATED):
• TSMC (Taiwan): 60-70% of advanced defense chips
• Samsung (South Korea): 10-15%
• Intel (United States): 10-15%
• Other: 5-10%
WAR SCENARIO: US-CHINA CONFLICT OVER TAIWAN
Day 1: TSMC fabs destroyed or inaccessible
Week 1: Existing chip stockpiles begin depletion
Month 3: Cannot replace destroyed systems
Month 6: Weapons production capacity at 30-40%
Year 1: Defense industrial base crippled
The US cannot fight a prolonged war with China
without access to Taiwan's semiconductor production.
This is the strategic trap.
The Long Game: TSMC's Strategy Is Working Perfectly
Step back and look at what TSMC has accomplished:
1. Captured US Subsidies ($6.6B) While Maintaining Monopoly
TSMC is getting American taxpayers to fund their expansion while keeping cutting-edge production in Taiwan. Arizona gets older-generation chips; Taiwan keeps 2nm/1nm development. TSMC gets both subsidy money AND maintained competitive advantage.
2. Diversified Geopolitical Risk Without Sacrificing Position
If Taiwan fabs go offline, TSMC has backup capacity in Arizona and Japan (TSMC is also building fabs in Kumamoto, Japan). These fabs can't match Taiwan's output, but they preserve TSMC's customer relationships and partial production capability.
Meanwhile, Taiwan remains essential—the insurance policy (Arizona/Japan) doesn't eliminate dependence, it just reduces catastrophic risk from 100% loss to 60-70% loss.
3. Locked In Customers for Next Decade
Apple, Nvidia, AMD, Qualcomm have all signed long-term supply agreements with TSMC running through 2030+. These contracts include penalties for switching foundries, volume commitments, and technology roadmap coordination.
Even if Samsung or Intel achieve technological parity (unlikely before 2028-2030), customers can't easily switch—they're contractually locked to TSMC.
4. Maintained Taiwan's Geopolitical Leverage
The Silicon Shield is stronger than ever. As AI and advanced computing become more critical (ChatGPT, autonomous vehicles, advanced weapons), dependence on TSMC intensifies. China's incentive to invade decreases as the economic cost of losing TSMC access increases.
TSMC isn't just a company—it's Taiwan's primary deterrent against Chinese invasion.
The 2030 Question: Can Intel Come Back?
Intel, once the undisputed semiconductor leader, has fallen behind TSMC by 3-5 years. CEO Pat Gelsinger (hired 2021) has an ambitious plan: "IDM 2.0"—regain technology leadership by 2025 and compete with TSMC as both manufacturer and designer.
The Plan:
- Intel 4 (7nm equivalent): In production 2023-2024
- Intel 3 (enhanced 7nm): Production 2024
- Intel 20A (2nm equivalent): Target 2024-2025
- Intel 18A (1.8nm equivalent): Target 2025-2026, aimed at matching TSMC's 2nm
If Intel hits these targets, they'd close the gap to 1-2 years behind TSMC by 2026-2027—competitive enough to win back customers.
The Problems:
1. Intel has missed technology targets for 10 consecutive years. Every node since 14nm (2014) has been delayed 1-3 years. Why should we believe this time is different?
2. Intel's foundry business (Intel Foundry Services) has no major customers. Qualcomm signed a "collaboration agreement" but hasn't committed to high-volume production. Without customers, Intel can't achieve the scale needed to compete with TSMC economically.
3. Intel is trying to do two things simultaneously: catch up technologically AND build a foundry business. TSMC only does manufacturing—they've specialized for 35 years. Intel is attempting to match TSMC while also designing their own chips (Core, Xeon). This divided focus is a structural disadvantage.
The Realistic Outcome (2030):
Intel probably achieves parity at mature nodes (7nm, 5nm) by 2027-2028 and wins some foundry customers for legacy production. But TSMC maintains leadership at cutting-edge nodes (2nm, 1nm, beyond). Intel becomes a viable alternative for some applications—but not the leader.
The monopoly weakens to oligopoly (TSMC 60%, Samsung 20%, Intel 15%, others 5%)—better than today's 90% concentration, but still far from distributed manufacturing.
The Ultimate Endgame: What Happens If Taiwan Is Invaded?
Let's war-game the scenarios:
Scenario A: Invasion + Fab Destruction (Most Likely)
Timeline:
- Day 1: China invades, TSMC fabs destroyed by fighting or sabotage
- Week 1: Global chip shortage panic, semiconductor stocks crash 40-60%, tech stocks fall 30-50%
- Month 1: Existing chip inventories begin depleting; TSMC Arizona/Japan ramp to maximum output (insufficient to meet demand)
- Month 3: Apple delays iPhone launch; automotive production down 40%; datacenter expansion halted
- Month 6: Consumer electronics shortages widespread; AI development slows; defense production compromised
- Year 1: Estimated $1-2T global economic loss; recession in US, Europe, Japan; China's economy contracts 5-8%
- Year 2-5: Samsung expands production, Intel accelerates (both receiving emergency government funding); gradual recovery but years of lost technological progress
Geopolitical outcome: US/Europe sanction China (total trade embargo); China faces international isolation; Taiwan destroyed but China gains no functioning fabs; everyone loses.
Scenario B: Blockade Without Invasion (China's Smart Play)
Strategy: China doesn't invade—just blockades Taiwan, cutting off shipping/air access. No shots fired; TSMC fabs remain intact but can't ship products.
Timeline:
- Week 1: Chip shipments from Taiwan stop; global shortage begins
- Month 1: China offers deal: "Remove blockade if US/Taiwan accept reunification terms"
- Month 2-3: US/allies face choice: military confrontation to break blockade (risks war) or negotiate (accepts Chinese terms)
This is China's leverage play: Use Taiwan's semiconductor monopoly as hostage without destroying the asset. Force US/Taiwan to negotiate from weakness.
Counter: US/Taiwan could airlift critical materials to TSMC and evacuate chips by air (expensive but possible for limited volumes). Buys time but doesn't solve underlying problem.
Scenario C: Cyber Attack on TSMC (21st Century Warfare)
Strategy: China launches sophisticated cyber attack targeting TSMC's manufacturing control systems. No physical destruction, but fabs go offline as software/equipment becomes inoperable.
Advantage: Deniability (hard to prove attribution definitively); reversible (if China achieves political goals, they can "help" restore operations); doesn't destroy physical assets.
Timeline:
- Day 1: TSMC manufacturing systems compromised; production halts
- Week 1-2: TSMC attempts recovery; unclear if systems can be restored or need complete rebuild
- Month 1: If recovery fails, same cascade as Scenario A but slower onset
Defense: TSMC reportedly has air-gapped backup systems and extensive cybersecurity (Taiwanese military cooperation). But no system is invulnerable.
IMMEDIATE IMPACT (MONTH 1):
• TSMC production: -90% (Arizona/Japan at 10% of Taiwan capacity)
• Global advanced chip supply: -85%
• Semiconductor stock index: -40 to -60%
• Tech company market cap loss: $2-3T
• Consumer panic buying: Electronics, chips
QUARTER 1 (MONTHS 1-3):
• Apple: iPhone production -70%, Mac -60%
• Automotive: Production -35% (chip shortage)
• Datacenters: Expansion halted, maintenance threatened
• AI development: New chip deployment stops
• Gaming: PS5/Xbox production -80%
• GDP impact: US -1.5%, China -2%, Global -0.8%
YEAR 1:
• Cumulative GDP loss: US $600-800B, China $400-600B,
Europe $300-400B, Global $1.5-2T
• Unemployment: Tech sector +15-20%
• Stock market: S&P 500 -25%, NASDAQ -35%
• Recession: Global, -2% to -3% GDP
YEARS 2-5 (RECOVERY PHASE):
• Samsung capacity expansion: +50% (govt subsidies)
• Intel emergency acceleration: Achieves 3nm by 2026
• New fabs: US, Europe, Japan (emergency construction)
• Total new capacity investment: $500B+ globally
• Timeline to pre-crisis capacity: 4-6 years
• Lost technological progress: 3-5 years
(Innovation slows without cutting-edge chips)
LONG-TERM OUTCOME:
Geographic diversification achieved through crisis.
But at catastrophic cost: $5-10T total economic damage,
5-10 years lost progress, geopolitical chaos.
China gains destroyed Taiwan, loses global economy access.
Nobody wins.
Why TSMC's Monopoly Persists: It's Not an Accident
The semiconductor concentration in Taiwan isn't a market failure—it's the result of decades of deliberate strategy by Taiwan, TSMC, and indirectly by American policy.
Taiwan's Strategy: Economic Defense
Taiwan recognized in the 1980s-1990s that it couldn't defend itself militarily against China. Solution: make Taiwan economically indispensable. TSMC was explicitly supported by Taiwanese government policy—subsidies, infrastructure investment, education pipeline, regulatory support—to create a "too important to invade" asset.
It worked. TSMC is Taiwan's primary defense.
TSMC's Strategy: Maintain Lead Through Continuous Investment
TSMC spends $30-40 billion annually on R&D and capital expenditures—more than Intel and Samsung combined in foundry-specific investment. This relentless spending maintains their 2-3 year technology lead.
Morris Chang's insight: In semiconductors, second place is first loser. If you're not the technology leader, customers choose you only for price. TSMC would rather spend to maintain leadership than compete on price.
America's Complicity: Outsourcing Was Cheaper
US fabless chip companies (Apple, Nvidia, AMD, Qualcomm) chose TSMC because it was economically optimal:
- Better technology than Intel
- Lower costs than building own fabs
- Flexibility (only pay for capacity you use)
- No capital risk (TSMC absorbs fab construction costs)
American companies maximized shareholder returns by outsourcing manufacturing. This was rational economically—but created strategic dependence.
The US government allowed this because free market ideology dominated policy from 1990s-2010s. Semiconductors were treated as commercial products, not strategic assets.
That changed only recently (2018-2022) as China threat became undeniable and COVID exposed supply chain fragility.
The Future: Slow Diversification, Persistent Concentration
Over the next 10-15 years, semiconductor manufacturing will slowly diversify:
- 2025-2027: TSMC Arizona/Japan operational; Intel achieves 5nm; Samsung expands
- 2028-2030: US/Europe reach 20-30% of global advanced chip capacity (vs. ~5% today)
- 2030-2035: Continued expansion; Taiwan's share drops from 90% to 60%
- 2035+: Possible equilibrium: Taiwan 50%, US 20%, South Korea 15%, Europe 10%, Japan 5%
This is better than today's concentration—but still leaves Taiwan as critical node. Complete diversification would require 20-30 years and $1+ trillion in sustained investment.
The question: Will China wait that long?
The Semiconductor Fortress Holds
TSMC's monopoly isn't breaking anytime soon. The CHIPS Act will reduce catastrophic risk, but won't eliminate dependence. Taiwan remains the world's semiconductor fortress—and everyone knows it.
China knows invading Taiwan triggers economic MAD. The US knows defending Taiwan is existential for technological supremacy. Taiwan knows TSMC is their shield.
As long as this equilibrium holds, Taiwan is safe. But equilibriums are fragile. One miscalculation—Chinese leadership believing they can seize fabs intact, American overconfidence in Arizona fabs, Taiwanese sabotage plans failing—and the entire global economy collapses.
That's not a supply chain. That's a geopolitical weapon.
And TSMC is holding it.
No comments:
Post a Comment