Back to blog

The Semiconductor Wars: How Chips Became More Important Than Oil

ConvertAndEdit TeamJanuary 4, 202513 min read0 views
The Semiconductor Wars: How Chips Became More Important Than Oil
semiconductorstechnologygeopoliticsmanufacturingchips
Everything runs on chips. Your phone, car, refrigerator, military defense systems, AI models—everything. And 92% of the most advanced ones come from a single company on an island China claims as its own. This is the story of the most important war nobody talks about.

The Stakes: Why Chips Matter More Than Oil

The New Oil: Silicon

Global dependency comparison:

ResourceCan Survive WithoutTime to CollapseAlternative Available
OilDifficult but possible6 monthsSolar, nuclear, coal
FoodNo30 daysNone
WaterNo3 daysNone
Advanced ChipsNo90 daysNone at scale
What stops without chips:
- Every car manufactured (1,000+ chips each)
- Every smartphone, computer, server
- All cloud services (AWS, Google, Azure)
- Military equipment (F-35 has 3,000 chips)
- Medical devices
- Power grids
- Internet infrastructure

The Numbers That Define Power

Semiconductor industry 2025:
- Market size: $680 billion
- Growth rate: 12% annually
- Jobs dependent: 50 million globally
- R&D spending: $100 billion/year
- Economic multiplier: 5x (every $1 creates $5 GDP)

The Players: Who Controls the Future

TSMC: The Most Important Company You've Never Heard Of

Taiwan Semiconductor Manufacturing Company:

MetricTSMC DominanceNext CompetitorGap
3nm Production92%Samsung 8%11.5x
5nm Production85%Samsung 15%5.7x
Revenue$80BIntel $65B23% lead
Market Cap$550BIntel $190B3x
CustomersEveryoneLimitedMonopoly
Who depends on TSMC:
- Apple: 100% of A-series chips
- NVIDIA: 100% of AI chips
- AMD: 100% of CPUs/GPUs
- Qualcomm: 100% of Snapdragon
- Broadcom: Network chips
- Even Intel: Outsourcing advanced nodes

The terrifying reality: If TSMC stops, the global economy stops.

The Chip Manufacturing Hierarchy

Cutting Edge (3nm-5nm)

CompanyCountryCapabilityMarket ShareInvestment
TSMCTaiwan3nm leader85%$40B/year
SamsungSouth Korea3nm struggling12%$30B/year
IntelUSA5nm behind3%$25B/year

Mature Nodes (7nm-28nm)

CompanyFocusStrengthWeakness
SMICChina independenceGovernment backingSanctions limited
GlobalFoundriesSpecialty chipsAuto partnershipsNo leading edge
UMCCost efficiencyReliableNot advancing

The Design Giants

NVIDIA: The New King

The AI monopoly:

ProductMarket ShareRevenueGrowthMoat
AI Training95%$50B+200% YoYCUDA ecosystem
AI Inference80%$20B+150% YoYSoftware stack
Gaming GPU83%$15B+20% YoYBrand loyalty
Data Center75%$35B+180% YoYPerformance lead
Why NVIDIA wins:
- CUDA: 10-year software advantage
- H100/H200: 10x faster than competition
- First mover in AI
- $10B R&D budget
- Jensen Huang: Visionary CEO

AMD: The Comeback Kid

Lisa Su's transformation:

YearStock PriceMarket Capvs IntelKey Product
2014$4$3B1/50thStruggling
2017$14$13B1/10thRyzen launch
2020$90$110B1/2RDNA2, CDNA
2024$180$290B2x biggerMI300X AI chip
2025$210$340B2.5xZen 5, MI400
AMD's strategy:
- Chiplet architecture (revolutionary)
- TSMC partnership (access to best nodes)
- Attack all markets (CPU, GPU, AI)
- Undercut NVIDIA pricing
- Open-source software (ROCm)

Intel: The Fallen Giant

The decline timeline:

EraStatusMistakeConsequence
2010-2015DominantIgnored mobileLost to ARM
2016-2020Struggling10nm delaysLost to AMD
2021-2023CrisisFab problemsLost to TSMC
2024-2025RebuildingFoundry pivotUnknown
Pat Gelsinger's rescue plan:
- IDM 2.0: Become foundry
- 5 nodes in 4 years (impossible?)
- $100B US government support
- Intel 18A: Revolutionary or hype?
- Partner with TSMC (humiliating)

ARM: The Invisible Empire

ARM's dominance:
- 99% of smartphones
- 90% of tablets
- Growing in servers (AWS Graviton)
- Entering PCs (Apple M-series)
- Edge AI everywhere

The business model:
- Design, don't manufacture
- License to everyone
- Collect royalties forever
- $3B revenue, $150B market cap

The Technology Race

Moore's Law: Dead or Evolving?

Traditional Moore's Law: Transistors double every 2 years

Reality in 2025:

NodeTransistors/mm²YearCost per TransistorReal Improvement
10nm100M2017$0.01Baseline
7nm200M2019$0.0082x density
5nm300M2021$0.0091.5x density
3nm450M2023$0.0121.5x density
2nm600M2025$0.0181.3x density
1.4nm750M?2027$0.025?Diminishing
The new paradigm:
- More Moore: Shrinking continues
- More than Moore: 3D stacking
- Beyond Moore: New materials
- Around Moore: Chiplets

The Manufacturing Miracle

How Chips Are Made

The impossibility of chip making:

  1. Silicon wafer: 99.9999999% pure (9 nines)
  2. Photolithography: EUV machines ($380M each)
  3. Etching: Atomic-level precision
  4. Deposition: One atom at a time
  5. Testing: Billions of transistors checked
  6. Packaging: 3D assembly
  7. Yield: 90% for mature, 60% for cutting edge

Why it's so hard:
- Features smaller than COVID virus
- Cleanrooms 1000x cleaner than hospital
- 3,000 process steps
- 4 months start to finish
- Single dust particle = ruined chip

EUV: The $380 Million Bottleneck

ASML's monopoly:

MetricASMLCompetitionMoat
EUV Machines100%NoneAbsolute
DUV Advanced90%Nikon 10%Dominant
Revenue$30B-Growing
Order Backlog$40B-2 years
Price per Tool$380MN/AOnly option
The EUV miracle:
- 13.5nm wavelength light
- 100,000x brighter than sun
- Mirrors instead of lenses
- 250kW power consumption
- 1 year to build
- 40 trucks to ship
- 6 months to install

The Geopolitical Battleground

The US-China Tech War

US Strategy: Containment

Export controls escalation:

YearRestrictionTargetImpact
2019Huawei ban5G leadershipLost global share
2020SMIC restriction7nm capabilityFrozen at 14nm
2022AI chip banComputing powerCan't train GPT-4
2023Tool banAll advanced10-year setback
2024Cloud accessAI servicesIsolation
2025Talent restrictionEngineersBrain drain
The coalition:
- USA: Policy and design
- Netherlands: ASML equipment
- Japan: Materials and tools
- Taiwan: Manufacturing
- South Korea: Memory and fabs

China's Response: Self-Sufficiency

Made in China 2025 → 2035:

InvestmentAmountProgressChallenge
National Fund$150BDeployedCorruption
Local Funds$200B+ActiveWaste
R&D$50B/yearGrowingTalent shortage
Fab Construction30+ newBuildingEquipment lacking
Talent Program100K engineersTrainingExperience gap
Breakthrough attempts:
- SMEE: EUV alternative (failing)
- SMIC: 7nm without EUV (low yield)
- Huawei: Chip design (successful)
- Loongson: CPU architecture (limited)
- YMTC: Memory (competitive)

Taiwan: The World's Most Important Island

Why Taiwan matters:
- 92% of sub-7nm production
- 65% of all chip production
- Would take 10 years to replace
- China claims sovereignty
- US committed to defense(?)

The invasion scenario:
- Global chip production: -60%
- Economic damage: $10 trillion
- Supply recovery: 5-10 years
- Tech advancement: Stopped
- Winner: Nobody

The Chips Act: America's $280B Bet

US reshoring effort:

CompanyInvestmentLocationCapacityTimeline
TSMC Arizona$40BPhoenix20K wafers/month2025-2027
Samsung Texas$25BTaylor15K wafers/month2025
Intel Ohio$20BColumbus30K wafers/month2026
Micron NY$20BSyracuseMemory2025
GlobalFoundries$12BNew YorkExpansion2024
Challenges:
- Cost: 50% higher than Asia
- Talent: Shortage of engineers
- Culture: Different work ethic
- Supply chain: Still Asian
- Scale: 10% of global needs

The Memory Wars: The Other Battlefield

DRAM Oligopoly

CompanyCountryMarket ShareTechnologyRevenue
SamsungKorea45%DDR5, HBM3$45B
SK HynixKorea28%HBM leader$28B
MicronUSA22%Advanced nodes$22B
OthersChina5%Behind$5B

The HBM Revolution

High Bandwidth Memory for AI:
- 10x bandwidth of DDR5
- Critical for AI training
- $1000 per GB
- Sold out through 2026
- SK Hynix dominates

NAND Flash Consolidation

The survivors:

  1. Samsung: 35% share
  2. Kioxia (Toshiba): 20%
  3. Western Digital: 15%
  4. SK Hynix: 15%
  5. Micron: 10%
  6. YMTC (China): 5%

The Money: Following the Capital

Company Valuations

CompanyMarket CapP/E RatioGrowthMoat Rating
NVIDIA$1.5T65100% YoYExtreme
TSMC$550B2525% YoYExtreme
ASML$400B4530% YoYAbsolute
AMD$340B12040% YoYStrong
Broadcom$700B3520% YoYStrong
Intel$190BNegative-10% YoYWeakening
Qualcomm$180B2815% YoYStrong
ARM$150B20050% YoYGrowing

The CapEx Arms Race

Annual capital expenditure:

Company202320242025 (planned)Purpose
TSMC$36B$40B$44B2nm, 3nm expansion
Samsung$31B$35B$38BFoundry catch-up
Intel$21B$25B$28BUS fabs
SK Hynix$10B$14B$18BHBM expansion
SMIC$7B$9B$12BChina self-sufficiency

Future Technologies: Beyond Silicon

The Next Paradigm

Gate-All-Around (GAA)

The evolution:
- FinFET → GAA
- Better control
- Lower leakage
- Samsung first (3nm)
- TSMC following (2nm)

Chiplets: Lego for Chips

The revolution:

AdvantageImpactLeaderExample
Yield2x betterAMDZen 4
Cost40% lowerIntelMeteor Lake
FlexibilityMix and matchAppleM2 Ultra
PerformanceSpecializedNVIDIAGrace-Hopper

New Materials

Beyond silicon:
- Gallium Arsenide: 5x faster
- Graphene: 100x faster (theoretical)
- Carbon nanotubes: Lower power
- 2D materials: Atomic thickness

Quantum + Classical

Hybrid computing:
- Classical control
- Quantum acceleration
- Error correction
- 2030 timeline

Investment Strategies

The Picks and Shovels

Safer bets:

CategoryCompaniesWhyRisk
EquipmentASML, KLAC, AMATMonopoliesLow
FoundryTSMCNo alternativeMedium
EDA SoftwareSNPS, CDNSDesign necessityLow
MaterialsSUMCO, Shin-EtsuCritical suppliesLow

The Risky Plays

Higher reward, higher risk:

CompanyBull CaseBear CaseVerdict
IntelFoundry successExecution riskGamble
AMDAI competitionNVIDIA moatPromising
Chinese ChipsDomestic marketSanctionsAvoid
ARMEverything edgeValuationOverpriced

The Talent War

The Engineering Crisis

Global shortage:
- Need: 1 million engineers
- Available: 400,000
- Training time: 5-10 years
- Starting salary: $200K+

Where they go:

CompanyEngineersAvg CompensationPerks
NVIDIA26,000$500KStock appreciation
Apple15,000$450KProducts
Google12,000$420KResearch
Meta8,000$480KRemote
TSMC50,000$150KTaiwan cost of living

Scenarios: The Next 10 Years

Scenario 1: Status Quo Continues

Probability: 40% - TSMC maintains dominance
- China stays behind
- Gradual US reshoring
- Innovation continues

Scenario 2: China Breakthrough

Probability: 30% - Domestic EUV equivalent
- Leapfrog to new technology
- Global supply chains split
- Tech cold war intensifies

Scenario 3: Taiwan Crisis

Probability: 20% - Military conflict
- Global chip shortage
- Economic depression
- 10-year recovery

Scenario 4: Paradigm Shift

Probability: 10% - Quantum/optical computing
- Silicon obsolete
- New leaders emerge
- Current players disrupted

Survival Guide for Investors

What to Watch

Leading indicators:

  1. TSMC order book
  2. ASML shipments
  3. Memory prices
  4. Taiwan tensions
  5. US-China relations
  6. AI demand growth
  7. Auto chip shortage

Portfolio Allocation

Conservative approach:
- 30% ASML (monopoly)
- 25% TSMC (critical)
- 20% NVIDIA (AI leader)
- 15% AMD (challenger)
- 10% Others

Aggressive approach:
- 40% AMD (undervalued)
- 30% Chip equipment
- 20% Emerging (ARM, etc)
- 10% Turnaround (Intel)

Conclusion: Silicon Sovereignty

The semiconductor industry has become the defining battleground of the 21st century. Control over chip production means control over the future of technology, economy, and military power.

Key takeaways:
- Chips are the new oil, but harder to replace
- Taiwan holds the world hostage (unintentionally)
- China's catch-up will take 10+ years minimum
- US reshoring is necessary but insufficient
- Investment in chips = investment in future

The next decade will determine:
- Whether democracy or autocracy controls technology
- If globalization survives or splits
- Who leads in AI and quantum
- The balance of global power

Remember: Every device you touch, every AI you use, every innovation you see—it all runs on chips. And whoever controls the chips, controls the future.

The war has already begun. Choose your side wisely.


In chips we trust. Everything else is just software.