Skip to main contentSkip to navigationSkip to search

Building a Strategic Semiconductor Capacity

1. What It Takes to Make Chips

The Semiconductor Manufacturing Process (Simplified):

1. Chip Design

  • Engineers Design Circuit: Using specialized software (EDA tools)
  • Simulation: Test design virtually before manufacturing
  • Mask Creation: Convert design to photomasks (templates for chip layers)

2. Wafer Fabrication ("Fab")

  • Silicon Wafer: Start with ultra-pure silicon disc (12 inches diameter)
  • Photolithography: Use UV light to etch circuit patterns (like photography, but nano-scale)
  • Deposition: Add layers of materials (conductors and insulators)
  • Etching: Remove unwanted material (create transistors and wires)
  • Doping: Add impurities to control electrical properties
  • Repeat: 50-100+ layers to create complex circuits
  • Result: One wafer has hundreds of chips

3. Testing & Packaging

  • Test Each Chip: Identify defects (30-70% yield is normal for cutting-edge chips)
  • Cut the Wafer: Separate individual chips ("dies")
  • Package: Encase chip in protective casing with connectors
  • Final Test: Ensure packaged chip works

4. Assembly into Products

  • Circuit Boards: Chips soldered onto boards
  • Integration: Boards assembled into phones, computers, cars, etc.

2. The Infrastructure Required

What We Need to Build:
1. Fabrication Plants ("Fabs")

Current State:

  • US Has ~30 Fabs (mostly older technology, 10nm+ nodes)
  • Advanced Fabs: Only Intel has a few 7nm/5nm fabs (behind TSMC's 3nm)

What "Advanced" Means:

  • Node Size: 3nm, 5nm, 7nm (smaller = more transistors per chip = more powerful, efficient)
  • TSMC 3nm: Most advanced commercially available (used in iPhone 15, newest chips)
  • Intel Is Stuck at 7nm: Years behind (technological failure)

Target Capacity:

Minimum Strategic Reserve:

  • 10 Advanced Fabs (3nm-5nm capability)
    • Each produces ~100,000 wafers/month
    • Total: 1 million wafers/month = ~10 billion chips/year
    • Enough for: Critical needs (military, medical, and infrastructure) if Taiwan is cut off

Robust Domestic Industry:

  • 30 Advanced Fabs (match Taiwan's capacity)
    • 3 million wafers/month = 30 billion chips/year
    • Enough for: Near self-sufficiency (still import some commodity chips, but not dependent)

Budget:

Per Fab Cost:

  • Cutting-edge (3nm): $20-30 billion each (TSMC's Arizona fab = $40 billion for latest tech)
  • Mature (7nm): $10-15 billion each
  • Older (28nm+): $5-10 billion each

Strategic Reserve (10 Fabs):

  • Mix of Technologies: 3 cutting-edge (3nm) + 4 mature (7nm) + 3 older (28nm)
  • Cost: (3 × $25B) + (4 × $12B) + (3 × $7B) = $75B + $48B + $21B = $144 billion

Robust Capacity (30 Fabs):

  • Mix: 10 cutting-edge + 12 mature + 8 older
  • Cost: (10 × $25B) + (12 × $12B) + (8 × $7B) = $250B + $144B + $56B = $450 billion
2. Equipment & Materials

Specialized Equipment Needed:

Lithography Machines:

  • Extreme Ultraviolet (EUV): Required for 3nm-7nm chips
  • Only Supplier: ASML (Dutch company, monopoly!)
  • Cost: $150-200 million per machine
  • Lead Time: 2-3 years to build one
  • US Dependence: Must buy from ASML (or develop own, which would take 10+ years)

Other Critical Equipment:

  • Deposition Tools: Apply thin films (Applied Materials, US company)
  • Etching Systems: Remove material (Lam Research, US company)
  • Ion Implanters: Dope silicon (Axcelis, US company)
  • Metrology: Measure nano-scale features (KLA, US company)

Total Equipment Cost per Fab:

  • Cutting-Edge: $8-12 billion (EUV machines + supporting equipment)
  • Mature: $4-6 billion
  • Older: $2-3 billion

Materials:

  • Ultra-Pure Silicon: Grown as single crystals, sliced into wafers
  • Chemicals: Hundreds of ultra-pure chemicals (photoresists, etchants, and dopants)
  • Gases: Ultra-high-purity gases (nitrogen, argon, silane, etc.)
  • Water: Billions of gallons of ultra-pure water (18 megohm-cm resistivity)

Material Costs: Included in fab operating costs (~$2-5 billion/year per fab)

3. Human Capital

Workforce Needed:

Per Advanced Fab:

  • Engineers: 1,000-2,000 (process, equipment, design, and test)
  • Technicians: 2,000-4,000 (operate equipment, maintain cleanrooms)
  • Support Staff: 500-1,000 (logistics, admin, security)
  • Total per Fab: 3,500-7,000 workers

For 30 Fabs:

  • Total Workforce: 105,000-210,000 workers (let's say 150,000 average)

Skills Required:

  • Electrical Engineering: Circuit design and process engineering
  • Materials Science: Thin films, crystallography, and doping
  • Physics: Optics (lithography) and quantum mechanics
  • Chemistry: Cleanroom processes, etching, and deposition
  • Mechanical Engineering: Equipment design and maintenance
  • Software Engineering: Automation, testing, and design tools

Training Pipeline:

Current Problem:

  • Limited Programs: Few US universities have semiconductor programs (shut down in 1990s-2000s)
  • Talent Shortage: 67,000 semiconductor job openings in US (can't fill them)
  • Knowledge Loss: Engineers who built 1980s fabs are retired/dead

Solution:

National Semiconductor Workforce Program:

  • University Partnerships: Fund 50 universities to create/expand semiconductor programs
    • Budget: $5 billion (endowments for programs, equipment, faculty)
  • Community College Training: 200 community colleges offer technician programs
    • Budget: $2 billion (equipment and curriculum development)
  • Apprenticeships: On-the-job training at fabs (2-year programs)
    • Guarantee: Every apprentice hired full-time after completion
    • Wage: $60k during apprenticeship, $80k+ after
  • K-12 Pipeline: STEM education focused on materials science, engineering
    • Budget: $1 billion/year (integrated with the platform's education policy)

Total Workforce Development: $10 billion over 10 years

4. Supply Chain

Critical Inputs:

Silicon Wafers:

  • Current Suppliers: Japan (Shin-Etsu, SUMCO), Germany (Siltronic), and Taiwan (GlobalWafers)
  • US Capacity: Minimal (GlobalWafers building plant in Texas, but not enough)
  • Need: 3-5 wafer manufacturing plants in US
  • Cost: $5 billion each × 4 plants = $20 billion

Chemicals:

  • Current Suppliers: Japan, Europe (BASF, Merck), and some US (DuPont, Honeywell)
  • Need: Expand US chemical production for semiconductor-grade purity
  • Cost: $10 billion (new production facilities)

Gases:

  • Current Suppliers: Mostly US (Air Liquide, Linde, and Air Products)
  • Need: Expand capacity for ultra-high-purity gases
  • Cost: $5 billion

EUV Lithography Machines:

  • Critical Bottleneck: ASML (Netherlands) only supplier
  • Options:
    • A. Buy from ASML: Pay $150-200M per machine (ongoing dependence)
    • B. Develop an US Alternative: 10-15 year R&D project, $50+ billion (risky, slow)
    • C. Joint Venture with ASML: Co-develop next-gen machines (share tech and reduce dependence)
  • Recommended: Option C + stockpile Option A machines as backup

Total Supply Chain Investment: $35 billion

3. Timeline to Build Capacity

Realistic Timeline for Strategic Reserve (10 Fabs):

Years 1-3: Planning & Site Preparation

  • Site Selection: Identify 10 locations (need water, power, and access to a skilled workforce)
  • Environmental Review: NEPA compliance (can be streamlined for national security)
  • Infrastructure: Build roads, power substations, and water treatment
  • Workforce Recruitment: Begin university/community college programs
  • Equipment Orders: Place orders for EUV machines (2-3 year lead time)

Years 4-6: Construction

  • Fab Construction: 2-3 years to build cleanrooms, install equipment
  • Parallel Development: All 10 fabs start construction simultaneously (not sequentially)
  • Equipment Installation: As buildings complete, install tools
  • Workforce Training: Engineers/technicians train at existing fabs, equipment vendors

Years 7-8: Ramp-Up

  • Process Development: Fine-tune manufacturing processes (each fab takes 1-2 years to reach full yield)
  • Qualification: Test chips meet specs, certify for customers
  • Volume Production: Scale from pilot to full production

Year 9-10: Full Operation

  • All 10 Fabs Producing: 1 million wafers/month capacity achieved
  • Meets Strategic Reserve: US can sustain critical chip needs without Taiwan

For 30 Fabs (Robust Capacity):

  • Years 1-10: Build first 10 (as above)
  • Years 8-15: Build second wave of 10 (lessons learned from first wave)
  • Years 13-20: Build third wave of 10 (by now, process is streamlined)
  • Year 20: Full 30 fabs operational

4. Total Cost Breakdown

Strategic Reserve (10 Fabs, 10 Years):

Category Cost
Fab construction $144 billion
Equipment (included in fab costs) (included above)
Supply chain (wafers, chemicals, gases) $35 billion
Workforce development $10 billion
R&D (next-gen processes) $20 billion
Contingency (10%) $21 billion
TOTAL $230 billion

Robust Capacity (30 Fabs, 20 Years):

Category Cost
Fab construction $450 billion
Supply chain expansion $70 billion
Workforce development (larger scale) $30 billion
R&D (continuous innovation) $50 billion
Contingency (10%) $60 billion
TOTAL $660 billion

Annual Operating Costs (Once Built):

  • Per Fab: $2-5 billion/year (materials, labor, utilities, and maintenance)
  • 30 Fabs: $90-150 billion/year (average $120 billion/year)

But Revenue Offsets Costs:

  • Chip Sales: $200-300 billion/year (30 fabs at full capacity)
  • Net Profit Potential: $80-180 billion/year (if managed efficiently)