Weakness of Advanced manufacturing in America

[chinaexpert1]

Investigate America’s Failure to Scale Advanced Manufacturing

Overview

The United States remains a world leader in scientific research, software, intellectual property, semiconductor design, biotechnology, and venture formation. However, it often struggles to convert inventions and prototypes into high-volume domestic production.

This gap is visible in advanced technology trade. The United States moved from an approximately $13 billion advanced-technology-product trade surplus in 1995 to a record $297 billion deficit in 2024, even while retaining a strong surplus in intellectual property. This suggests that the country remains effective at invention and design but captures less of the manufacturing, supplier development, workforce learning, and production capacity generated by those innovations.

The problem is not simply that manufacturing moved abroad. Production ecosystems contain:

• Specialized suppliers;
• Tooling and machine builders;
• Skilled technicians;
• Process engineers;
• Testing and certification facilities;
• Critical minerals and refined materials;
• Pilot-production lines;
• Logistics networks;
• Experienced project managers;
• Customers capable of supporting early-scale production.

Once these systems disappear, rebuilding a factory does not automatically recreate the surrounding capability.

The United States has recently attracted substantial advanced-manufacturing investment, especially in semiconductors, batteries, data-center equipment, and related infrastructure. Semiconductor companies have announced hundreds of billions of dollars in U.S. projects since 2020, but announced investment is not equivalent to completed, competitive production.

The central research question is:

Why does the United States struggle to manufacture some strategically important technologies at scale, and which constraints most strongly determine whether domestic innovation becomes commercially competitive production?

The investigation should distinguish:

• Invention: creation of new scientific or technical knowledge;
• Prototype: proof that a product or process can work;
• Pilot production: limited manufacturing used to refine the process;
• Scale-up: expansion to commercially meaningful output;
• Mass production: reliable, repeatable, cost-competitive manufacturing;
• Industrial capacity: maximum sustainable production;
• Industrial capability: knowledge and systems required to produce;
• Supply-chain resilience: ability to continue production through disruptions.

The recommended primary units of analysis are:

• Product-country-year;
• Industry-state-year;
• Manufacturing establishment-year;
• Federal award or industrial project;
• Technology-development pathway;
• Supply-chain input.

---

Action Items

Initial Evidence

Several established trends justify the investigation:

• The U.S. share of global manufacturing output fell from approximately 28.4% in 2001 to 17.4% in 2023. Manufacturing employment also declined by roughly five million jobs between 1997 and 2024, although employment is an imperfect measure because automation can increase output with fewer workers.
• The United States runs a large advanced-technology-product trade deficit despite continuing strength in research and intellectual-property exports.
• Roughly $3 trillion in manufactured goods are imported annually. A recent analysis estimated that approximately one-quarter of those imports have some combination of national importance, concentrated supply, or dependence on geopolitically distant producers. For the most exposed products, domestic production would need to approximately double on average to satisfy current U.S. demand.
• The United States depends on imports for critical minerals used in batteries and semiconductors. GAO concluded that recycling could reduce some battery-mineral dependence within several years, but that substitution and recycling are less likely to solve short-term semiconductor-material dependence.
• U.S.-headquartered semiconductor firms remain globally strong in sales and R&D, while the United States continues to rebuild fabrication, materials, equipment, and packaging capacity. This demonstrates that the country has not uniformly lost advanced-manufacturing capability.
• New manufacturing construction does not guarantee operational success. Projects must still obtain equipment, energy, permits, suppliers, trained workers, customers, and acceptable production yields.
• Some current manufacturing measures may overstate strength because computer and electronics output indexes incorporate rapid quality improvements. Physical production, domestic value added, and supply-chain depth may provide different conclusions.
• Recent manufacturing growth associated with AI infrastructure shows that U.S. capacity can respond rapidly in selected sectors, but the recovery remains uneven.

These findings establish dependence and scale-up challenges, but do not prove that domestic production of every imported product would be economical or desirable.

Hypotheses

H1 — The United States has a missing pilot-to-scale financing system

Venture capital may fund invention and early prototypes, while conventional lenders avoid unproven factories and capital-intensive production.

Evidence to seek:

• Firms fail disproportionately between pilot and commercial production;
• Hardware startups raise less follow-on capital than comparable software firms;
• Financing costs rise sharply at the scale-up stage;
• Loan guarantees or purchase commitments improve project completion;
• Foreign competitors provide more patient industrial capital.

H2 — Supplier ecosystems matter more than the final assembly plant

A factory may remain dependent on imported materials, components, tooling, chemicals, and maintenance expertise.

Evidence to seek:

• Domestic content remains low after final assembly is reshored;
• Production delays are concentrated among second- and third-tier suppliers;
• Supplier proximity improves yield, lead time, and innovation;
• Import concentration is greatest upstream;
• Local supplier development predicts sustained production.

H3 — Loss of manufacturing weakens future innovation

Design and production knowledge may be complementary. Engineers learn from production failures, process variation, tooling limitations, and customer feedback.

Evidence to seek:

• R&D activity follows manufacturing relocation;
• Regions losing production later lose patents and technical employment;
• Co-location of R&D and production accelerates process innovation;
• Firms with domestic production improve products faster;
• The effect differs between modular and process-intensive technologies.

The proposition that innovation follows manufacturing is frequently asserted in current industrial-policy debate, but it should be tested rather than treated as universal.

H4 — Construction, permitting, and utility delays impede factory scale-up

Advanced factories require large sites, clean rooms, water, reliable electricity, environmental controls, transport, and specialized equipment.

Evidence to seek:

• U.S. factory projects take longer or cost more than foreign equivalents;
• Utility connection time predicts project delay;
• Permitting duration predicts cost escalation;
• Standardized industrial sites shorten delivery;
• Delays differ substantially across states.

H5 — Skilled technicians and production engineers are a binding constraint

Advanced manufacturing depends not only on doctoral scientists but also on technicians, electricians, machinists, welders, operators, maintenance workers, and process engineers.

Evidence to seek:

• Vacancies remain open despite high wages;
• New factories recruit workers from existing plants rather than expanding the workforce;
• Community-college programs predict improved staffing;
• Technician shortages delay equipment qualification;
• Skill demand differs from available educational credentials.

H6 — U.S. firms underinvest in process innovation

Companies may prioritize product design, software, marketing, and financial returns over manufacturing methods.

Evidence to seek:

• Process-related R&D grows more slowly than product R&D;
• Capital equipment is older than in competitor countries;
• Robotics and automation adoption lag in comparable industries;
• Process patents predict productivity;
• Investment responds to longer planning horizons and stable demand.

H7 — Market uncertainty prevents commercially efficient scale

A factory cannot reach competitive unit costs without dependable demand.

Evidence to seek:

• Projects with long-term purchase agreements reach completion more often;
• Demand volatility predicts canceled facilities;
• Government procurement accelerates early scale;
• Subsidies without customer commitments produce idle capacity;
• Production costs decline predictably with cumulative volume.

H8 — Industrial policy builds individual plants without complete supply chains

Grants and tax credits may produce visible facilities while leaving upstream and downstream dependencies unresolved.

Evidence to seek:

• Awarded plants retain high imported-input shares;
• Missing suppliers delay production;
• Incentives are weakly coordinated across supply-chain stages;
• Domestic value added remains low;
• Clusters outperform isolated subsidized facilities.

H9 — Trade protection can preserve capacity but may also raise downstream costs

Tariffs or price floors may support domestic producers while increasing input costs for U.S. manufacturers.

Evidence to seek:

• Protected industries increase output and investment;
• Downstream industries experience higher costs or lower exports;
• Supply dependence falls;
• Consumer prices rise;
• Net employment and value-added effects vary by product.

Current proposals involving polysilicon illustrate this trade-off: protecting domestic input production could improve security while raising costs for semiconductor or solar manufacturers.

H10 — Import dependence is dangerous only when combined with concentration and low substitutability

A high import share is not automatically a vulnerability when several reliable suppliers exist.

Evidence to seek:

• Disruptions are most severe where imports are concentrated;
• Alternative suppliers can or cannot be qualified quickly;
• Inventories provide insufficient protection;
• Substitution is technically difficult;
• Upstream chokepoints create cascading production losses.

H11 — Manufacturing success depends on productivity rather than employment alone

An automated factory may create relatively few jobs while providing high output, strategic capacity, and technical learning.

Evidence to seek:

• Output and value added rise despite flat employment;
• Highly automated plants remain internationally competitive;
• Employment-based evaluations undervalue strategic capacity;
• Local economic benefits depend on supplier formation and wages;
• Productivity differs by plant age and technology.

H12 — The scale-up problem differs greatly by technology

Semiconductors, batteries, pharmaceuticals, machine tools, shipbuilding, and advanced materials have different capital, workforce, regulatory, and supply-chain requirements.

Evidence to seek:

• Constraint profiles differ by sector;
• The same policy produces different outcomes;
• Capital intensity and learning curves explain variation;
• Import dependence has different strategic meaning;
• Sector-specific institutions outperform general subsidies.

---

Research Plan

1. Select strategic technology case studies

Begin with four contrasting sectors:

1. Semiconductors;
2. Batteries and critical materials;
3. Advanced machine tools and industrial equipment;
4. Pharmaceuticals or medical supply chains.

Optional extensions:

• Shipbuilding;
• transformers;
• solar manufacturing;
• aerospace;
• nuclear components;
• advanced materials.

Selection should be based on:

• Economic importance;
• Supply concentration;
• national-security relevance;
• import dependence;
• public-data availability;
• feasibility of substitution.

2. Define industrial capability metrics

Measure more than manufacturing employment.

Core metrics:

Additional measures:

• Domestic capacity;
• capacity utilization;
• production volume;
• physical output;
• productivity;
• workforce;
• plant age;
• supplier count;
• time to build;
• yield;
• import-source concentration;
• inventory resilience.

3. Map complete supply chains

For each technology, identify:

• Raw materials;
• refining and processing;
• chemicals;
• components;
• production equipment;
• software;
• assembly;
• testing;
• packaging;
• logistics;
• repair and maintenance.

Create a directed network:

Calculate:

• Node concentration;
• country concentration;
• criticality;
• substitutability;
• lead time;
• domestic capacity;
• downstream dependency.

4. Trace innovation-to-production pathways

For selected technologies:

Record:

• Funding;
• dates;
• investors;
• production location;
• technology transfer;
• plant status;
• capacity;
• employment;
• customers;
• failure point.

Use survival analysis to estimate the probability of reaching commercial production.

5. Measure announced versus operational capacity

Classify projects as:

• Announced;
• site selected;
• permitted;
• financed;
• under construction;
• equipment installation;
• pilot production;
• commercial operation;
• delayed;
• canceled.

Do not count announced investment as productive capacity.

For each project:

6. Analyze industrial-project delivery

Measure:

• Announcement-to-groundbreaking;
• groundbreaking-to-equipment installation;
• equipment-to-qualification;
• qualification-to-commercial production;
• initial versus final cost;
• initial versus achieved capacity.

Compare U.S. projects with foreign projects where definitions are sufficiently comparable.

7. Analyze workforce capacity

Map occupational demand for:

• Engineers;
• technicians;
• machinists;
• electricians;
• welders;
• operators;
• quality specialists;
• maintenance workers;
• production managers.

Compare:

• Job postings;
• employment;
• wages;
• credentials;
• program completions;
• apprenticeship capacity;
• regional worker supply.

Estimate whether shortages are caused by:

• Insufficient training;
• geographic mismatch;
• inadequate wages;
• experience requirements;
• security-clearance requirements;
• rapid simultaneous project growth.

8. Evaluate federal industrial policy

Track grants, loans, tax credits, procurement, and loan guarantees.

For each supported project, measure:

• Public dollars;
• private investment;
• completion;
• capacity;
• output;
• domestic value added;
• jobs and wages;
• supplier formation;
• imports displaced;
• cost per unit of capacity.

Use matched comparisons where feasible:

Avoid comparing subsidized strategic projects directly with ordinary factories without adjusting for risk and complexity.

9. Study learning curves

Estimate whether unit costs decline with cumulative production:

Compare U.S. and foreign learning rates.

A domestic plant may appear uncompetitive initially but become competitive after sufficient production. Conversely, persistent subsidies may conceal the absence of learning.

10. Measure supply-chain resilience

Simulate disruptions such as:

• Loss of a major source country;
• Port closure;
• mineral embargo;
• natural disaster;
• cyberattack;
• shipping interruption;
• sudden demand surge.

Estimate:

• Time to stockout;
• lost output;
• alternative sourcing time;
• price increase;
• substitution capacity;
• recovery time.

Use Monte Carlo simulations rather than one deterministic scenario.

11. Evaluate domestic-content quality

A product assembled domestically may still depend heavily on imported components and equipment.

Create a tiered classification:

• Domestic final assembly;
• Domestic component production;
• Domestic material processing;
• Domestic production equipment;
• Domestic intellectual property;
• Domestic maintenance and repair capability.

Report both final-assembly share and full domestic value-added share.

12. Identify successful industrial clusters

Find U.S. regions that have sustained:

• Production growth;
• supplier formation;
• productivity;
• workforce development;
• R&D;
• exports;
• plant reinvestment.

Compare with isolated projects that did not develop an ecosystem.

Potential explanatory variables:

• Anchor firms;
• community colleges;
• universities;
• energy cost;
• logistics;
• permitting;
• customers;
• supplier density;
• public procurement;
• state industrial organizations.

13. Test policy alternatives

Compare:

• Production tax credits;
• capital grants;
• loan guarantees;
• public procurement;
• tariffs;
• local-content requirements;
• worker training;
• supplier-development grants;
• industrial-site preparation;
• R&D support;
• recycling and substitution;
• strategic stockpiles.

Evaluate:

• Public cost;
• additional capacity;
• resilience;
• downstream prices;
• employment;
• productivity;
• international retaliation;
• time to impact.

14. Conduct robustness checks

Repeat analysis:

• Using trade value and physical quantity;
• With and without computer-quality adjustments;
• By final product and upstream inputs;
• Using employment, output, and value added;
• Using announced and operational capacity;
• Under different disruption scenarios;
• With alternative definitions of strategic importance;
• With and without defense-related demand.

15. Produce the Final PowerPoint

Recommended structure:

1. America invents but does not always scale;
2. Advanced-technology trade and production;
3. What industrial capability means;
4. Supply-chain maps;
5. The pilot-to-scale financing gap;
6. Plants announced versus operating;
7. Workforce and production knowledge;
8. Critical materials and upstream chokepoints;
9. Federal industrial-policy results;
10. Successful manufacturing clusters;
11. Which hypotheses survived testing;
12. Policy options ranked by cost and resilience.

The presentation must distinguish:

• Invention from production;
• Final assembly from domestic value added;
• Employment from productivity;
• Announced investment from operational capacity;
• Import dependence from actual vulnerability;
• Factory construction from competitive production;
• Strategic capacity from economic self-sufficiency.

---

Resources

Data, APIs, and Where to Get Data

Census International Trade API

Use for:

• Imports;
• exports;
• product classification;
• trading partner;
• customs value;
• quantities where available;
• advanced-technology products.

This should be the principal source for product- and country-level dependence.

USA Trade Online and Census Trade Downloads

Use for more detailed commodity, geography, and historical trade analysis where API endpoints are insufficient.

Bureau of Economic Analysis API

Use for:

• Manufacturing value added;
• input-output relationships;
• industry output;
• GDP;
• domestic supply and use;
• foreign multinational activity.

BEA input-output accounts can help estimate how an upstream disruption propagates through domestic industries.

Census Annual Survey of Manufactures

Use for:

• Shipments;
• value added;
• capital expenditure;
• employment;
• payroll;
• energy use;
• materials;
• industry productivity.

Economic Census

Use for:

• Detailed industry structure;
• establishments;
• employment;
• output;
• concentration;
• geographic production.

Census County Business Patterns

Use for:

• Establishment counts;
• employment;
• payroll;
• supplier density;
• regional industrial clusters.

Census Business Dynamics Statistics

Use for:

• Plant and firm entry;
• exit;
• job creation;
• firm age;
• establishment survival;
• regional manufacturing dynamics.

Manufacturing Extension Partnership Data

NIST’s Manufacturing Extension Partnership supports small and medium manufacturers.

Use available reports or program data for:

• Technology adoption;
• supplier development;
• productivity assistance;
• workforce challenges;
• firm outcomes.

Bureau of Labor Statistics APIs

Use for:

• Manufacturing employment;
• occupational employment;
• wages;
• productivity;
• producer prices;
• job openings;
• labor turnover.

O*NET Web Services

Use for:

• Occupation skills;
• tools;
• tasks;
• credentials;
• technology requirements.

This can connect industrial projects with required workforce competencies.

IPEDS and College Scorecard

Use for:

• Engineering and technician program completions;
• institution location;
• degree and certificate fields;
• workforce-pipeline analysis.

Apprenticeship.gov Data

Use for:

• Registered apprenticeships;
• occupations;
• sponsors;
• completions;
• geographic availability.

USAspending API

Use for:

• Manufacturing grants;
• loans;
• contracts;
• defense procurement;
• agency support;
• recipient;
• obligations;
• award modifications.

CHIPS Program Office Data

Use for:

• Semiconductor awards;
• project commitments;
• announced capacity;
• milestones;
• private investment;
• workforce commitments.

Cross-check agency announcements with construction and operating evidence.

Loan Programs Office Project Data

The Department of Energy Loan Programs Office provides project, loan, sector, and location information.

Use for:

• Battery plants;
• critical materials;
• energy manufacturing;
• loan guarantees;
• project status.

Department of Energy Supply Chain Data

Use for:

• Batteries;
• critical materials;
• transformers;
• grid equipment;
• energy technologies;
• manufacturing capacity.

USGS Mineral Commodity Summaries and Data

Use for:

• Domestic production;
• imports;
• reserves;
• refining;
• consumption;
• import reliance;
• critical minerals.

GAO Critical-Minerals Research

GAO’s 2026 analysis provides current evidence on import exposure, substitution, and recycling opportunities for battery and semiconductor supply chains.

USPTO Open Data

Use for:

• Patents;
• inventors;
• assignees;
• technology classifications;
• patent citations;
• ownership changes.

Link patenting with later domestic and foreign production.

OpenAlex API

Use for:

• Scientific publications;
• institutions;
• authors;
• topics;
• research collaboration.

This helps trace research strength preceding industrial development.

SEC EDGAR APIs

Use for:

• Firm R&D;
• capital expenditure;
• plant investments;
• risks;
• supply dependencies;
• acquisitions;
• business segments.

Company claims should be checked against physical production and trade data.

ImportYeti or Customs-Based Sources

Potentially useful for supplier relationships and shipment patterns, but coverage and licensing should be reviewed before inclusion in a fully open project.

Federal Procurement Data

Use for:

• Government demand;
• suppliers;
• domestic-content requirements;
• contract concentration;
• procurement timelines.

Government purchase commitments may be particularly important in defense, energy, medical, and infrastructure markets.

Federal Reserve Industrial Production Data

Use for:

• Manufacturing output;
• capacity;
• capacity utilization;
• industry trends.

Federal Reserve Economic Data

Use for:

• Industrial production;
• manufacturing construction;
• interest rates;
• energy costs;
• exchange rates;
• macroeconomic controls.

UN Comtrade API

Use for international product-level trade comparisons and supplier concentration.

OECD TiVA and STAN Data

Use for:

• Trade in value added;
• global value chains;
• industry structure;
• international productivity comparison.

World Bank and UNIDO Data

Use for broader international manufacturing-output, value-added, and industrial-development comparisons.

Foundational Resources

• CSIS, The Growing Gap Between U.S. Technology Innovation and Production.
• McKinsey Global Institute, Ramping Up Manufacturing in America?
• GAO, Critical Minerals: Reducing U.S. Import Reliance with Substitution and Recycling Technologies.
• Semiconductor Industry Association, 2026 State of the Industry Report.
• Census advanced-technology trade statistics;
• BEA input-output and multinational-enterprise accounts;
• NCSES research and development statistics.

Central Scientific Caution

A trade deficit does not automatically prove industrial weakness. Imports may reflect specialization, consumer demand, lower costs, or efficient exchange with reliable partners. Domestic production is not automatically more resilient, especially if it depends on one domestic plant or heavily imported inputs.

The correct question is:

Which manufacturing dependencies create unacceptable economic or strategic risk after accounting for supplier diversity, substitution, inventories, production lead times, and the cost of domestic alternatives?

The strongest investigation will compare:

1. Import share;
2. Supplier-country concentration;
3. Substitutability;
4. Inventory duration;
5. Domestic productive capacity;
6. Domestic technical capability;
7. Time required to expand production;
8. Economic consequences of disruption;
9. Cost of resilience;
10. Downstream effects of intervention.

The most valuable result may be a product-level industrial vulnerability index, rather than a general conclusion that all manufacturing should return to the United States.

• If this issue requires access to 311 data, please answer the following questions:
  • Do you need a one-time or ongoing dump of the data?
  • Do you need subset of data (i.e. certain years) or the entire data set (approx. 4 million rows or 11 GB)?
    • If a subset is needed, please define subset characteristics (i.e. date range, etc.)
  • Do you need online access via an API or a download of data?