The Semiconductor Supply Chain: Chokepoints, Controls, and the Race to Fabricate Independence

Scale, Growth, and the AI Tailwind

Global semiconductor sales reached $627.6 billion in 2024, a 19.1% increase from 2023's $526.8 billion. Industry projections for 2025 range from $707 billion (MarketsandMarkets) to $789 billion (Statista), with 92% of semiconductor executives forecasting revenue growth. For the first time, artificial intelligence has displaced automotive as the primary demand driver, with Nvidia's data centre revenue alone reaching $39.1 billion in 2024 and projected to exceed $170 billion in fiscal year 2026.

The market structure is defined by extreme concentration. TSMC's consolidated revenue reached $90.08 billion in 2024 — a 30% increase — with its foundry market share rising to 67.1% in Q4 2024. Samsung Foundry holds approximately 12%, struggling with yield issues in advanced nodes that have caused customer hesitation. Intel Foundry Services commands less than 1% of external revenue and does not rank in the top ten foundries. SMIC, China's largest foundry, holds approximately 6.3% of global foundry revenue, concentrated in mature nodes with limited but advancing capabilities at 7nm using DUV lithography.

Six Chokepoints That Define the Industry

The semiconductor supply chain contains six critical chokepoints, each controlled by two to three suppliers with 83-100% combined market share. These concentrations are structural — the result of decades of specialisation, capital intensity, and intellectual property accumulation — and cannot be replicated within any commercially or politically meaningful timeframe.

EUV Lithography: ASML's Absolute Monopoly

ASML holds a 100% monopoly on extreme ultraviolet (EUV) lithography systems, which are required for manufacturing chips at 7nm and below at competitive cost. ASML shipped approximately 40 EUV systems in 2024, each costing $150-200 million, generating 38% of the company's €28 billion revenue. No alternative supplier exists; no credible competitor is in development. China's attempts to develop domestic EUV capability have not produced a commercially viable system. The Netherlands expanded export restrictions in September 2024, requiring government licences for ASML's TWINSCAN NXT:1970i and 1980i DUV systems — further constraining China's access to even non-EUV advanced lithography.

The Export Control Escalation

Since October 2022, the United States has enacted four rounds of semiconductor export controls, each progressively broader in scope and more restrictive in application. The initial October 2022 controls restricted exports of advanced GPUs (notably Nvidia's A100 and H100) and semiconductor manufacturing equipment, while prohibiting US persons from supporting advanced Chinese chip facilities. The October 2023 update tightened performance thresholds using total processing performance (TPP) metrics, expanded the Entity List, and closed workarounds that had allowed modified chips (A800, H800) to circumvent initial restrictions.

The December 2024 round marked the most comprehensive expansion: controls on 24 types of semiconductor manufacturing equipment, three categories of development and production software tools, new restrictions on high-bandwidth memory (HBM), and 140 Entity List additions targeting Chinese tool manufacturers, fabs, and investment companies. In parallel, the Netherlands and Japan aligned their export control regimes, restricting the sale of advanced DUV lithography equipment and other critical manufacturing tools. The January 2026 imposition of a 25% tariff on advanced semiconductor chips adds a cost dimension to the regulatory barrier.

For international businesses, the export control regime creates a compliance maze. Companies must now verify not only the end-use of semiconductor products but the nationality of technical personnel, the location of design activities, and the ownership structure of customers. The extraterritorial reach of US controls — extending to any product containing US-origin technology, software, or equipment — means that virtually every advanced semiconductor in the world falls within scope.

The $200 Billion Subsidy Race

Six major jurisdictions have committed over $200 billion in semiconductor-specific subsidies, triggering the largest industrial policy competition since the space race. The United States leads with $52.7 billion under the CHIPS and Science Act, of which $39 billion is allocated to direct manufacturing subsidies. Key recipients include TSMC ($6.6 billion for three Arizona fabs in a $65 billion investment), Intel ($7.86 billion for facilities in Arizona, Ohio, Oregon, and New Mexico), and Samsung ($4.75 billion for advanced 2nm fabs in Taylor, Texas).

China's Big Fund III, established in May 2024 with 344 billion yuan ($47.5 billion) in registered capital — backed by 19 state-owned investors led by the Ministry of Finance — represents the largest single semiconductor investment fund in history. Its initial $12.7 billion allocation targets semiconductor materials, silicon wafers, and fab equipment manufacturers, signalling a strategic shift from finished chips to the upstream supply chain components that export controls have made most difficult to acquire.

Japan has committed approximately $27 billion (¥3.9 trillion) in cumulative semiconductor funding from 2021 to 2023 — equivalent to 0.71% of GDP, the highest among major economies relative to economic size. The EU Chips Act has catalysed €69 billion in investments across R&D and manufacturing facilities. South Korea's K-Chips Act provides 15-25% tax deductions for semiconductor investments, though notably without direct subsidies. India's Semicon India Programme commits ₹76,000 crore ($9.1 billion), offering up to 50% capital expenditure support for semiconductor fabs.

Progress and Constraints

China's semiconductor self-sufficiency rate stands at approximately 14% by value — against an original 2025 target of 70% that is now acknowledged as unachievable. Goldman Sachs projects this rate reaching 37% by 2030, though even this estimate assumes sustained investment and no further tightening of export controls. The gap is starkest in equipment: China's semiconductor equipment self-sufficiency stands at 13.6%, with TrendForce projecting a target of 50% by 2025 that remains aspirational.

SMIC has demonstrated 7nm chip manufacturing using DUV lithography — a technical achievement that requires 34 lithography steps compared to 9 with EUV, resulting in higher costs and lower yields. Reports indicate SMIC has progressed to 5nm-class production using its N+3 process, though yields remain below competitive levels. Huawei's HiSilicon division has launched Kirin 9020 processors using SMIC's 7nm process, with performance reportedly approaching TSMC 5nm levels despite the node disadvantage. CXMT, China's leading memory manufacturer, has begun mass production of DDR5 and LPDDR5X memory, reportedly achieving 80% yield rates by Q4 2024 (a figure disputed by South Korean industry analysts) — a significant milestone for domestic memory self-sufficiency.

Strategic Implications for International Businesses

The semiconductor supply chain presents three irreducible risks for international businesses. First, geographic concentration in Taiwan — 60% of global fab capacity and 90%+ of advanced node capacity — creates a systemic vulnerability that no subsidy programme can resolve within a decade. Companies must assess their semiconductor dependency not as a procurement issue but as a strategic risk requiring board-level attention and contingency planning.

Second, export controls are directionally permanent. Each round has been more restrictive than the last; no round has been reversed. Companies operating across the US-China technology boundary must build compliance architectures that assume further tightening, not stabilisation. The extraterritorial reach of US controls means that even non-US companies using US-origin technology, software, or equipment face enforcement risk.

Third, the subsidy race is creating a two-tier industry: companies with access to multiple subsidy jurisdictions can optimise manufacturing costs across the US, EU, Japan, and South Korea, while companies locked into single-jurisdiction supply chains face structural cost disadvantage. The strategic imperative is not merely to source semiconductors but to understand the political geography of the supply chain — because that geography now determines both access and cost.

This analysis draws on TrendForce foundry market data, Semiconductor Industry Association revenue statistics, ASML and TSMC company reports, US Bureau of Industry and Security regulatory filings, CSIS export control analysis, Congressional Research Service reports, Goldman Sachs semiconductor research, and IEA critical minerals data. All figures in US dollars unless otherwise noted.