The Semiconductor Supply Chain: Concentration, Fragility, and Geopolitics
The semiconductor supply chain is one of the most geographically concentrated and technically complex manufacturing networks on Earth. A single earthquake, a drought in Taiwan, or a policy change in Washington can ripple through the entire electronics industry within weeks. Understanding where the chokepoints are isn't just academic — it's increasingly a matter of national security.
The Geographic Reality
Here are the numbers that keep defense planners up at night: TSMC manufactures roughly 60% of the world's semiconductors by revenue and over 90% of the most advanced chips (sub-7nm). All of TSMC's leading-edge production is in Taiwan, an island 100 miles from mainland China. Samsung's advanced fabs are in South Korea, within artillery range of North Korea. These aren't hypothetical risks — they're geographic facts.
The concentration extends beyond foundries. ASML, headquartered in the Netherlands, is the sole supplier of EUV lithography machines. No EUV machine, no advanced chips. Period. ASML's supply chain itself depends on Carl Zeiss (Germany) for optics, Trumpf (Germany) for lasers, and Cymer (USA, ASML subsidiary) for light sources.
Semiconductor-grade silicon wafers come from five companies: Shin-Etsu and SUMCO (Japan), Siltronic (Germany), SK Siltron (South Korea), and GlobalWafers (Taiwan). Japan's Shin-Etsu alone controls about 30% of the market.
The Chemical and Gas Dependencies
Advanced chip manufacturing requires hundreds of specialty chemicals and gases, many with single-source suppliers. Neon gas, used in DUV lithography, was historically sourced primarily from Ukraine (about 50% of global supply pre-2022). The war disrupted this supply, and while alternative sources have come online, it illustrated how obscure inputs can create critical vulnerabilities.
Related reading: 5G and mmWave Chip Design: RF Front-End, Beamforming, and Po.
Photoresist — the light-sensitive material that patterns chip features — is dominated by three Japanese companies: Tokyo Ohka Kogyo, JSR, and Shin-Etsu Chemical. For EUV photoresist, the supplier base is even narrower. There's essentially no non-Japanese source for production-grade EUV resist.
High-purity hydrogen fluoride, used in wafer cleaning and etching, prompted a geopolitical incident in 2019 when Japan restricted exports to South Korea, threatening Samsung and SK Hynix's production. South Korea has since diversified some suppliers, but the episode demonstrated how chemical supply restrictions can function as economic weapons.
Equipment Supply Chain
Chip fabrication equipment (wafer fab equipment, or WFE) is a $100+ billion annual market dominated by a handful of companies:
- ASML (Netherlands) — lithography, monopoly on EUV
- Applied Materials (USA) — deposition, etch, CMP, inspection
- Lam Research (USA) — etch and deposition
- Tokyo Electron (Japan) — etch, deposition, coater/developer
- KLA (USA) — process control and inspection
Lead times for new EUV machines are approximately 18-24 months. ASML can produce roughly 50-60 EUV systems per year, and demand consistently exceeds supply. A new fab can cost $20+ billion and take 3-5 years from notable to volume production.
This connects to the ideas in Automotive Chip Requirements: ISO 26262, Temperature Ranges,.
The Reshoring Push
Multiple governments are spending billions to build domestic chip manufacturing capacity. The US CHIPS Act allocates $52.7 billion in subsidies. The EU Chips Act targets €43 billion. Japan has committed ¥3.9 trillion. Each is trying to reduce dependence on Taiwan and East Asia for advanced chips.
The results so far are mixed. TSMC's Arizona fab (currently being built) has faced construction delays, labor disputes, and cost overruns. Intel's Ohio fab complex is proceeding but on a revised timeline. Samsung's Taylor, Texas fab has also slipped from original schedules.
The fundamental challenge: you can build a fab anywhere, but you can't easily replicate the ecosystem of suppliers, experienced engineers, and institutional knowledge that's developed over decades in existing semiconductor hubs. A TSMC fab in Arizona will produce chips, but the ramp to full yield and efficiency will likely take longer than an equivalent expansion in Tainan.
Export Controls and Technology Denial
The US has implemented increasingly aggressive export controls targeting China's semiconductor industry. The October 2022 rules restricted sales of advanced chips and manufacturing equipment to Chinese entities. Subsequent updates in 2023 and 2024 tightened the restrictions and extended them to cover more countries and technologies.
Related reading: SRAM Cell Design: Bitcell Scaling, Read Stability, and Embed.
The impact on China's semiconductor ambitions is real but complicated. SMIC has demonstrated 7nm-class production using older DUV multi-patterning techniques (without EUV), but at significantly higher cost and lower yield than TSMC's EUV-based process. China can make chips at older nodes domestically, but the most advanced processes remain out of reach.
The equipment restrictions matter more than the chip restrictions. China can stockpile chips, but it can't stockpile the ability to manufacture advanced chips. ASML's EUV machines can't be sold to China, and even some DUV immersion lithography systems are now restricted.
Supply Chain Risk Mitigation
For companies that depend on semiconductors (which is essentially every technology company), supply chain risk mitigation has moved from a procurement concern to a board-level strategic issue. Common approaches:
- Dual-sourcing critical components across different foundries and geographies
- Building strategic inventory buffers (12-18 months for critical chips, up from just-in-time days of 2-3 weeks)
- Qualifying alternative components that can substitute if primary supply is disrupted
- Long-term supply agreements with committed capacity and pricing
None of these eliminate the risk. They reduce it, at the cost of higher inventory carrying costs and more complex supply chain management. The 2020-2022 chip shortage demonstrated that even companies with sophisticated supply chain teams can get caught when disruptions cascade across the entire industry simultaneously.
What Happens Next
The semiconductor supply chain is gradually becoming less concentrated, but "gradually" is the operative word. New fabs take years to build and qualify. New chemical and equipment suppliers take even longer to develop. The geopolitical tensions that drove the reshoring push aren't easing.
My honest assessment: the supply chain will be more geographically diversified in 2030 than it's today, but Taiwan and South Korea will still produce the majority of advanced chips. Complete supply chain self-sufficiency isn't achievable for any single country — the technology is too complex and the supply chain too specialized. The best realistic outcome is "sufficiency for critical needs" — enough domestic capacity to keep defense, infrastructure, and essential industries running even if international trade is severely disrupted.