TSMC to build second Japan chip factory
TSMC’s plan to build a second chip fabrication plant in Japan is framed as part of a broader effort to diversify advanced semiconductor production away from Taiwan and China, amid geopolitical risk and supply-chain shocks. Commenters compare Japan, the US and EU as alternative fab locations, weighing factors like earthquake resilience, work culture, labor costs, government subsidies and political stability. There is broad agreement that leading‑edge manufacturing will remain highly concentrated and strategically vital, even as countries push reshoring and “friend‑shoring” to reduce dependence on any single region.
Earthquakes, Infrastructure, and Fab Engineering
- Fabs in Japan raise questions about vibration sensitivity of lithography tools.
- Commenters note both Taiwan and Japan are earthquake-prone; fabs use seismic and pneumatic isolation, and buildings can be made earthquake‑resistant.
- Earthquakes still halt production and can ruin in‑process wafers; recovery involves realignment and downtime, not equipment destruction.
- Power stability is highlighted as critical for wafer ingot growth; the 2011 quake’s power issues hurt wafer supply for a long period.
TSMC’s Global Expansion: Japan, US, EU
- Japan seen as attractive due to work culture, existing optics/semiconductor ecosystem (Sony, Nikon, Canon), water resources, and weak yen.
- New Japan fabs: mid‑to‑advanced nodes (40nm down to ~6–7nm), with strong auto/electronics customers like Sony, Toyota, Denso.
- TSMC is also building in the US (Arizona) and Germany; construction is ongoing but slower in the US due to labor, logistics, and CHIPS Act uncertainty.
- Debate whether TSMC “gave up” on the US; others say they’re proceeding but struggling with conditions and talent.
Labor, Culture, and Cost
- Major tension around Asian “996”/long-hours fab culture vs. US/EU expectations on pay, safety, unions, and work–life balance.
- Some argue US labor is too expensive and unwilling to adopt TSMC-style workloads; others point out Intel, GlobalFoundries, and others already run fabs in the US, so it’s a management/culture issue, not impossibility.
- Disagreement on whether Taiwanese wages are “stagnant” or improving; both low pay and recent growth data are cited.
Taiwan, China, and Geopolitics
- Long, contentious discussion of Taiwan–China relations: language ties vs. political risk, divestment from China, and how much Taiwanese still see themselves as “Chinese.”
- Many see TSMC’s overseas fabs (Japan, US, Germany) as geopolitical hedging and reducing China’s leverage.
- Some argue China’s economic and demographic headwinds make a Taiwan invasion less likely; others fear Xi’s legacy ambitions and point to Hong Kong as a warning.
Reshoring and De-risking from China
- Multiple links and stats claim significant US/EU reshoring and declining Chinese exports; skeptics question methodology (“some reshoring” vs. meaningful scale).
- General view: globalization is shifting toward “friend‑shoring,” not ending, with moves to India, Vietnam, Mexico, etc.
Dependence on TSMC and Alternatives
- Thread clarifies roles: ARM designs IP, most fabless firms (Apple, AMD, Nvidia, etc.) rely on TSMC for leading‑edge nodes.
- Alternatives exist (Intel, Samsung, GlobalFoundries, Micron, many others) but are often on older nodes or smaller scale.
- Consensus: loss of TSMC’s leading‑edge capacity would be a severe global shock, but not total semiconductor collapse.
Chip Tech Trajectory and 3D
- Some expect eventual commoditization and easier low-volume runs; others note cost per transistor has risen at newer nodes and may keep rising.
- Discussion of 3D NAND, stacked cache, HBM, and thermal limits; 3D is advancing but heat and yield remain major constraints.
- Debate whether we’re near an asymptote or still have long runway via 3D, chiplets, and AI‑assisted design.
Defense, Drones, and Node Requirements
- Question raised: do militaries really need cutting-edge chips?
- Responses: yes for radar, autonomy, and advanced drones; but examples also show Russia using many cheap consumer microcontrollers instead of high‑end FPGAs.
- View: war is competitive—being a few generations behind can be exploitable, but not all systems need the latest node.