Strategies

Never compete with customers (the foundational refusal)

TSMC has never designed or sold its own chips. 39 years of compounding customer trust because every customer — Apple, NVIDIA, AMD, Qualcomm, Broadcom — knows TSMC will never become their competitor. This single self-restraint enables every other strategic move.

• Ben:TSMC is totally fine with [Apple taking the credit for a 3nm process] because their job is not to market. It's to empower their customers.
  [Acquired TSMC, ch. Pure-play philosophy]
  EpisodeCite
• David:Morris: 'What very few people saw, and I can't tell you that I saw, was the rise of the fabless industry. I only hoped for it. But I had better reasons for hoping for it than the people at Intel, TI, and Motorola because I was now standing outside. When I was at TI and General Instruments, I saw a lot of these IC designers wanting to leave, start their own businesses, and the constraint is setting up their own fabs.'
  [Acquired TSMC, ch. Founding (quoting Morris Chang)]
  EpisodeCite

Geographic concentration AS the strategic asset

TSMC's leading-edge fabs cluster in Hsinchu and Tainan. The cluster effect — talent density, supplier ecosystem, government coordination — is itself a moat. The geopolitical concentration risk (Taiwan-China) is the cost; the cluster productivity is the benefit. Foreign fabs (Arizona, Japan) are insurance, not the engine.

Government-aligned strategic asset

From founding through today TSMC operates as a quasi-national-champion. Taiwanese government as early investor, ongoing utilities + zoning support, talent pipeline via universities, post-2022 fab subsidies in US/Japan/Germany. The model: a private company operating at strategic-infrastructure scale with state backing.

• Ben:ASML — they make the most advanced chip manufacturing machines in the world. They're the only company that makes them. They're located still in the Netherlands. Their biggest customer is TSMC.
  [Acquired TSMC, ch. ASML dependency]
  EpisodeCite
• ASML annual reports document the EUV equipment shipments that gate every leading-edge node.
  [ASML investor relations]
  SourceCite

Engineer-first culture; 12-hour work shifts

TSMC's process engineers run intense rotating shifts (the famous four-shift system). High-discipline, high-output culture; engineering rigor compounds across nodes. The cost is well-known burnout pressure; the benefit is yield rates competitors struggle to match.

Open Innovation Platform — co-design IP libraries with customers

Customers don't just send specs to TSMC; they jointly develop process design kits, IP libraries, and EDA tool integrations. Once a fabless company designs their next chip against TSMC's libraries, switching to another foundry requires re-doing years of design work. Switching cost = total IP redesign.

• David:As they pushed the manufacturing process technology forward, they're manufacturing better chips with smaller process links. They're enabling their customers, which are the fabless companies, to get better and better performance. As they get better performance, the fabless companies can address more of the market. Their existing customers get bigger and new fabless customers start, which gives them more revenue, which repeats the whole cycle. It goes slow. Like any flywheel, it takes a lot of effort and a lot of time to start turning it.
  [Acquired TSMC, ch. The flywheel]
  EpisodeCite

Process technology leadership at any capex cost

TSMC outspends Samsung and Intel on capex by ~2x. Each leading-edge node costs $15-20B+. The bet: be 12-24 months ahead on the process curve, win every leading-edge customer, fund the next node from those revenues. Compounds for 15+ years.

• Cumulative 2015-2024 capex: TSMC $198.5B, Samsung DS $238.5B, Intel $169B. The leading-edge race is a multi-hundred-billion-dollar arms race that only three companies on earth can credibly fund.
  [/api/semiconductor-capex (live data) + TSMC annual reports]
  SourceCite