TSMC

Everyone credits TSMC's dominance to better engineering. The real moat is a pledge made in 1987: never compete with a customer. That single promise turned every chip designer's R&D into a switching cost - and built 69.9% of the world's foundry market on trust.

Everyone reads A16 as TSMC's leap to 1.6nm. It isn't a new transistor at all - it's the N2 node with the power wiring moved to the back. TSMC took a feature it pulled from a cheaper node, gave it a new name, and built a premium AI-only lane that charges more without touching its mobile roadmap.

TSMC's $165B Arizona buildout and its Kumamoto fabs look like an exit from Taiwan concentration. They aren't. The cutting edge stays home: A16 volume production was confirmed for Taiwan first in 2027, while Arizona's bleeding edge waits and Japan makes mature chips.

TSMC just published a roadmap through a '1.2nm' node in 2029 — and its rivals are still slipping into 2027. With 71% of the foundry market and 66% gross margins, it can out-invest everyone at once. The moat isn't a node. It's the cadence.

Everyone names geography as TSMC's edge. Its own filings call Taiwan the principal risk. The real moat is a compounding loop of yield and co-design lock-in — which is why customers paid a premium to stay even after Samsung shipped 3nm first.

The standard read is that TSMC wins on the best transistors. The real weapon is its node cadence - made the industry's design calendar through a tools-and-IP ecosystem that helped lift advanced nodes to 74% of wafer revenue and over 60% global share.

TSMC's Arizona bill is quoted as ballooning from $12 billion to $165 billion. But that jump is four separate commitments across 2020–2025, not a runaway budget. The one fab that actually overran doubled to ~$20B and slipped about a year — chiefly for lack of skilled local labor.