✦ The Future BetsNarrow moat

TSMC's future bets are unlike anyone else's in this collection, because the company is betting mostly on itself: four wagers that the physics, the geography, the packaging and even the light can be made to keep cooperating.

The first is the roadmap into the angstrom era: N2 ramping now, A16 with its Super Power Rail backside-power delivery ready in 2026 — though volume production has slipped to 2027 — and A14 slated for 2028, with A12 and A13 already sketched for 2029¹. The second is geographic: the Arizona gigafab, $165 billion committed in March 2025 and since expanded toward $265 billion — six wafer fabs and two advanced packaging plants that would make Phoenix the first credible leading-edge cluster outside Taiwan². The third is packaging: CoWoS capacity driving toward roughly 130,000 wafers a month by late 2026 — nearly quadruple late-2024 — as the AI bottleneck migrates from transistor to package³. And the fourth is photonics: COUPE, the silicon-photonics platform entering production in 2026 to put optical interconnect directly into the package⁴.

None of these is optional speculation; each is the moat extending itself into the next decade's terrain, funded by $60-64 billion of annual capex⁵. But they are bets all the same: the angstrom nodes must yield, the American fabs must approach Taiwanese economics, the packaging expansion must not overshoot demand, and light must behave in volume manufacturing. Watch four dates the company itself set: A16 volume in 2027, Arizona's Fab 2 producing in 2027, the CoWoS supply gap closing through 2026, and COUPE shipping this year. TSMC rarely misses its own calendar — which is exactly why the calendar is the thing to watch.