⚠ Capital Intensity Meets a Cyclical IndustryHigh threat

The scale-and-capital moat has a matching vulnerability: TSMC must sink vast, fixed, up-front capital into a business whose demand is violently cyclical, and getting the timing wrong is enormously costly. Fabs take years to build and cost tens of billions, so TSMC must commit capacity long before it knows whether the demand will be there — and the semiconductor industry has a long history of building into booms that then turn to gluts, leaving expensive capacity idle. TSMC's revenue fell in the 2023 downturn, and its enormous current build-out¹ for AI is a bet that today's demand persists.

The tension is structural and permanent. The capital intensity that protects TSMC from competitors also chains it to a boom-and-bust cycle with the highest fixed costs in the industry: magnificent operating leverage when demand is strong, painful under-utilization when it is weak. The company is now spending sixty billion dollars and more a year on the assumption that AI-driven demand keeps accelerating, and while that assumption looks well-founded today, a cyclical downturn or an AI digestion phase would leave TSMC with the most expensive idle capacity in the industry. TSMC manages this risk better than any peer — its leading-edge focus, cost discipline, and diversified customers cushion the swings — but it cannot escape the fundamental hazard of pouring irreversible tens of billions into a cyclical business years ahead of the demand it hopes to serve. The moat and the risk are the same wall of capital.