Concentration but StickinessNarrow moat

TSMC's customer base at the leading edge is concentrated among a handful of giants — Apple has long been its largest customer¹, and Nvidia has surged alongside it in the AI era, together with AMD, the hyperscalers, and a few others; TSMC's 2025 filing shows a customer that was below the 10% disclosure threshold in 2023 reaching 19% of revenue and overtaking the long-standing leader, which fell to 17%². This concentration is a double-edged feature of the moat. On one hand, these customers are deeply dependent on TSMC and locked in by the switching costs and co-development described elsewhere; the relationships are extraordinarily sticky, and the customers are the most valuable in the industry. On the other, the concentration means TSMC's fortunes ride on a small number of firms whose own cycles, product successes, and strategic choices it must absorb.

The stickiness dominates in practice: Apple and Nvidia cannot get comparable chips elsewhere, so their dependence on TSMC is real and durable, and it has deepened as their products (iPhones, AI accelerators) have grown more central to the technology economy. But the concentration hands these giants genuine negotiating power despite their dependence — a customer representing a large share of TSMC's leading-edge revenue is one TSMC cannot afford to lose or badly disappoint, and such customers extract favorable terms even from a dominant supplier. And it exposes TSMC to their fortunes: a stumble at a major customer, a downturn in smartphones, or a pause in AI spending would land heavily. The concentration is thus a moat (deep, sticky, high-value relationships with the best customers) wrapped around a risk (dependence on a few powerful buyers) — and the balance, for now, favors TSMC, because the giants have nowhere comparable to go.