The Learning CurveWide moat

TSMC (TSM) — moat facet

Decades of accumulated process know-how — the asset money cannot buy, only time.

Underpinning TSMC's yield and manufacturing excellence is the 'learning curve' — the accumulated, experience-based knowledge of how to run leading-edge fabs at high yield and scale, built up over decades and impossible to acquire any way but by living it. Every wafer TSMC has ever made, every defect it has hunted down, every process it has tuned, has added to a body of tacit manufacturing knowledge that resides in its engineers, its processes, and its organizational memory. This is why a competitor can buy the same equipment from ASML and still not match TSMC's output: the machines are available, but the knowledge of how to wring high yield from them at scale is not.

Wafers shipped since 2014, cumulative (million)17201537201758201984202111220231402025Computed: sum of annual 12-inch-equivalent shipments, TSMC quarterly management reports
About 140 million wafers in twelve years: the experience a new entrant could only acquire by running fabs.

The learning curve is the deepest and most durable part of the scale-and-manufacturing moat because it is human and organizational rather than physical. It cannot be leapfrogged with capital, cannot be fully documented or transferred, and cannot be hurried — it is accumulated sequentially, node by node, wafer by wafer, over years. A rival starting the climb today is not just behind on the current node; it lacks the decades of accumulated learning that TSMC brings to solving each new node's problems faster and better. This tacit, experience-based mastery is what turns TSMC's technology lead into a durable moat: the process advantage might, in principle, be matched, but the manufacturing wisdom to execute it at world-leading yield and scale is the product of a climb no competitor has made, and cannot shortcut — a climb TSMC began as the world's first pure-play foundry in 19871.

Moat trajectory: Widening

Widening. Every wafer and every node adds to accumulated manufacturing know-how that a rival cannot buy or hurry — the experience gap compounds in TSMC's favor.

The number that tests this moat
Reported
Annual production capacity
More than 17 million 12-inch equivalent wafers (2025)

Every wafer made teaches the fabs something, so capacity is also the size of the learning base. Capacity growing faster than rivals' keeps the learning curve steeper.

Source: TSMC Form 20-F, FY2025 ↗
⚠ Threats to the moat
References
  1. ReportedTSMC pioneered the pure-play foundry model in 1987.
    TSMC — founded 1987 by Morris Chang on the pure-play foundry model (manufacture only, never design) — 1987-present · publ. 1987-2026 · source ↗
Sources
Generated September 23, 2026