◆ What the Market Isn't Pricing In

TSMC presents a fascinating valuation puzzle: it is arguably the best manufacturing business in the world's most important industry, dominant and widening its lead, riding the AI supercycle — and yet it trades at around twenty-seven times earnings, well below ASML at about 54 times³, though after Nvidia's own de-rating to about 26 times⁴ no longer below the largest AI chip designer it manufactures for. At roughly $420 a share and a $1.9 trillion market value¹ in mid-2026, TSMC's multiple is rich by its own history and modest beside most of its peers, and the reason for the gap is no mystery. It is written on the map: the market applies a permanent, visible discount for the risk that TSMC's crown-jewel fabs sit on an island China has vowed to take. The whole investment question reduces to how one weighs that discount against the extraordinary quality of the business.

The bull case is that the market over-discounts the geopolitical risk and under-appreciates the business. TSMC is not merely a great foundry; it is the indispensable, near-exclusive manufacturer of the leading-edge and AI chips the entire technology economy depends on, with a moat — technology, scale, yield, ecosystem, lock-in — that has widened against subsidized rivals for a decade. Its recent results are spectacular: 2025 revenue up 36% past $122 billion², net income up 51%, record margins, and 2026 guided above 40% growth on insatiable AI demand. The 'silicon shield' argument holds that the world's dependence on TSMC actually reduces the odds of a catastrophic conflict, since an attack would destroy the prize and devastate all sides. On this view, a wide-moat, AI-critical franchise at twenty-seven times earnings is cheap — the geopolitical fear, real but low-probability in any given year, has left the best business in tech trading at a discount its quality does not deserve.

The bear case is that the discount is warranted and perhaps insufficient. The Taiwan risk is not a normal risk that can be modeled and diversified; it is a binary, catastrophic tail — a genuine conflict would be calamitous for TSMC in a way no multiple can price — and its probability, while low, is not negligible and arguably rising with US-China tension. Layered on top are the ordinary risks: the AI boom driving two-thirds of revenue is cyclical and untested by a downturn; the capital intensity is crushing and the overseas diversification dilutive; and subsidized rivals, backed by governments determined to break the dependence, are trying to close the gap. On this view, twenty-seven times earnings on peak-cycle AI profits, for a business with a catastrophic tail risk it cannot control, is not obviously cheap at all.

What the market may be under-appreciating, in either direction, is the sheer indispensability of the asset. There is no substitute for TSMC — no other company can build the leading-edge and AI chips the world needs at the required scale and yield, and none will emerge for years despite every government's efforts. That indispensability is what makes the silicon-shield argument credible and the business so valuable; it is also exactly what makes the concentration risk so acute, since the world cannot route around a disruption. The honest verdict is that TSMC is a wonderful, wide-moat, AI-critical business available at a reasonable price, with the entire discount — and then some — explained by a single, unhedgeable geopolitical risk. Whether that makes it a bargain or a value trap depends less on the moat, which is not in doubt, than on a question no analyst can answer: what happens across the Taiwan Strait. The market prices TSMC as the best business in tech, minus a large charge for the map. Whether that charge is too big or too small is, in the end, the whole of the investment case.