Read the stack from the bottom up
A 2-nanometre chip is the output of a vertical assembly line of monopolies and oligopolies. At the base, Shin-Etsu grows the silicon wafer. ZEISS SMT polishes the EUV optics that go inside ASML’s lithography machines, the only tools that can print the smallest features. Lasertec verifies the photomask. Applied Materials supplies the deposition, etch and packaging tools that build the transistors layer by layer. TSMC runs the fab that turns all of it into chips — and SMIC tries to do the same inside China, without the one machine it is barred from buying. Every layer feeds the one above it, and the entire column ultimately feeds TSMC.
Leverage concentrates at the choke-points — and so does valuation
The stack is not flat. Three layers are genuine sole-source monopolies: ASML ships ~100% of production EUV, ZEISS is the only maker of its optics, and Lasertec is the only supplier of actinic mask inspection. Two are oligopolies — Applied holds ~18% of all fab-tool spend, and Shin-Etsu sits in a top-four wafer club. TSMC is a near-monopoly at the leading edge (~70% of foundry, >90% of the frontier). The market prices the hardness of each choke-point almost directly: ASML trades at ~58× earnings and TSMC at ~38×, premiums no broad toolmaker or wafer supplier commands. The lesson the cluster makes concrete is the oldest one in supply chains — the narrower and less substitutable your link, the more of the value you keep.
Monopoly and single-customer risk are the same coin
But indispensability has a hidden cost, and reading the studies side by side exposes it: the most irreplaceable layers are also the most dangerously concentrated. ZEISS sells its world-monopoly optics to essentially one customer — ASML, which owns 24.9% of it. Lasertec draws ~77% of revenue from just three fabs, and just watched orders fall ~60% on a single cancellation. ASML’s entire buyer base is a handful of names. Being irreplaceable up the stack means being hostage to a tiny set of buyers down it — which is why ZEISS’s own teardown asks whether being indispensable is the same as being safe. The answer the cluster suggests: only while the customers keep spending.
The export-control wall cuts the stack in two
The US-China controls do not hit the stack evenly — and the asymmetry is the most important thing the cross-section reveals. EUV was never sold to China, so the Western monopolies’ crown jewels were never exposed there; ASML’s China revenue is falling toward ~20%, but it is the lower-margin DUV tier that leaves, not the irreplaceable franchise. The controls instead do two things: they trim the oligopolies’ China sales (Applied has quantified ~$1B of lost revenue; Shin-Etsu faces Chinese oversupply), and they cap the one Chinese rung — SMIC — below the EUV line, leaving it to build 7nm on old DUV tools at a fraction of TSMC’s yield and margin (19% gross vs TSMC’s 62%). Lithography is the bottleneck, and SMIC’s ceiling, TSMC’s lead and ASML’s valuation are all downstream of the same fact: who has EUV.
Where they agree — and where they split
The studies broadly agree that the leading-edge choke-points are durable through this decade: ASML, ZEISS, Lasertec and TSMC all lean “monopoly intact” (independent assessments put a Chinese EUV-class capability beyond 2030). They split on three things demand cannot settle. The first is valuation — ASML, TSMC and Applied all land their own valuation questions as genuinely contested, the multiples pricing in an uninterrupted AI super-cycle. The second is High-NA timing: ZEISS’s and ASML’s next growth leg waits on TSMC, which is deferring volume High-NA to 2029-plus at ~$380M a tool. The third is cyclicality — Lasertec’s −60% order collapse and Applied’s sharper-than-market drawdowns are the reminder that even irreplaceable suppliers ride a brutally lumpy capex cycle. The demand for chips is the easy part. Whether these specific suppliers get paid smoothly for them is not.