The two-sided chip-chokepoint war — Western equipment vs China's materials (and who's racing to break it)

Web-verified 2026-06-11. Structured + sources: geopolitics-chip-chokepoint-war.json. Overlay — evidence-graded, excluded from the proofs. Sits upstream of geopolitics-taiwan-silicon-shield; connects to geopolitics-allied-intel-nodes, macro-critical-minerals-equities, defense_chokepoint, macro-crqc-quantum-landscape.

Upstream of Taiwan's fabs the chain is a two-sided chokepoint war: the West controls the equipment to make advanced chips and denies it to China; China controls the raw materials the West's fabs and weapons need and retaliates. Each side holds a gun to the other's supply chain. But — per the caution the question rightly raises — the chokepoint is sticky, not permanent: China and others are racing to break it, and the deepest moat is the one nobody can fully see.

Side 1 — the Western equipment chokepoint (NL + DE + JP)

• ASML's EUV monopoly (Netherlands): the sole maker of EUV lithography — ~100% of EUV, no competitor — and EUV is required below ~7nm. ~80% of ASML's EUV sales go to TSMC (fact).
• Zeiss optics (Germany): ASML's EUV optics come from a single partner, Zeiss SMT; ~80% of all microchips are made with Zeiss optics + ASML systems. High-NA EUV enters series production in 2026 — an even narrower chokepoint. A monopoly resting on a monopoly.
• Export controls + a reported kill switch: since 2019 the US + Netherlands have barred advanced EUV (and increasingly immersion DUV) from China — "the single most consequential decision in protecting Western AI dominance"; 2026 curbs tightened further. ASML reportedly built a remote kill switch into TSMC's EUV machines, able to disable them via a maintenance update if China invades (contested — reported, not confirmed). Implication: seizing Taiwan would not hand China working EUV — but it would brick the world's leading edge.
• Japan adds a parallel chokepoint: etch/deposition tools (Tokyo Electron) and >50% of 14 critical materials (~75% of high-end photoresist).

Side 2 — China's materials counter-leverage

• The bans: 3 Dec 2024 China banned gallium, germanium, antimony + superhard materials to the US (direct chip-control retaliation); 4 Apr 2025 added seven heavy rare earths (Tb, Dy, Sm, Gd, Lu, Sc, Y) — the magnets for motors and weapons.
• Extraterritorial reach (9 Oct 2025): modeled on the US Foreign Direct Product Rule — any foreign product with ≥0.1% Chinese-origin rare earths, or made using Chinese processing tech, needs a Chinese license. Beijing extended its writ across the whole global chain (fact).
• The truce: Nov 2025 China suspended the US-focused controls until 27 Nov 2026 — but the architecture remains and is reinstatable (same "suspend but keep the gun loaded" pattern as the antimony ban in macro-critical-minerals-equities). It bites because China processes ~90% of REE/antimony and most gallium/germanium/graphite.

Can the chokepoint be broken? China's indigenous push + alternatives + the tacit-knowledge wall

This is where it would be wrong to treat the Western lock as permanent — the question's caution is correct.

• China's multi-track EUV effort (graded — reported/projected):
• LDP-EUV (Huawei + SMEE + Harbin Institute): a Laser-Induced Discharge Plasma EUV source reportedly hit 100–150 W in mid-2025 (vs ASML's 250W+ for high-volume manufacturing) — enough for "first-light" testing of 5nm-class logic. A Shenzhen "whole-of-nation" prototype coordinated with SiCarrier is reportedly in testing. Reported timelines (SMIC "EUV-refined" chips ~late 2026; 2nm by 2028; parity "by end of decade") are aggressive and contested — treat as ambition, not delivery.
• SSMB-EUV (Tsinghua): a radically different path using a steady-state-micro-bunching particle accelerator (synchrotron) to generate a continuous high-power EUV beam — a potential leapfrog if it works, years out.
• DUV multipatterning: SMIC already ships 7nm (Huawei Mate 60) without EUV via DUV multipatterning — at a yield/cost penalty; the near-term workaround.
• Alternative approaches (anyone, graded): Canon Nanoimprint (NIL) — its FPA-1200NZ2C "stamps" patterns at ~14nm linewidth (~5nm-node), with a path toward ~10nm (~2nm-node), at ~1/10th the cost of an EUV scanner and ~90% less power; China is chasing NIL as an ASML alternative (300+ firms mobilized, first domestic NIL tool shipped). But the consensus (SemiAnalysis, Bits&Chips) is NIL won't rival EUV at the leading edge for the foreseeable future — defectivity/overlay/throughput problems "without a clear way forward." Silicon photonics / photonic computing is floated as a longer-shot bypass (compute without bleeding-edge transistors) — speculative. The quantum/CRQC frontier (macro-crqc-quantum-landscape, macro-quantum-computing) has its own manufacturing bottlenecks (cryogenics, error correction, qubit fab) and is not a near-term substitute.
• The deepest bottleneck — tacit knowledge & trade secrets (the "unknown unknowns"): the hardest moat isn't a machine you can reverse-engineer; it's the undocumented institutional know-how — process recipes, yield engineering, metrology, the thousands of supplier-ecosystem relationships, and decades of embodied expertise inside ASML, Zeiss, TSMC, and the Japanese materials houses. Espionage and reverse-engineering get you blueprints, not the recipe. This is why the chokepoint is sticky (years, not months) — and also why it is not permanent: tacit knowledge erodes under a sustained, well-funded "whole-of-nation" effort, talent poaching, and time. The honest read: the West's lead is real and measured in years, not in a permanent switch; China is climbing the curve on multiple tracks at once, and at least one (SSMB) could change the slope. Anyone claiming the chokepoint is permanent — or that China is one breakthrough away — is overstating a genuinely uncertain race.

The two-sided deterrence, and the full stack

Mutual chokepoint deterrence: the West can deny China the machines to make advanced chips (and reportedly kill-switch the ones in Taiwan); China can deny the West the materials to run its fabs and build its weapons. Each "suspension" is a truce over a loaded weapon, not disarmament — and both sides are racing to make the other's chokepoint obsolete.

The complete stack — each step in a different jurisdiction, most a single point of failure, all assumed open at once by the machine-verified AI loop:

Seven steps, seven jurisdictions, one assumption: that they all stay open, and cooperative, simultaneously.

← Research index · structured data: geopolitics-chip-chokepoint-war.json · geopolitics-chip-chokepoint-war.md