ASML / China-DUV displacement is overpriced — the sell-off prices near-full substitution off a 5-machine data point

The July-2026 ASML de-rating priced a single load-bearing premise: that domestic Chinese immersion-DUV (Shanghai Aishengna / SMEE) displaces ASML immersion lithography roughly 1-for-1. The tool's own numbers contradict that — ~5 units in 2026 and ~20 in 2027, production-unproven, against ASML's 131 immersion systems in 2025, with ASML sold out through 2026-2027 (likely 2028). Even a mechanically-generous WORST-CASE 1:1 model (~190 cumulative machines by 2030, EUR 60M/tool, 60% GM) is only a ~7% 2030 EPS hit — far below the ~12% sell-off. This thesis holds the DEBATE: the near-term overreaction is fairly clear-cut, but the multi-year China DUV ramp is genuinely uncertain and is what the scored calls track.

DUV capacity converts to domestic output only if the BACK END is also available, and as of 2026-08-01 it is not VALID

• Chinese domestic immersion-DUV fab-shipment capacity0.75 strong
  Domestic immersion DUV entering mass production is the fact the overreaction thesis rests on, and it is real.
• Bonding Not Litho — signal 2026-08-010.80 strong
  80%
  Japan placed the complete back-end advanced-packaging chain under licence effective 1 August 2026 — hybrid bonders, TSV, thinning, dicing, high-end test — at a ~80% rejection rate, with original-manufacturer MAINTENANCE ending after a one-month transition.
• Hybrid Bonding0.70 moderate
  Chinese domestic share in bonding and sub-0.5um overlay metrology is near zero, which is what makes the control bite rather than merely annoy.
• HBM40.70 moderate
  HBM requires TSV and stacking, so a memory answer to litho constraints routes through exactly the equipment now controlled.

THE THESIS HAS BEEN ARGUED INSIDE THE EQUIPMENT LAYER ALONE and that is what made it thin. Domestic DUV is a genuine achievement and it is a FRONT-END achievement. Converting wafers into competitive parts requires thinning, TSV, bonding and test, and Japan controlled precisely that chain five days ago. So the overreaction call may be right about lithography and wrong about the system: relief on one constraint arriving as another closes is not relief. WHAT WOULD SETTLE IT is a domestically-packaged advanced part at volume, not a domestically-exposed wafer.

The bear case's load-bearing premise — domestic Chinese immersion-DUV displaces ASML immersion lithography ~1-for-1 — is contradicted by the tool's own ramp and unproven status VALID

• China Domestic Immersion-DUV Lithography (SMEE / Shanghai Aishengna)0.99 strong
  REACTIVE from derived:china-duv-lithography.demand_pull (today 83.2, band 33.29-133.16 -> p 0.98-0.84).
  why

  Pull on the domestic DUV programme is the displacement signal. RISING pull is evidence the substitution is real and the sell-off was NOT an overreaction, so p moves DOWN as the cell moves up — the inverse wiring is deliberate. Anchors are centred on today's reading so this makes the premise LIVE without re-rating the thesis. The domestic tool programme exists; not in dispute.

• Incentive × Capacity — the indigenization / margin-compression generator0.85 strong
  The prior itself is medium-conviction and explicitly NOT backtested (see its open_questions).
  why

  Scored below the facts it reasons over.

• China Domestic Immersion-DUV Lithography (SMEE / Shanghai Aishengna) — displacement pace slow0.75 strong
  5-machine data point is the whole evidence base.
  why

  Slow is the reasonable read, but the sample is tiny and a ramp is exactly the thing that surprises.

• China Domestic Immersion-DUV Lithography (SMEE / Shanghai Aishengna) — displacement ratio sub-1-to-10.70 moderate
  Domestic tools are lower-throughput, so sub-1-for-1 substitution is technically sound — but no fab has published a layer-level swap, so this is inference not observation.
• Chinese domestic immersion-DUV fab-shipment capacity — value nascent0.80 strong
  Desk measurement, qualitative stage.
  why

  Corroborated by absence of volume shipment reports.

Five premises must ALL hold for the bear case to be wrong on its own terms. The product is deliberately unforgiving: this is a conjunctive argument about a tool programme with a five-machine evidence base.

ASML's China-revenue structure and its backlog/policy context bound the real hit far below full substitution VALID

• ASML Holding0.99 strong
  Not in dispute.
• ASML Holding — china 290.90 strong
  Company-disclosed geographic revenue split.
  why

  High confidence; the residual doubt is that bill-to geography is not end-demand.

• US Export Controls on Advanced Semiconductors supplies ASML Holding1.00 strong
  Export-control exposure is established and ongoing.

The bounding argument is the strongest leg of the thesis because it rests on disclosure rather than on estimates of a competitor's ramp.

Therefore the ~12% sell-off off a 5-machine report is a scored, falsifiable mispricing, not a fact VALID

• ASML Holding0.99 strong
  Not in dispute.
• China Domestic Immersion-DUV Lithography (SMEE / Shanghai Aishengna)0.99 strong
  Not in dispute.
• Incentive × Capacity — the indigenization / margin-compression generator0.85 strong
  Medium-conviction, un-backtested prior.
• China Domestic Immersion-DUV Lithography (SMEE / Shanghai Aishengna) — production validation unproven0.85 strong
  No production-validated domestic immersion layer has been reported.
  why

  Strong, but it is an absence-of-evidence claim and would flip on a single credible qualification report.

The mispricing claim inherits the weakness of the displacement argument above it. Composed honestly, this is a real but not high-confidence call — which is consistent with the stated `medium` conviction rather than the confident tone of the conclusion text.