Day05 | Global Semiconductor Manufacturing Equipment: Profits, Bottlenecks, and Falsifiable Judgments

One-sentence thesis

Semiconductor equipment is not a business in which everyone profits simply by “selling shovels”; it is a collection of high-switching-cost businesses built around pattern transfer, materials control, defect discovery, and qualification for volume production. Profit concentration is related to supply bottlenecks, but the two do not correspond one-to-one.

This report strictly separates actual results, forecasts, and assumptions. KLA’s “FY2027 Q1” is a disclosed quarter ended September 30, 2026, not calendar-year 2027 performance. Full-year 2026 is not yet complete; no 2027–2029 figures may be presented as actual results.

I. Core Judgments

1. The strongest pricing power comes from “the losses caused by an unsuccessful replacement,” not merely from the equipment’s selling price. Chip yield, cost per good wafer, process recipes, and qualification history jointly determine equipment value. KLA and ASML have relatively high operating margins in the financial-report sample, but this does not establish any unmeasured share of global value-chain profits captured by metrology or lithography. A multiproduct company’s margin is not a pure process-step margin.
2. Incremental demand in 2026 has spread from advanced front-end logic to DRAM/HBM, back-end testing, and advanced packaging. SEMI’s OEM forecast published in July 2026 projects total equipment sales of USD165.9bn, comprising WFE of 143.9, test equipment of 15.3, and assembly and packaging equipment of 6.7. These are forecasts, not full-year actual results. Broader demand also does not mean that every equipment subsector is simultaneously raising prices or experiencing longer lead times.[E23]
3. The clearest structural single-source exposure is in lithography optics, but vacuum valves are also among the constrained areas with direct order/delivery evidence in 2026. ASML discloses that Zeiss SMT is its sole supplier of critical optics and that its capacity limits system output; VAT’s orders are growing faster than revenue is being realized. By contrast, TRUMPF’s EUV business revenue fell 23.2% in the fiscal year ended June 2025, showing that “technologically critical” does not mean “in shortage in every period.”[E09][E11][E12]
4. Service compounds with the installed base; it is not a perpetual annuity detached from fab utilization and export licenses. Long-term agreements can stabilize maintenance, spare-parts, and upgrade demand, but equipment downtime, customer pricing pressure, upgrade deferrals, and licensing restrictions still transmit through to the business. AMAT’s service-segment margin is not inherently higher than its systems-segment margin either.[E03][E04][E13]
5. Globalization has not disappeared; it is being repriced. Headquarters location, critical-component production location, final-system assembly location, customer equipment installation location, sales recognition geography, and export-control jurisdiction must be separated. Mainland China’s large equipment purchases do not imply a large share for Chinese equipment suppliers; progress in some etch, deposition, and cleaning steps does not imply substitutability across the entire advanced-node process chain.
6. The next question should first be “How do orders become qualified output?” and only then “How large is the market?” Orders, delivery, acceptance, installation, production-line qualification, yield ramp, and actual fab output are different states. A supplier’s capacity-expansion announcement proves only that investment has occurred or a plan has been formed; it does not directly prove that a bottleneck has been removed.

II. Correcting the Boundaries: Align the Denominators and the Clocks First

2.1 Research Scope

Critical upstream inputs: optics/lasers, vacuum/fluids, RF power, precision motion, detection sensors, specialty materials
           ↓
Equipment OEMs: lithography—coating/developing—etch—deposition/epitaxy—cleaning—CMP—ion implantation—thermal processing
           ↕ Metrology/inspection, process control, and embedded software throughout the flow
           ↓
Wafer fabrication → wafer test/known-good-die (KGD) selection → advanced packaging/bonding/dicing → post-packaging test
           (Some testing and inspection occur throughout fabrication and packaging, not only at the end.)
           ↖ Spare parts, maintenance, refurbishment, and productivity upgrades span the full life cycle

• Included: Wafer fabrication equipment, mask-related equipment, manufacturing-process metrology/inspection, and assembly, packaging, and test equipment; direct critical components and services are included to explain pricing power.
• Adjacent, but not included in the equipment-market total: Wafers, photoresist, electronic gases, packaging substrates, cleanroom construction, fab-building civil works, utility-infrastructure construction expenditure, pure EDA, foundry revenue, and chip sales revenue. These may be bottlenecks, but they do not share the same equipment-sales denominator. The exclusion here concerns construction expenditure and does not exclude fab facilities equipment within SEMI’s WFE definition.
• Do not double-count: Critical-component sales are already embedded in OEM system costs. Adding component-supplier revenue, OEM sales, fab capex, and chip revenue does not produce the value chain’s “market size.”

2.2 Six Common Misreadings and Their Corrections

Conflict record [conflict E22-vs-E23]: adjudication deferred. SEMI’s WWSEMS statistics published on April 7, 2026 report actual 2025 equipment sales of USD135.1bn.[E22] SEMI’s OEM forecast published on July 14, 2026 projects 2026 sales of USD165.9bn and 23.2% year-over-year growth, mechanically implying a 2025 base of approximately USD134.66bn.[E23] A difference exists, but it is not possible to confirm whether it arises from a version revision, statistical-scope differences, or another cause. In the absence of underlying bridging evidence, adjudication is deferred. This report retains each source’s date and value and does not replace the April actual statistic with the back-calculated figure.

Different products must not create false conflicts. The above difference is distinct from the WFE/300mm boundary issue. July OEM WFE and October 300mm fab equipment spending cannot be directly subtracted and labeled an “upward demand revision.”[E23][E24]

2.3 Financial Comparison Rules

The main profit-map table uses reported operating income ÷ revenue. Four companies use US GAAP; ASM uses IFRS and TEL uses Japanese accounting standards, explicitly labeled. Gross margin, operating margin, net margin, and adjusted margins are not mixed. ASML’s annual report contains both IFRS and US GAAP materials; gross margins from the two bases must not be spliced together. A subscription share must also specify that its denominator is service revenue, not total company revenue.

III. Four Falsifiable Hypotheses: Verdicts, Counterevidence, and Monitoring Thresholds

Confidence is a subjective assessment of evidence strength, not a statistical confidence interval. “High” means multiple primary sources point in the same direction; “medium” means the mechanism is credible but production-line-level or cross-cycle validation is lacking.

H1 | Profits Concentrate in Hard-to-Replace Processes and Yield Control, Rather Than Being Distributed Evenly Across All Equipment

Verdict: Supported, with high confidence; however, “irreplaceability alone explains every profit difference” is not supported.

• Supporting evidence: Operating margins in the 2025 financial-report sample were approximately 39.3% for KLA, 34.6% for ASML, 32.0% for Lam, and 29.2% for AMAT. In the latest comparable quarters through September 2026, KLA and ASML remained at high levels. This reflects the combined effects of product mix, installed-base service, R&D leverage, supply and demand, and operating leverage—not a single-variable experiment.[E01][E02][E03][E04][E05][E06][E07][E08]
• Inferred mechanism: On an expensive volume-production line, equipment price is only one part of the decision. Qualification time, defect escapes, downtime risk, and recipe rework caused by replacement create high switching costs around qualified suppliers. For expensive chips, early defect detection has particularly high economic value.
• Counterevidence: Besi achieved a 65.2% gross margin in Q3 2026, showing that profit opportunities are not exclusive to the front end. But this is a gross margin, which cannot be ranked directly against the operating margins in this section; it is only supplementary evidence that the back end also has value-capture ability. TRUMPF’s technological position in EUV did not prevent related revenue from declining in FY2024/25. Single-customer dependence, component-level contracts, and different business models can change profit allocation.[E12][E14]
• Falsification conditions: Over four consecutive quarters, process-control and advanced-lithography samples no longer earn operating margins above the equipment-sample median on a consistent basis and after removing one-off effects; or new entrants pass cross-customer volume-production qualification, materially lower prices, and cause incumbent suppliers to lose share persistently. Either would warrant lowering the assessment of “moat stability,” rather than masking the evidence with industry growth.
• Current evidence gap: A global operating-profit pool by process step on a uniform accounting basis has not been obtained; therefore, it is not valid to claim that “lithography plus metrology account for X% of global profits.”

H2 | Bottlenecks Have Expanded Beyond “EUV Alone” to Critical Components, Process Control, and Packaging Coordination

Verdict: Supported conditionally, with medium confidence. An EUV-only narrative is clearly insufficient, but physical shortages have not been established in every area.

• Supporting evidence: ASML’s annual report explicitly states that Zeiss’s optical capacity limits system output. In July 2026, Zeiss disclosed an Oberkochen expansion adding approximately 25,000 square meters of space in phases. VAT’s H1 orders increased while revenue declined, indicating that conversion of orders was constrained by the capacity ramp.[E09][E10][E11]
• Latest signal: For Q3 2026, VAT reported orders of CHF696.6m, revenue of 376.4m, and backlog of 979.5m. Customers extended how far in advance they place orders; this is not the same as equipment delivery lead time. Customers may also be securing supply early, requiring attention to duplicate-order and cancellation risk.[E19]
• Counterevidence: TRUMPF’s EUV business contracted. Besi reported order growth while also disclosing that some customers scheduled deliveries around their H1 2027 capacity plans. The former rejects the claim that “critical components must remain continuously scarce”; the latter shows that delivery timing partly reflects customers’ deliberate demand scheduling.[E12][E14]
• Precise language for metrology and packaging: Evidence of growth and product complexity is strong; public evidence of “industry-wide longer lead times/shortages” is weaker. These can only be described as potential yield and process-integration bottlenecks, not declared shortages on customers’ behalf.
• Falsification conditions: Suppliers’ actual quarterly output, on-time delivery, and customer qualification progress improve persistently, while advance-order windows/backlog decline without deteriorating orders. Physical bottlenecks should then be downgraded. Completion of a new plant alone, without qualified output, does not constitute falsification.

H3 | Installed-Base Service and Upgrade Revenue Make Equipment Suppliers More Resilient, but Cannot Eliminate Cyclicality

Verdict: Supports a “buffer” but weakens the claim of “immunity,” with high confidence.

• Supporting evidence: AMAT disclosed that more than two-thirds of FY2025 service revenue came from long-term subscription agreements, with an average contract term of 2.9 years. KLA’s 2025 service materials disclosed subscription, renewal, and contract-duration metrics using 2024 as the baseline. The installed base and contract coverage do improve visibility.[E13][E15]
• Economic mechanism: New-equipment sales are first-order capital expenditure, whereas maintenance service is more closely tied to existing production activity. Productivity upgrades may both increase OEM service revenue and postpone customers’ purchases of new machines; therefore, “upgrade growth” does not necessarily mean “stronger new-equipment demand.”
• Counterevidence: AMAT’s FY2026 Q3 operating margin was 37.7% in its systems segment and 30.1% in AGS; service is not necessarily more profitable. Starting in FY2026 Q1, 200mm equipment moved from AGS to systems. Also from FY2026 Q1, the company fully allocated corporate support costs to its segments, with comparative periods recast. Recast comparable figures must be used; reclassification or cost-allocation changes must not be mistaken for changes in the service cycle.[E04]
• Another countervailing mechanism: Low utilization can depress some consumables, spare parts, and time-and-materials service revenue. Changes in export licenses may restrict the delivery of maintenance and upgrades. Long-term contracts create rights and obligations, not unconditional cash inflows.
• Falsification conditions: During a meaningful downturn in new-equipment revenue, if service revenue and service profit on a consistent basis fall by the same or a greater magnitude, while renewals, coverage, or installed-base utilization deteriorate, the strength of the “cycle buffer” should be revised downward. At least four quarters of observation are needed; a single-quarter rise in service mix does not prove growth in absolute service revenue.

H4 | Domestic Substitution in China Is Deepening Process by Process, but Does Not Yet Establish a Fully Independent Advanced-Node Process Chain

Verdict: Supported conditionally, with medium confidence; neither “substitution is complete” nor “there has been no progress” is supported as a blanket claim.

• Evidence of deeper substitution: NAURA’s 2025 annual report discloses expansion of its equipment portfolio and deliveries for volume production, showing that local platforms are broadening their coverage. However, company revenue growth also reflects end-market investment, product mix, and acquisition consolidation; it cannot directly be treated as the magnitude of share gains in a given process step.[E16]
• Evidence of remaining bottlenecks: ASML’s position in commercial EUV supply, concentration in Zeiss optics, and US, European, and Japanese regulatory frameworks indicate that dependence on advanced lithography, some process-control tools, and critical components still merits separate scrutiny. The ability to produce certain advanced chips does not establish full-chain substitution at equivalent cost, throughput, yield, and equipment life.[E09][E26][E27][E28]
• Counterevidence: Usable alternatives may exist for mature and some advanced process steps; import restrictions may also create more opportunities to qualify local suppliers. The absence of a commercial EUV substitute cannot be used to infer that all etch/deposition/cleaning suppliers lack competitiveness. Conversely, one company’s revenue growth cannot prove that imported equipment has exited every process.
• Falsification conditions: Upgrade the substitution assessment for a process step only when at least two independent fabs disclose repeat volume-production orders, capacity, defect-rate/yield and reliability performance for the relevant equipment within the complete flow, alongside verifiable availability of critical components. A fully independent advanced-node process chain must additionally cover lithography, inspection, masks, spare parts, and long-term service.
• Policy boundary: Historical rules and press releases cannot substitute for a product-by-product licensing assessment as of the cutoff. This report offers no routes for evading export controls, nor does it portray licensing requirements as a blanket ban on all equipment sales.[E26][E27][E28][E29]

IV. Profit Map: Put Scale, Margins, Moats, and Risks on the Same Map

4.1 Financial Anchors for Six Sample Companies

Units: millions of local currency. Operating margin is operating income/revenue, rounded. The latest quarter is not a full-year run rate and has not been annualized.

Sources: [E01][E02][E03][E04][E05][E06][E07][E08][E17][E18]. KLA’s operating income can be recalculated by subtracting cost of revenue, R&D, selling/general/administrative expenses, and operating impairment charges for the period from revenue. FY2025 additionally includes a deduction of USD239.100m for goodwill/intangible-asset impairment; net income must not be substituted. TEL materials scheduled for release on October 30 are not included before their actual publication.

Do not label the difference between an annual figure and a single quarter as year-over-year improvement. Genuine same-quarter comparisons are: ASML operating margin of 32.8% in 2025 Q3 → 44.4% in 2026 Q3; AMAT 30.6% in FY2025 Q3 → 33.7% in FY2026 Q3; and Lam 34.4% in the quarter ended September 28, 2025 → 40.0% in the quarter ended September 27, 2026. Moreover, TEL’s reported FY2026 annual operating margin was approximately 25.6%, below FY2025’s 28.7%. Net income was affected by gains on securities disposals, so net-profit performance cannot establish improvement in the core business’s operating margin. This counterexample shows that “all leaders’ profitability is moving uniformly upward” is not valid.[E02][E04][E06][E17]

What magnitude of profit can be established? The above FY2025 operating income of the three US-dollar-reporting samples—AMAT, Lam, and KLA—totals USD18.965bn, on combined revenue of USD58.960bn, for a weighted sample operating margin of approximately 32.17%. ASML is shown separately with EUR11.301bn of operating income; no US-dollar aggregation is made without a specified exchange rate. These totals are neither for the same calendar-year period nor the global equipment industry’s profit pool, and cannot be divided by SEMI equipment sales to derive an “industry margin.”

4.2 Which Layer Captures Value, Why, and When It Could Lose It

This is not a binary split between a “high-end front end” and a “low-end back end.” Difficult back-end bonding and testing can command substantial premiums. Critical upstream components, even when hard to replace, may have profits constrained by a single large customer, contract terms, and expansion costs. Purchasing scale, price, revenue, profit, and the consequences of supply interruption are five different rankings.

V. Bottleneck Ranking: Global Risk Priorities and Regional Perspectives

5.1 Global Ranking

The following is a qualitative risk priority based on “production-line interruption impact × difficulty of short-term replacement × expansion/qualification time.” It is not a statistical probability of supply interruption or an industry-standard lead time.

References: [E09][E10][E11][E12][E14][E07][E23].

5.2 Regional Ranking Must Ask “What Constrains Local Capacity Expansion?”

Regional purchase figures: [E22]; construction pipeline: [E24]; production network: [E11]. This report avoids assigning seemingly precise “bottleneck scores” to regions lacking uniform data.

VI. The 2023–2026 Timeline: From Downturn to AI-Led Demand, Without Deleting the Counterexamples

Sources: [E20][E21][E22][E23][E24][E25][E26][E27][E02][E06][E08][E10][E11][E12][E13][E14].

Read the cycle chain this way: Expected end demand for compute/memory → chipmakers’ budgets and process roadmaps → orders → critical-component/system production → shipment and acceptance → installation and qualification → effective good output. Every arrow contains a time lag. A surge in quarterly orders may mean stronger demand, or it may mean advance ordering to secure supply; subsequent cancellation rates, delivery rates, cash, and customers’ capex execution rates are needed to distinguish the two.

VII. Three Reproducible Stress Tests

The following are not earnings forecasts, nor do they claim knowledge of the true AI share, capacity shortfall, or proportion affected by licensing. Their purpose is to show transmission paths and thresholds. Every parameter not taken from financial reports is explicitly identified as an analytical assumption.

Test One: How Much WFE Revenue Could Be Lost If AI-Related Capex Is Delayed?

Known baseline: SEMI’s July 2026 WFE forecast of USD143.9bn.[E23]

Assumptions: 55% is sensitive to the timing of AI-related investment; 20% of this portion’s originally scheduled deliveries are delayed; and 30% of delayed orders can be replaced by other projects in the current period. Permanent order cancellations are not assumed.

Net current-period deferral = 143.9 × 55% × 20% × (1 − 30%) = USD11.0803bn
Scenario WFE revenue = 143.9 − 11.0803 = USD132.8197bn
Decline from baseline = 11.0803 / 143.9 = 7.7%

If the incremental operating-profit conversion rate for the net deferral is assumed to be 35%–50%, the operating-profit impact is approximately USD3.88–5.54bn. This is only a transmission sensitivity and cannot be divided by an unknown industry profit pool to produce a precise percentage decline in profits. Delivery of the orders in the following year would represent a timing shift; permanent demand loss arises only if customers cancel. The same shock is not imposed on test and packaging, to avoid repeatedly applying it across different markets.

Monitoring thresholds: Customer budget cuts, supplier disclosures of cancellations/deferrals, customer advances and backlog quality, and the speed at which substitute projects absorb the gap. If deliveries are merely delayed while budgets remain unchanged, first reduce the current-period revenue cadence rather than automatically cutting the long-term market opportunity.

Test Two: More Factory Space Does Not Equal Qualified Output—A Weakest-Link Calculation for a Serial System

Assumed units: All capacity is expressed as a “monthly equivalent wafer-flow index,” not actual data from a particular factory. Initial lithography capacity is 100, with a planned addition of 20. Optical constraints limit the actual addition to 12, while productivity upgrades to the old installed base add another 3. Nominal capacities at other steps are: etch 125, metrology 120, and packaging planned at 120 but actually 110. Final combined yield is 90%.

Planned good output = min(120,125,120,120) × 90% = 108
Actual lithography capacity = 100 + 12 + 3 = 115
Stress-scenario good output = min(115,125,120,110) × 90% = 99
Shortfall versus plan = (108 − 99) / 108 = 8.33%

• Raising only packaging to 120 increases good output to 115×90%=103.5, still 4.17% below plan.
• Restoring only lithography to 120, while packaging remains at 110, leaves good output at 99: the lithography increment is “absorbed” by the back-end bottleneck.
• If packaging improves and combined yield rises from 90% to 94%, output becomes 115×94%=108.1, close to the original plan. Yield improvements can offset some hardware shortages, but this number is a scenario, not a promise.

Implication: Using ASML output or the number of individual bonding tools as a direct proxy for total AI-chip capacity misses the minimum capacity across the full chain and its yield. The model does not address re-entrant process steps, product mix, queuing, equipment availability, or independent yields at each stage; it only illustrates the weakest-link mechanism. Assumptions should be replaced with actual wafer starts, equipment availability, production-line cycle times, and final good-unit counts.

Test Three: How Much Revenue and Profit Can Service Protect When Export Restrictions Combine with Local Substitution?

Assume a representative equipment supplier has baseline revenue of 100, comprising 75 from new machines and 25 from service. Assume a China revenue share of 30% for both categories. Of the relevant new-machine revenue, 40% is affected; 25% is replaced in the current period through lawful approvals or demand in other markets. For service, 10% is affected and cannot be replaced in the current period. These are not any company’s actual controlled proportions.

Net new-machine impact = 75 × 30% × 40% × (1 − 25%) = 6.75
Net service impact = 25 × 30% × 10% = 0.75
Scenario total revenue = 100 − 6.75 − 0.75 = 92.5

Further assume a 50% gross margin for new machines, 55% for service, and fixed operating expenses of 20:

Baseline operating income = 75×50% + 25×55% − 20 = 31.25
Scenario operating income = (75−6.75)×50% + (25−0.75)×55% − 20 = 27.4625
Operating-income decline = 12.12%; scenario operating margin = 27.4625 / 92.5 = 29.69%

Service loses only 0.75, so it does buffer revenue. But fixed R&D/selling expenses still make the decline in operating income greater than the decline in revenue. If service is completely unaffected, revenue is 93.25 and profit is 27.875—still not immunity. If the replacement share falls to zero, revenue declines to 90.25 and profit to 26.3375.

Model legal and commercial states separately: Already restricted, awaiting a license, license approved, supplier voluntarily stopping supply, and customer switching to a local supplier are different states. Do not set all China revenue to zero. Conversely, do not assume that production outside the country is unrestricted. Regulatory scope and product classifications require item-by-item professional confirmation.[E26][E27][E28][E29]

VIII. Evidence Gaps and the Most Valuable Work for the Next Update

The most effective monitoring dashboard: ① Order-to-revenue ratios and order age; ② supplier on-time delivery and qualified capacity; ③ customer equipment-capex execution rather than headline budgets; ④ service revenue on a consistent basis, coverage, and renewals; ⑤ repeat volume-production orders for critical new processes; ⑥ actual export-license approvals/refusals and disclosures of affected revenue. If only one question can be tracked, track “At which state does the delay occur?” rather than broad bullish industry headlines every day.

IX. Evidence Ledger: Source Dates and Independent Verification Status

“Independent verification” means cross-checking different entities or different underlying evidence, not two media outlets reproducing the same press release. Financial figures can be recalculated, but a financial audit cannot replace customers’ independent validation of product performance.

X. References by Source Tier

All of the following were accessible before the research cutoff. Future years in linked titles do not represent future actual results. Reference IDs correspond one-to-one with the main text.

A. Financial Reports, Annual Reports, and Original Company Disclosures

• [E01] ASML Q4/FY2025 investor materials, 2026-01-28.
• [E02] ASML Q3 2026 SEC financial materials, 2026-10-14; company quarterly press release.
• [E03] Applied Materials FY2025 and fourth-quarter results.
• [E04] Applied Materials FY2026 third-quarter results, 2026-08-13.
• [E05] Lam FY2026 year-end results and FY2025 comparatives, 2026-07-29.
• [E06] Lam results for the quarter ended September 27, 2026, released 2026-10-21.
• [E07] KLA FY2025 fourth-quarter and full-year results.
• [E08] KLA FY2027 first-quarter results, 2026-10-28.
• [E09] ASML 2025 annual report: critical suppliers and risks.
• [E10] Zeiss SMT capacity expansion, 2026-07-24.
• [E11] VAT H1 2026 letter to stakeholders.
• [E12] TRUMPF FY2024/25 annual report, EUV business and fiscal year-end explanation.
• [E13] Applied Materials 2026 Proxy, AGS contract structure.
• [E14] Besi Q3 2026 results, 2026-10-22. Note that “2025” in the URL path is not the publication date of these results.
• [E15] KLA service-business update, 2025-11.
• [E16] NAURA 2025 annual report, 2026-04-18.
• [E17] Tokyo Electron FY2026 annual financial results and FY2025 comparatives; financial results for the quarter ended June 30, 2026.
• [E18] ASM CY2025 fourth-quarter and full-year report; ASM Q3 2026 report.
• [E19] VAT Q3 2026 order and sales update, 2026-10-15.

B. Industry Associations and Statistical Methodology

• [E20] SEMI: 2023 equipment sales, 2024-04-10.
• [E21] SEAJ/SEMI: 2024 regional equipment-market table, 2025-04-10.
• [E22] SEMI: actual 2025 equipment sales of USD135.1bn, 2026-04-07.
• [E23] SEMI: OEM equipment-sales forecast, 2026-07-14.
• [E24] SEMI: 300mm fab equipment-spending forecast, 2026-10-13.
• [E25] SEMI official market-information archive: 2026-04-01 edition of the 300mm Fab Outlook.

C. Media and Third-Party Observations

• [E30] Tom’s Hardware: Zeiss expansion and ASML output constraints, 2026-07-26. For contextual cross-reading only; the annual reports and company announcements it cites have been checked separately, and republication is not treated as independent evidence.
• Media reports about domestic EUV prototypes, undisclosed customer yields, and “comprehensive breakthroughs” have not been used as the basis for quantitative conclusions. Source claims, equipment demonstrations, and reproducible volume-production capability must be distinguished.

D. Original Regulatory Texts

• [E26] BIS: 2024-12-02 update to equipment, software, and HBM controls.
• [E27] Dutch government: 2024-09-06 expansion of licensing requirements for advanced manufacturing equipment.
• [E28] Dutch government: 2025-01-15 announcement of expanded licensing for certain metrology/inspection technologies.
• [E29] BIS: 2025-08-29 changes to VEU treatment of foreign-owned fabs in China.

This is the Day05 research working draft. All data are governed by the time anchors in the evidence ledger; unresolved conflicts must not be presented as settled conclusions in the finished report.