ASE Technology

Phase A — Understand the business

Lens 1 · Company Overview

Business model. ASEH sells manufacturing services, priced per unit of work, not silicon. In packaging, the customer consigns the wafer — "the cost of the silicon die, typically the most costly component of the packaged semiconductor, is usually not reflected in our costs (or revenues) since it is generally supplied by our customers on a consignment basis". That is the whole margin story: ASEH's revenue is the value-add (bumping, RDL, molding, test time), so a modest revenue line sits on top of enormous customer-owned die value flowing through its fabs. Testing is priced by CPU-seconds of automated-test-equipment time; EMS and packaging by cost-plus vs. prevailing market price.

FY2025 scale:

• Operating revenues NT$645,387.7M = US$20,573.4M (+8.4% YoY vs NT$595,409.6M in 2024)
• Gross profit NT$114,192.9M = US$3,640.2M, gross margin 17.7% (up from 16.3%)
• Profit from operations NT$51,420.8M = US$1,639.2M, 8.0% margin (up from 6.8%)
• Net profit to owners NT$40,015.7M = US$1,275.6M (+23.6%)
• Diluted EPS NT$8.75/share; diluted EPS per ADS NT$17.49 (US$0.56) — one ADS = two common shares

Products/services. Packaging spans commodity wirebonding through the leading-edge stack (below); testing spans wafer probe, final test, SiP/module test, and system-level test; interconnect materials = in-house IC substrate production (supplies ~8.7% of its own substrate needs by value); EMS via USI = motherboards, AI cards, Wi-Fi/SiP modules, automotive electronics.

Customers. Five largest customers = 46.5% of revenue in 2025 (48.4% 2024, 48.0% 2023 — modestly de-concentrating). One customer (incl. OEM-on-behalf) >10% of revenue in each of 2023–25. customers.csv is empty on the shelf; the 20-F does not name the >10% account, but the AI-packaging trade press points to NVIDIA/AMD as anchor advanced-packaging customers. 48% not disclosed — treat the specific names as ``.

Suppliers. Substrates, leadframes, bonding wire (gold, being displaced by cheaper copper), molding compound. Wafers are consigned (no inventory cost). EMS raw materials = 78.7% of EMS revenue (PCBs, ICs, components) — the mechanical reason EMS margin is single-digit.

Contract structure. No take-or-pay disclosed. Orders confirmed "several weeks before delivery" against blanket orders + rolling forecasts — i.e., short-cycle, not long-dated backlog. This matters for the bear case: the AI order book is a forecast, not a contracted annuity.

Lens 2 · Supply Chain

ASEH sits in the back-end, between the foundry and the system OEM:

Upstream inputs → ASEH:

• Consigned wafers ← foundries (TSMC dominant; also Samsung, UMC, GlobalFoundries) and IDMs. This is the critical dependency: ASEH's advanced-packaging volume is a derivative of foundry front-end output.
• IC substrates ← Unimicron, Nan Ya PCB, Ibiden, Shinko (external) + ASEH's own interconnect-materials unit (~8.7% self-supply). Substrate is a known industry chokepoint for advanced packaging.
• Bonding wire (gold/copper), leadframes, molding compound, HBM stacks (customer-consigned) ← materials suppliers.
• Test equipment ← Teradyne, Advantest (testers, partly leased to manage utilization); bonders/steppers ← ASM Pacific, BESI, Applied Materials.

ASEH → downstream:

• Packaged/tested devices → fabless designers (NVIDIA, AMD, Broadcom, MediaTek, Qualcomm) and IDMs → then to hyperscaler buyers (Microsoft, Amazon, Google, Meta) inside AI servers, and to smartphone/auto/industrial OEMs.
• Drop-shipment: ASEH ships tested parts directly to the customer's designated end user, bypassing customer inspection — a stickiness mechanism.

Chokepoints & single-source dependencies:

1. The TSMC-overflow relationship is the whole AI thesis. TSMC's CoWoS capacity runs a persistent 15–20% supply/demand gap; TSMC outsources on-substrate assembly for NVIDIA GPUs to ASE and Amkor. ASEH is tripling its CoWoS-equivalent capacity to 25,000 wafers/month and breaking ground on six new plants in the US, Malaysia, Japan, Germany, Taiwan in 2026.
2. Substrate supply is the recurring back-end bottleneck; ASEH's partial vertical integration is a hedge but it is still net-short substrates.
3. Geographic concentration in Taiwan — most capacity, and the cross-strait (ROC/PRC) risk the 20-F flags repeatedly, is the single largest tail risk to the chain.

Names present → this lens passes. The chain's defining feature: ASEH is a pure back-end play whose AI upside is gated by TSMC's front-end/CoWoS output and by substrate supply — it does not control either.

Lens 3 · Competitive Advantages (moats)

Moat 1 — Scale as the #1 OSAT. FY2024 OSAT share: ASE 44.6% (US$18.54B), Amkor #2 (US$6.32B), JCET #3 (US$5.0B), PTI/Powertech #5 (US$2.28B). ASE is ~3× the #2. The 20-F's own cost argument: "we benefit from specialization and economies of scale by providing services to a large base of customers across a wide range of products... better able to reduce costs and shorten production cycles through high-capacity utilization". Scale in a high-fixed-cost, utilization-driven business is a genuine, self-reinforcing moat.

Moat 2 — Advanced-packaging technology leadership + qualification lock-in. ASEH's leading-edge toolbox — FOCoS, FOCoS-Bridge (3,000–14,000 leads), 2.5D/3D IC (TSV), Co-Packaged Optics, FOWLP, FOPoP — is explicitly aimed at "ASICs and HBM for HPC, networking, server and AI/ML" and "AI accelerators for AI training". Each customer must qualify and correlate every facility (weeks-to-months per site) — a switching cost that gets stickier as packages get more complex. Advanced packages (Bumping/FC/WLP/SiP) rose to 59.6% of packaging revenue (from 51.3% in 2023) — the mix shift is the moat compounding.

Moat 3 — Bargaining power, asymmetric by product. In leading-edge AI packaging, ASEH currently holds pricing power — it raised advanced-packaging quotes >20% in mid-2026 because capacity is a hard constraint. In commodity wirebond and EMS, it is a price-taker in a "highly competitive" market. So the moat is concentrated in the AI-relevant sliver, not the whole P&L.

The moat's limit (critical): ASEH's most dangerous competitor is not another OSAT — it is its own supplier, TSMC, whose InFO and CoWoS are in-house advanced packaging. The 20-F names it directly: "TSMC has offered advanced packaging technologies such as integrated fan-out (the 'InFO') technology". ASEH's advanced-packaging share is, in large part, TSMC's spillover — a moat rented from the leader, not owned outright.

Lens 4 · Segments

Segment revenue and gross margin, FY2023→2025:

ATM (Packaging+Testing) blended GM = 23.8% in 2025 (22.9% 2024). EMS GM = 9.2%.

The single most important structural fact in the whole dossier: ASEH is two businesses fused for reporting. ATM is a rising-margin, AI-levered ~59% of revenue growing double-digits; EMS is a shrinking, single-digit-margin ~40% anchor tied to "slow recovery of general communication and automotive products". Consolidated gross margin (17.7%) badly understates the AI business (ATM 23.8%, testing 31%). Any bull-case multiple must be applied to ATM, not the blended entity — and any bear case leans on EMS diluting both the growth rate and the margin the market is willing to pay for.

Geography (by customer HQ): US 56.7% (falling from 63.6% in 2023), Taiwan 18.0% (rising), Asia 15.2%, Europe 10.0%. Export sales = 82.0% of total. US demand is the center of gravity — an AI-demand tell, but also a tariff/policy-exposure tell.

End-use mix (ATM): Computing jumped to 24.0% (from 18.1%) as AI ramped; Communications fell to 45.8% (from 50.8%); Consumer/industrial/auto 30.2%. The computing surge is the AI signal inside the segment data.

---

Phase B — Measure performance

Lens 5 · Earnings Result

Latest audited full year (FY2025, 20-F) — the print vs. the company's own history:

• Revenue +8.4% to US$20,573.4M. Composition tells the story: Packaging +17.8%, Testing +31.8%, EMS −5.2%. AI-levered ATM carried a soft EMS line.
• Gross margin 17.7% (+140bp), driven by "higher packaging and testing revenue mix and higher factory utilization," partly offset by an appreciating NT dollar and higher utility costs.
• Operating margin 8.0% (+120bp); operating profit +27.5% — operating leverage working.
• R&D +13.9% to US$1,047.2M (5.1% of revenue, up from 4.8%) — spending into leading-edge advanced packaging.
• Net income to owners +23.6% to US$1,275.6M.
• A drag to note: net non-operating swung from +NT$1,394.4M income (2024) to −NT$117.7M expense (2025) on lower FX-hedging gains + higher finance costs; effective tax rate rose to 19.70% (from 18.97%) partly on Pillar Two.

Latest quarter (Q1 2026): Net revenues NT$173,662M, +17.2% YoY, −2.4% QoQ (normal Q1 seasonality). ATM record NT$112.4B, +29.7% YoY. Reported EPS US$0.195, beat the US$0.17 consensus; stock +1.7% on the print. (A separate feed showed a spurious "EPS $0.0063 miss" — a per-ADS/per-share and currency-unit artifact; the $0.195 beat is the correct read. Conflict surfaced, not silently resolved.)

Balance-sheet flags: Cash & equivalents US$2,947.7M + current financial assets US$247.2M; total debt US$8,700.8M (US$1,381.2M short-term, US$6,824.4M long-term); unused credit lines US$12,770.7M; working capital US$2,082.1M. Net debt ≈ US$5.75B. Operating cash flow US$4,534.6M; capex US$5,470.7M (PP&E basis) — capex exceeded operating cash flow, so FY2025 free cash flow was negative. Financing activities flipped to +US$1,443.1M (net borrowing) vs. −US$233M-equivalent outflow in 2024. This is the number the bull case must own: the AI ramp is not self-funding at current capex.

Lens 6 · Earnings Calls (sentiment trend)

transcripts/ is empty on the shelf — this lens is ``, labeled. Across the FY2025 call and the Q1/Q2 2026 calls the tone has moved from cyclical-recovery to capacity-constrained boom:

• Recurring, escalating phrase: "demand exceeds supply." COO Tien Wu (media, 2026): the US$8.5B 2026 capex "will likely need to be revised up again"; advanced-packaging revenue "will double this year".
• New in 2026: explicit AI/LEAP framing. Management coined "LEAP" (Leading-Edge Advanced Packaging) as a reported metric and lifted the 2026 target repeatedly — from ~$3.2B to >$3.5B (~+118% YoY). (Source conflict: $3.2B and $3.5B both appear across dates; the trajectory — a raised target, roughly doubling — is the signal.)
• Stopped saying: the 2023–24 "slow recovery / soft general demand" language now applies only to EMS; ATM commentary is uniformly bullish.
• Capital-intensity honesty: management is openly guiding to a second major capacity ramp in 2027 and ~15 new factory projects for 2029–30 demand — signaling this is a multi-year capex supercycle, not a one-year pop.

Sentiment trajectory: decisively more bullish, and increasingly supply-constrained — the classic late-cycle tell that is bullish for near-term pricing and a yellow flag for eventual over-capacity.

Lens 7 · Comps

OSAT / advanced-packaging peer set. Multiples `` with source; no multiple fabricated — gaps marked n/a.

Lens 8 · Stock-Price Catalysts

`` throughout. What moves this stock:

• The whole 2025–26 re-rating is one macro trade: AI advanced-packaging scarcity. ASX is +230% over the trailing 12 months (+308% on a 52-week-change basis per stockanalysis.com; +232.8% per Yahoo — same order of magnitude). This is not an earnings-multiple story quarter-to-quarter; it is a narrative re-rating from "cyclical Taiwanese OSAT" to "AI infrastructure gatekeeper."
• Specific up-catalysts (>5% type moves): (1) TSMC CoWoS-overflow / NVIDIA-AMD outsourcing headlines; (2) monthly Taiwan revenue prints beating (May 2026 revenue drove a jump); (3) the >20% advanced-packaging price-hike report; (4) capex raises reframed as demand signals ($7B→$8.5B); (5) the Analog Devices Penang acquisition (announced Oct 2025, closing H1 2026) as capacity/diversification.
• Down-catalysts: NT-dollar appreciation (a real 2025 margin headwind, called out in the 20-F), EMS weakness, any AI-capex-digestion scare, cross-strait geopolitical headlines.
• Pattern: the market reacts to AI-packaging supply/demand and pricing far more than to consolidated EPS. The EMS 40% is largely ignored on the way up — which is exactly the asymmetry that reverses in a downdraft.

---

Phase C — Judge people & books

Lens 9 · Management

• Founders still run it, with real skin in the game. Chairman & CEO Jason C.S. Chang and Vice Chairman/President Richard H.P. Chang founded ASE in 1984. Jason Chang is the largest individual shareholder at ~25% of shares; insiders (incl. founding family) hold ~28%; institutions ~43%. Founder-led with a quarter of the equity is a strong alignment signal — this is an owner-operator, not a caretaker.
• Track record: built the world's #1 OSAT from a Kaohsiung startup; executed the contested SPIL acquisition (Siliconware) to consolidate the Taiwanese OSAT industry and formed the ASEH holding structure in 2018; bolted on USI/EMS for systems-level reach; serial capability M&A (HCC 2023 automotive; Infineon's Philippines + South Korea plants 2024 for power packaging; Analog Devices Penang 2025–26).
• Capital allocation: aggressive reinvestment — capex NT$171.6B (US$5.47B) in 2025, ~2× the NT$96B in 2024, funded partly by new borrowing (see Lens 5); steady dividend (NT$6.6/share declared 2026-03-27, ≈ NT$13.2/ADS; US$0.417 ADS cash dividend ex-date 2026-07-06 ); disciplined bolt-on M&A at appraised prices. ROE ~13.6% — respectable but not spectacular for the capital deployed, reflecting the capital-intensity and the EMS drag.
• Red flags (specific, from the filing):
• Related-party real-estate dealings with Hung Ching — ASEH repeatedly buys jointly-constructed plants/land from Hung Ching (NT$5,263M in 2024; a 72.15% building interest for NT$4,231M in Nov 2025) at "independent appraisal" prices. Appraised and disclosed, but a recurring related-party property pipeline warrants monitoring.
• COO Tien Wu insider-trading matter (see Lens 10) — a director/COO was indicted over SPIL-tender-offer trading; acquitted at first instance, but the case has bounced through appeals. Governance overhang on a key operator.
• Archetype: founder-owner in an industrial-scale, capital-intensive business — implies long-horizon capacity bets over short-term margin optics, and a controlling family whose interests dominate minority holders' (a double-edged alignment).

Lens 10 · Forensic Red Flags

Grounded in the 20-F + regulatory/regulatory-findings.md. Every figure labeled.

Accounting-risk scan:

• Revenue recognition: low risk on the packaging side — consigned wafers mean revenue is service value-add, hard to inflate; short order-to-delivery cycle. EMS is higher-volume pass-through (raw materials 78.7% of EMS revenue) which optically inflates the revenue line relative to economic value — a reason blended margins look thin, not a manipulation flag.
• Goodwill/intangibles: goodwill NT$52,541.9M (US$1,674.9M) as of 2025-12-31; a NT$132.8M (US$4.2M) goodwill impairment was taken in H1 2025 on an underperforming CGU. Impairment testing uses 11.96%–13.81% pre-tax discount rates — disclosed and conservative. SPIL PPA effects still add ~NT$2,955.5M/yr of D&A — a real, ongoing non-cash drag from the acquisition, transparently disclosed.
• Cash flow vs. earnings: OCF US$4,534.6M vs. net income US$1,313.2M — OCF comfortably exceeds earnings (D&A US$2,149.8M is the bridge). Healthy quality of earnings. The divergence to watch is capex > OCF (negative FCF), not earnings quality.
• Receivables/inventory vs. revenue: not flagged as outrunning revenue in the MD&A; working-capital swing was a NT$31.1B source of cash in 2025.
• SBC: share-based payment (employee compensation, restricted stock, options) dilutes — diluted share count 4,429.6M vs. basic 4,341.2M (≈2% overhang); flagged as an operating-expense driver. Modest.
• Debt/leverage: total debt US$8.7B against US$1.6B operating profit; variable-rate exposure (1.70%–4.48% LT). Manageable but rising as the ramp is debt-funded.
• FX: revenue mostly USD, costs mostly NT$/RMB/JPY/KRW/EUR — a strengthening NT dollar directly compresses margin (an explicit 2025 headwind). Structural, hedged imperfectly.

Regulatory findings (required sub-section):

• SEC: No SEC Litigation Releases and no AAERs naming ASE Technology in the 2021-07-06 → 2026-07-06 window, per EDGAR EFTS (LR + AAER).
• 10-K/20-F Item — Legal Proceedings:

1. SPIL insider-trading matter (Dr. Tien Wu, COO): a 2017 indictment alleged Wu tipped a friend ahead of the SPIL tender offers/MOU. Kaohsiung District Court found him NOT GUILTY (Feb 5, 2020); prosecutors appealed; High Court rejected (Jun 2021); Supreme Court reversed and remanded (Jan 2022); case retried on remand. Company states it strengthened internal controls; no other directors party. A related SFIPC "Director Removal Case" (Oct 2018) sought to remove Wu from the board. Ongoing governance overhang — not resolved cleanly.
2. K7 plant wastewater discharge: a 2013 Water-Pollution-Control-Act fine (originally NT$102M) that ping-ponged through appeals and was settled July 2024 — ASEH pays a token NT$0.3M fine + a NT$50M voluntary water-R&D contribution; the NT$47M reimposed fine was refunded. Effectively closed, immaterial.

• Non-SEC (web search — FTC/DOJ/FDA/consent-decree/settlement): no material enforcement hits found in the trade/news record for ASE Technology as of 2026-07-06. (Environmental/labor items are routine industrial matters, not enforcement of the material kind.)

Net forensic read: clean accounting, high earnings quality, transparent disclosure. The two real flags are governance (founder-control + the Tien Wu insider-trading saga + recurring Hung Ching related-party property deals), and capital structure (negative FCF, rising debt to fund the ramp) — not accounting integrity.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Bottom-up from FY2025 actuals + 2026 guidance. Output ``; every input labeled. FY = calendar year (Dec-end). No forecast.ts create (watchlist mode).

Anchors:

• FY2025 net income to owners US$1,275.6M; diluted EPS/ADS US$0.56.
• 2026 capex US$8.5B, +60% vs 2025's $5.3B — a leading indicator of 2027 revenue.
• LEAP (advanced packaging) revenue doubling to >$3.5B in 2026 (from ~$1.6B).
• Sell-side consensus: earnings +33.2%/yr, revenue +17.8%/yr, EPS +33.9%/yr forward.

FY2026E (base):

• Revenue ~US$24.0–24.5B — ATM +~25%, EMS flat-to-slightly-up off a weak base.
• Blended gross margin ~18.5–19%. ATM GM pushing toward ~25%.
• Operating margin ~9%.
• Net income to owners ~US$1.6–1.75B; diluted EPS/ADS ~US$0.72–0.78.

FY2027E (base): Revenue ~US$28–29B; EPS/ADS ~US$0.95–1.05.

FY2028E (base): Revenue ~US$32–34B; EPS/ADS ~US$1.15–1.30.

Scenario paths (EPS/ADS):

All ``, arithmetic from the anchors above. Key sensitivity: at ~$41.87/ADS, base-case FY2026E EPS ~$0.75 implies a forward P/E of ~56x on the ADR — richer than the ~35x "forward P/E" quoted by data aggregators, which are averaging in higher out-year consensus. The gap means the market is discounting the FY2027–28 ramp, not FY2026.

Lens 12 · Bull vs Bear

Bull case. ASE is the indispensable physical chokepoint of the AI hardware buildout. TSMC cannot make enough CoWoS; the 15–20% supply gap flows to ASE and Amkor, and ASE — 3× larger than Amkor, with the full FOCoS/2.5D/3D/CPO toolbox already qualified on NVIDIA/AMD parts — is the primary overflow valve. Advanced-packaging revenue is doubling in 2026 and management is guiding to a second capacity ramp in 2027 and factory projects for 2029–30. It has pricing power (raised quotes >20%) precisely because capacity is the binding constraint of the entire AI supply chain. The mix shift (advanced packaging 51%→60% of packaging; computing 18%→24% of ATM; testing +32%) is structurally re-rating the margin profile. Founder-owner with 25% skin in the game reinvesting into an undersupplied secular boom. The multiple (mid-30s fwd P/E, 25x EV/EBITDA) is defensible if you believe the AI-packaging TAM compounds for 3+ years and the EMS anchor eventually gets spun/shrinks as a share of the whole.

Bear case (2–3 permanent-impairment risks):

1. The moat is rented from TSMC. ASE's advanced-packaging share is largely TSMC's spillover. If TSMC's own CoWoS/InFO capacity catches up (it is expanding aggressively, incl. in the US), the overflow that is the AI thesis shrinks. ASE does not control the front-end that gates its own upside.
2. AI-capex digestion + advanced-packaging over-capacity. ASE, Powertech, KYEC, Amkor and Samsung are all pouring record capex into advanced packaging (industry OSAT capex ~NT$370B in 2026). A back-end business is utilization-driven with high fixed costs — the 20-F is explicit that margins collapse when utilization drops. An AI-spending pause in 2027–28 lands on top of a just-built capacity wall. This is the classic OSAT cycle, and it has always mean-reverted.
3. The 40% EMS anchor + one >10% customer. Nearly half the company is a shrinking, 9%-margin contract-manufacturing business that dilutes both growth and the multiple. And a single customer >10% of revenue means the whole AI thesis rides on one order book (widely believed NVIDIA) that is a rolling forecast, not a contracted annuity.

Pre-mortem (18 months out, thesis broke): 2H2027 — NVIDIA/hyperscaler advanced-packaging orders are digested after the Rubin pull-forward; TSMC's expanded CoWoS absorbs demand it previously outsourced; ASE's six-new-plant capacity comes online into a softening market; utilization drops, advanced-packaging pricing gives back the 20% hike; NT$ strength compounds the margin hit; EMS is still soft. The stock, which had priced FY2028 perfection at +230%, de-rates hard as it reverts from "AI gatekeeper" narrative back to "cyclical OSAT" multiple.

Are multiples too high? On FY2026 base EPS the ADR is ~55x — yes, rich; it only works on the FY2027–28 ramp. The single sharpest market signal: the average of 27 analyst price targets is ~$34, versus a ~$42 price — the sell side, in aggregate, does not justify today's price on its own numbers.

Contrarian view (what the market refuses to see): the bull consensus treats ASE as a secular AI compounder and is ignoring that (a) its edge is structurally subordinate to TSMC's front-end decisions, and (b) OSAT is the most reliably cyclical link in the semi chain — this is a cyclical stock wearing a secular multiple. Conversely, the bear consensus underweights how genuinely hard advanced-packaging qualification and scale are — even in a digestion year, ASE keeps the #1 share and the AI mix is permanently higher than pre-2024. The truthful position is neither a secular compounder nor a value trap — a cyclical leader in the up-phase of a supercycle, priced as if the up-phase never ends.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• What structurally breaks the model: OSAT is a utilization + fixed-cost business. The 20-F says it plainly — "these costs do not decline when customer demand or our capacity utilization rates drop... A relatively modest increase or decrease in revenue can have a significant effect on our operating margins." ASE is building six new plants + ~15 factory projects into an AI-capex cycle that has never not mean-reverted. When it turns, the just-built capacity is a margin anvil.
• Revenue concentration: one customer >10%, top-5 = 46.5%. If the anchor AI account (widely presumed NVIDIA) reallocates to TSMC in-house CoWoS, to Amkor's new Arizona fab (co-located with TSMC Arizona), or simply digests, the AI revenue that drove the +230% re-rating is a forecast that can be cut with a phone call.
• Why the moat is weaker than bulls think: ASE's advanced-packaging position exists because TSMC is capacity-short. That is a supplier-created, temporary rent, not a durable moat. TSMC is expanding CoWoS faster than anyone; Amkor is building next to TSMC Arizona to capture US on-shore AI packaging; Samsung is pushing its own turnkey. The most dangerous competitor is ASE's own biggest supplier.
• Worst capital-allocation reads: funding a US$8.5B capex year with negative free cash flow and new debt at the top of a hype cycle; a recurring related-party property pipeline (Hung Ching); a controlling family (25%) whose interests can diverge from minorities.
• What must hold for today's price: advanced-packaging demand keeps doubling and pricing power persists and the six-plant ramp fills and EMS doesn't drag and NT$ doesn't strengthen — simultaneously, for 2+ years. At ~55x FY2026 base EPS and a price above the average target, that is priced for flawless execution.
• Growth disappoints by 20–30%: if FY2027 revenue grows ~12% instead of ~20% and utilization softens, EPS could stall near ~$0.70 (bear path) and the multiple compresses from ~35x to ~18–20x (PTI-like) — a plausible 40–50% ADR drawdown on multiple + estimate reset combined.
• Single scenario that permanently impairs: not bankruptcy (balance sheet is fine) — the impairment is to the narrative: TSMC in-sourcing + industry over-capacity converts ASE back to a mid-teens-P/E cyclical, permanently un-winding the AI-gatekeeper premium. Plausibility: moderate over 18–30 months; low over the next 6–12.

Lens 14 · Management Questions (ordered by information value)

1. What share of your advanced-packaging (LEAP) revenue is TSMC-outsourced CoWoS overflow versus packaging you win directly and independently — and how does that split change as TSMC's own CoWoS capacity expands through 2027? (This is the whole thesis — is the moat owned or rented?)
2. Your single >10% customer: what is that customer's share now, its trajectory, and how contractually committed (vs. rolling-forecast) is the AI order book behind it?
3. At what industry-wide advanced-packaging utilization rate does your incremental margin turn negative — and what is your plan for the six new plants if AI capex digests in 2027–28?
4. You raised advanced-packaging quotes >20%. How durable is that pricing once your own new capacity and competitors' capacity come online — and what did you assume for pricing in the 2027 ramp plan?
5. FY2025 free cash flow was negative and the ramp is part-debt-funded. What is the FCF and net-debt path through the 2026–27 capex peak, and what utilization sustains the dividend?
6. EMS is ~40% of revenue at ~9% margin and shrinking. What is the long-term strategic case for keeping USI/EMS inside the holdco rather than spinning it to re-rate the ATM business?
7. How much of the 2026 gross-margin improvement is mix/utilization (structural) versus the NT-dollar/pricing (cyclical) — i.e., what is the sustainable through-cycle ATM gross margin?
8. Co-Packaged Optics: what is the realistic revenue timeline, and does CPO widen or narrow your moat versus TSMC and versus optical-module incumbents?
9. Where are you most substrate-supply-constrained, and does the ~8.7% in-house substrate self-supply rise enough to matter as a bottleneck hedge?
10. What is the ROIC you underwrite on the US$8.5B 2026 capex and the 2027 second ramp, and over what payback period?
11. Geographic diversification (US, Malaysia, Japan, Germany) adds cost vs. Taiwan. What is the margin drag of the "China+1 / geopolitical-resilience" footprint, and who pays for it — you or the customer?
12. What is the resolution status and financial/governance exposure of the Dr. Tien Wu insider-trading remand and the Director-Removal case?
13. Please quantify the recurring Hung Ching related-party property transactions and the governance process that sets those appraised prices.
14. How exposed is the 2026–27 plan to US tariff / export-control policy given 56.7% of revenue is US-headquartered customers and you are building US capacity?
15. What is the succession plan for founder-Chairman/CEO Jason Chang, and how is the ~25% family stake governed relative to minority holders?

---