Novatek Microelectronics

Phase A — Understand the business

Company Overview

What it is. Novatek Microelectronics Corp. (聯詠科技), founded 1997 as a spin-off from foundry United Microelectronics (UMC), is the largest merchant (non-captive) display-driver IC designer in the world and one of the ten largest fabless chip companies globally by revenue. It runs a pure fabless model — it designs the ICs and outsources all wafer fabrication to foundries (principally UMC and TSMC) and assembly/test to OSATs.

How it makes money — two legs:

1. Display Driver ICs (DDICs) — the silicon that translates a video signal into the voltages that drive each pixel of an LCD or OLED panel. ~65–70% of revenue in FY2024–25. Split across:

• Large-panel DDICs (TVs, monitors, notebooks) — ~31% of 2024 revenue.
• Small/medium DDICs (smartphones, tablets, automotive, wearables), increasingly touch-and-display integrated (TDDI) and OLED — ~42% of 2024 revenue and the faster-growing sub-segment on OLED adoption.

1. SoC solutions — timing controllers (T-CON), TV/monitor SoCs, image-signal processors, CMOS-image-sensor companion chips, and a growing edge-AI / machine-vision SoC line. ~30–35% of revenue in FY2024–25, rising fast: SoC reached 44% of Q1 2026 revenue and a record 53.7% in April 2026.

Customers. Panel makers and OEMs: Chinese panel giants (BOE, TCL/CSOT, Tianma), Korean/Taiwanese panel and set makers, and — a 2024 diversification win — North American and South Korean smartphone brands. Concentration is by end-market geography (China > 50% of sales) more than by a single named account (see Lens 2). The contract structure is merchant/transactional (design-win + PO), not take-or-pay or long-term recurring — pricing resets each cycle, which is central to the bear case.

Contract/payment structure. Standard fabless: design-win → forecast → purchase orders priced per unit, re-negotiated on foundry-cost and panel-demand cycles. No meaningful recurring/subscription revenue; no take-or-pay backlog protection. ASP is mix-driven — OLED and TDDI drivers carry higher ASP/margin than legacy LCD drivers, so the mix-shift is the margin story.

Supply Chain

Novatek sits in the middle of the display value chain — it neither owns fabs nor sells to end-consumers. The full mapped chain, named:

Upstream (inputs to Novatek):

• Foundries (the critical dependency): UMC — Novatek's corporate parent-by-origin, still its largest single shareholder (~15–25%), and a key wafer supplier occupying "a major share of UMC's HV 40nm capacity". TSMC — for finer-node SoC and advanced OLED-driver work. Novatek's wafers run principally at 40nm–12nm. For AMOLED drivers specifically, the required HV 40nm / 28nm capacity is held ~90% by Samsung Foundry + TSMC + UMC (GlobalFoundries and SMIC the residual) — a genuine chokepoint.
• IP: Novatek licenses foundation IP and develops its own display/touch/imaging IP in-house.
• OSAT (assembly/test): Taiwanese/Chinese OSATs (COF/COG bumping and packaging houses such as Chipbond and Chipmos are the standard DDIC back-end partners in Taiwan).

Novatek (the node): fabless design of DDIC + SoC.

Downstream (Novatek → end demand):

• Panel makers: BOE, TCL CSOT, Tianma, AUO, Innolux, Samsung Display, LG Display, Visionox → who integrate the driver onto the panel/module.
• OEM/brands: smartphone brands (incl. newly-won NA + Korean names), TV/monitor/notebook OEMs, automotive Tier-1s → end consumer.

Chokepoints / single-source risks:

• Foundry HV-node capacity is the binding constraint — DDICs run on specialty mature nodes, not leading-edge, and that capacity is scarce and controlled by a 3-supplier oligopoly. When mature-node capacity tightens (as it did in the 2021 shortage, when TSMC raised 12-inch LCD-driver quotes ) Novatek's costs rise faster than it can re-price.
• China end-demand concentration (>50% of sales) is the single largest downstream chokepoint — a China consumer-electronics slowdown or a panel-maker inventory correction hits Novatek directly (management flagged exactly this in May 2025 — see Lens 5).
• Memory & gold input-cost inflation — 2026 guidance explicitly cites memory shortages and rising gold costs pressuring COGS.

This lens is named, not generic — but note the OSAT layer is industry-inference, not company-disclosed, and should be confirmed against a Novatek annual report before it's load-bearing.

Competitive Advantages (moats)

The moat is real but narrow — it is scale + design-cycle incumbency in a commoditizing product, not a software or network moat.

• Scale / #1 merchant share. Novatek is the largest merchant DDIC supplier and a top-3 supplier across TVs, monitors, notebooks and smartphones, with ~22% of the OLED-driver market and >NT$100B revenue. Scale buys foundry allocation priority on scarce HV mature-node capacity — the single most valuable structural advantage in this business (a smaller rival simply cannot get the wafers in a shortage).
• Design-win switching costs (moderate). A DDIC is co-designed with a specific panel; requalifying a new driver mid-program is costly and slow, so incumbency on a panel platform is sticky for that generation. But it re-opens every product cycle — this is a rolling moat, not a permanent one.
• TDDI + OLED integration IP. Novatek's move up-stack into touch-and-display integration and, in 2025, mass production of mobile OLED TDDI raises ASP and integration complexity, widening the gap vs. pure-LCD-driver rivals (Himax, Sitronix, Raydium, FocalTech).
• Bargaining power — asymmetric and weakening at both ends. Upstream: Novatek is a large, priority customer of UMC (and a UMC affiliate), which helps — but it is still a price-taker when mature-node capacity tightens. Downstream: its buyers are the enormous Chinese panel oligopoly (BOE, CSOT), who have real pricing power over a merchant component supplier. Novatek is squeezed in the middle — strong vs. tiny rivals, weak vs. its foundries and its mega-panel customers.

Vulnerabilities. DDICs are a mature, cyclical, increasingly-commoditized product; Korean captive suppliers (Samsung LSI, LX Semicon) and Chinese domestic drivers (Chipone, ESWIN) are structurally motivated to displace merchant share for supply-chain-sovereignty reasons. The durable-moat question is whether edge-AI SoC can become a large enough, higher-moat second leg before DDIC commoditizes the core.

Segments

Hard-data caveat: our figures is empty — there is no segment data. All splits below are from third-party profiles and must be treated as approximate; Novatek's own annual report is the authoritative source and was not ingested.

By product line (FY2024, approximate):

*The product buckets overlap in different vendor taxonomies (DDIC 65–70% vs SoC 30–35% at book level; the finer 42/31 split is within-DDIC). Treat as directional, not additive.

By geography:

• China > 50% of 2024 revenue (dominant panel-manufacturing + assembly base).
• China + Taiwan + South Korea > 85% combined.
• Asia-Pacific > 90% of total sales. Rest-of-world is a rounding error.

The trend that matters: the book is bifurcating — legacy large-panel LCD DDIC is the decelerating cash cow; SoC (edge-AI/machine-vision) is the accelerating, re-rating leg, now crossing 50% of monthly revenue in early 2026. Whether that is a genuine mix-shift toward a higher-margin, higher-moat business or just cyclical timing is the central analytical question of this name.

---

Phase B — Measure performance

Earnings Result

FY2025 (full year, self-calculated release 2026-02-06):

• Revenue: NT$100.663B, −2.07% YoY (from FY2024's NT$102.79B).
• Net profit after tax: NT$16.348B, −19.64% YoY (from FY2024's ~NT$20.34B).
• EPS: NT$26.87.
• Gross margin: 37.66%, down 2.69pp from FY2024's 40.35%.
• Operating profit: NT$18.237B (op margin ~18.1% ).

The headline: flat-to-down revenue but sharply lower profit — a ~20% earnings decline on a ~2% revenue decline is textbook operating deleverage plus gross-margin compression (mix + input costs), the classic signature of a mature semi name past its cycle peak.

Q4 2025: Revenue NT$22.817B (−7.15% QoQ, −9.68% YoY); net NT$3.69B (+0.96% QoQ, −23.15% YoY); EPS NT$6.07; GM 38.19% (up 1.9pp from Q3's 36.29% — margin bottomed and turned).

Q1 2026 (the recovery print):

• Revenue NT$23.15B, +1.4% QoQ but −14.7% YoY.
• GM 39.06% and op margin 17.4% — both ahead of guidance (guide was GM 36–39%, op margin 15–18% ).
• SoC = 44% of revenue in Q1, 53.7% in April — the mix-shift is the margin lever.

Q2 2026 guidance: Revenue NT$27.5–28.5B (implies +19–23% QoQ), GM 38–41%, op margin 16–19%. A large sequential step-up — the strongest forward signal in the file.

Reaction / what's priced in: despite the guidance-beat and a sharp Q2 ramp guide, the stock sits near NT$384.50 (Apr 2026), −33% over the trailing year and far below its NT$593–614 (May 2024) and NT$656 all-time-high (Apr 2021) levels. The market is pricing the structural deceleration, not the quarterly beat — YoY revenue is still down double-digits and the memory/gold cost pressure caps the margin-recovery story. Balance-sheet flags: none surfaced; Novatek is historically net-cash with a high dividend payout (see Lens 9), which is itself a signal (see verdict).

Earnings Calls (sentiment trend)

No transcripts are on the research shelf; this is ``-sourced sentiment from call summaries across Q4'25 → Q1'26.

Trajectory of management tone:

• May 2025: Cautious/defensive. Management "expects growth to moderate on weakening demand in China, US" — an explicit downgrade of the demand outlook. This is the low-water mark of tone.
• Feb 2026 (Q4'25 call): Cautiously constructive. "Q4 results met guidance with margin gains; 2026 growth to focus on AI and OLED products". The pivot in the narrative — from defending a soft cycle to selling the edge-AI/OLED growth story — dates to here. Management flags "memory and visual edge AI to be the main demand drivers in 2026".
• Apr/May 2026 (Q1'26 call): Increasingly confident on mix, honest on costs. Raised margin outlook "on stronger product mix and early shipments"; guided a big Q2 step-up; but candidly warned that memory-shortage and gold-cost price increases "will not fully offset margin pressure."

Recurring phrases (rising): "edge AI," "machine vision," "OLED TDDI," "product mix," "SoC." Things they stopped saying: the defensive China/US demand-moderation framing that dominated mid-2025 has been replaced by product-led growth language. Net read: tone has improved off the 2025 trough and is now mix-and-AI-led — but management is notably not claiming the cost pressure is behind them, which reads as credible rather than promotional.

Comps

Peer set pulled from research/companies/_index.json (hardware topic) + obvious DDIC/fabless names the index misses (Sitronix, Raydium, LX Semicon, Silicon Works).

Himax read (the one clean comp): FY2025 revenue $832.2M, net $43.9M ($0.25/diluted ADS, down from $79.8M/$0.46 in 2024), GM 30.6%, op margin 5.3% (from 7.5%), fwd P/E ~23x. The contrast is the whole point: Novatek is ~11x Himax's net income, carries a ~700bps-higher gross margin (37.7% vs 30.6%) and ~5x ROE (27.5% vs low-single-digit), yet trades at roughly half Himax's forward multiple (~14.5x vs ~23x) and pays a 7% dividend Himax can't. On quality-adjusted, cash-return-adjusted metrics Novatek is the cheaper, higher-quality asset of the two — the market discount reflects Novatek's lower growth and China exposure, not lower quality.

Where the table is thin: I could source clean multiples only for Novatek and Himax. The Taiwan/Korea DDIC rivals (Sitronix, Raydium, LX Semicon) and MediaTek are n/a rather than guessed. A proper relative-value call needs those pulled from a terminal.

Stock-Price Catalysts

``-sourced price history; Novatek trades as a cyclical, foreign-flow-sensitive Taiwan semi.

• Apr 2021 — all-time high NT$656. Peak of the pandemic display/panel super-cycle + the 2021 mature-node shortage that let DDIC suppliers raise price. This is the reference high — the market's memory of "peak Novatek."
• 2021–2022 — sharp de-rate as the panel/DDIC cycle rolled over (inventory glut, demand normalization post-COVID). The classic semi-cyclical top.
• May 2024 — ~NT$593–614 — a cyclical recovery + AI/edge-AI narrative lift; the last time the stock traded "expensive."
• May 2025 — de-rating catalyst: management's explicit "growth to moderate on weakening China/US demand" guidance marks the top of the recent leg; the stock began its slide into 2026.
• Feb 2026 — FY2025 print (−20% earnings, GM down 269bps) confirmed the deceleration; the stock stayed weak.
• Apr 2026 — ~NT$384.50, −33% YoY despite a Q1 guidance-beat and a +19–23% QoQ Q2 guide — the market is trading the YoY decline and cost pressure over the sequential recovery.

Pattern → what the market actually reacts to: (1) the mature-node foundry-cost / panel-price cycle (2021 shortage = up, 2022 glut = down) — this is the dominant driver; (2) China consumer-electronics demand (the May-2025 warning was the single biggest recent catalyst); (3) the edge-AI/OLED SoC narrative as a swing factor on multiple; (4) foreign institutional flows (Vanguard/BlackRock hold 2–4% each; Taiwan large-caps move on foreign buying/selling). Notably, quarterly beats have NOT been enough to move the stock up while the YoY trend is negative — the market is anchored on the structural story.

---

Phase C — Judge people & books

Management

• CEO & President: Steve Wang (Shou-Jen / Shou-Ren Wang) — Vice Chairman, CEO & President since October 2001 (~24-year tenure). Chairman: Ho Tai-Shung. An exceptionally long, stable operating tenure — Wang has run Novatek across the entire post-spinoff arc from small UMC affiliate to the world's #1 merchant DDIC house.
• Track record — strong, quantified. Under this team Novatek grew from a UMC spinoff to >NT$100B revenue, top-3 global DDIC across all major panel categories, ~22% OLED-driver share, and top-10 global fabless. They executed the up-stack moves that matter: TDDI integration and, in 2025, mobile OLED TDDI mass production — plus the pivot into edge-AI/machine-vision SoC. This is a management team that has repeatedly moved the product mix up the value chain ahead of commoditization.
• Skin in the game / ownership. Founder-affiliated but effectively a professional-management, UMC-anchored structure: UMC remains the largest single shareholder (~15–25%), institutions ~40–55%, general public ~39%. One-share-one-vote governance; 4 of 9 directors independent running audit/comp. Individual insider stakes are modest (professional-manager archetype, not a founder-owner controlling block).
• Capital-allocation history — disciplined, cash-return-led. ROE ~27.5%, ROIC ~18.8% — genuinely high returns on capital, consistent with an asset-light fabless model run well. The defining allocation choice is a very high dividend payout: NT$23/share proposed for FY2025, ~85.6% payout ratio, ~7% yield. Fabless + net-cash + minimal reinvestment need → return cash. This is rational for a mature business — but an ~86% payout is management telling you they don't see enough high-return reinvestment to retain the earnings. That is the honest read of the dividend: it's a maturity signal, not a growth signal.
• Red flags — few. No related-party abuse surfaced (the UMC relationship is disclosed and structural, not a hidden related-party deal). No promotional behavior, no strategy-pivot churn (24 years of the same CEO, same up-stack strategy). Governance screens clean.
• Archetype: seasoned professional operator with a founder-era tenure — the right profile for a mature, cash-generative, execution-driven business. The risk is not competence; it's that even excellent operators can't make a decelerating end-market grow.

Forensic Red Flags

Accounting-risk scan (necessarily limited — no filings on the shelf, no `` figures; this is a web-only screen and should not be read as a forensic audit):

• Revenue recognition: merchant/PO model, point-in-time recognition on shipment — low manipulation risk (no long-term percentage-of-completion or bundled-subscription judgment). Green.
• Margin quality: the FY2025 GM drop (40.35% → 37.66%) is disclosed and cost-driven (mix + memory/gold input inflation), not an accounting artifact — management pre-warned it. Consistent with peer Himax's margin compression, i.e. industry-wide, not company-specific. Green.
• Segment reporting: Novatek's public segment disclosure is coarse (DDIC vs SoC), and third-party segment splits conflict slightly across vendors — a data-quality flag for the analyst, not a fraud flag. Amber (need the annual report).
• SBC / non-GAAP: Taiwan reporters emphasize IFRS statutory figures; no evidence of aggressive non-GAAP adjustment to flatter earnings. Green.
• Cash vs earnings / receivables / inventory: no divergence surfaced, but this could not be verified — FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices., inventory-days and DSO were not on the shelf and were not cleanly sourced on the web. Amber — explicitly unverified. For a China-concentrated component supplier, inventory and receivables to Chinese panel makers are the line items to scrutinize in the annual report before any real capital is committed.
• Goodwill/intangibles: asset-light fabless — minimal goodwill risk. Green.

Regulatory findings (required sub-section):

• SEC (EDGAR): regulatory/regulatory-findings.md confirms Novatek has no CIK — it is not a US filer, so no EDGAR Litigation Releases or AAERs are possible. total_sec_findings: 0 by construction, not by clean bill.
• Non-SEC enforcement: web search on "Novatek Microelectronics" (FTC/DOJ/FDA/consent decree/settlement/fine/penalty) enforcement and a targeted patent/antitrust search (incl. vs. rival Sitronix) surfaced no material regulatory action, antitrust settlement, or enforcement finding. (Care: "Novatek" also names an unrelated Russian gas company PAO Novatek — those hits were filtered out.)
• 10-K Item 3 (Legal Proceedings): not available — Novatek files no 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes.. Taiwan TWSE/annual-report legal-proceedings disclosure was not ingested.
• Verdict: No material regulatory or legal findings surfaced — verified via SEC EDGAR EFTS (no CIK → not applicable), web search (FTC/DOJ/antitrust/patent, null), as of 2026-07-07. Caveat: unaudited by this dossier against Novatek's own TWSE annual-report legal disclosure, which is the authoritative source and was not on the shelf.

---

Phase D — Project & stress-test

Forward Projection

Built bottom-up from FY2025 actuals + FY2026 guidance. Fiscal year = calendar year. Currency NT$. Output ``; every input labeled. No our model create in the watchlist loop.

Anchors:

• FY2025 actual: revenue NT$100.663B, EPS NT$26.87, GM 37.66%, op margin ~18.1%.
• Q1'26 actual NT$23.15B; Q2'26 guide midpoint ~NT$28.0B.
• H1'26 run-rate implies ~NT$51B; management guides FY2026 revenue to grow YoY.
• Shares outstanding ~608M.

FY2026 (base):

• Revenue: ~NT$104–106B. Call NT$105B.
• GM ~38–39%. Op margin ~17.5%.
• Net margin ~16% → net ~NT$16.8B.
• EPS base ≈ NT$27.6.

FY2026 bull: stronger edge-AI SoC mix (SoC already 53.7% in April) drives GM to ~40% and revenue to ~NT$110B → net ~NT$18.5B → EPS ~NT$30.4.

FY2026 bear: China demand double-dips + memory/gold costs bite harder, GM slips to ~37%, revenue flat ~NT$101B → net ~NT$15.0B → EPS ~NT$24.7.

FY2027 (base): modest cyclical growth + continued SoC mix-up → revenue ~NT$110B, GM ~39%, net ~NT$18B → EPS ~NT$29.6. FY2028 (base): ~NT$115B, EPS ~NT$31, assuming edge-AI SoC compounds but DDIC stays flat-to-cyclical.

Three-year base EPS path (fiscal): FY2026 ~NT$27.6 · FY2027 ~NT$29.6 · FY2028 ~NT$31. This is a low-single-to-mid-single-digit EPS CAGR — a mature cash-cow trajectory, not a growth re-rating, unless edge-AI SoC inflects materially above trend. At ~NT$384.50 and base FY2026 EPS ~NT$27.6, the forward P/E is ~13.9x — cheap for the quality, appropriate for the growth.

(Forecast logged conceptually, not committed via our model per the watchlist rule. If promoted to a thesis, the base line to track is: 3034.TW FY2026 EPS ≥ NT$27, p≈0.6, resolves 2026-12-31.)

Bull vs Bear

Bull case. Novatek is the world's #1 merchant display-driver house — an asset-light, ~27% ROE, net-cash cash machine trading at ~13–14x with a ~7% dividend fully covered by earnings. The market is pricing it as a melting ice cube, but it has quietly executed the two up-stack moves that matter: mobile OLED TDDI (ramped 2025, structurally higher ASP) and edge-AI / machine-vision SoC (44% of Q1'26 rev, 53.7% in April — a genuine, accelerating second leg). Q1'26 beat on margin; Q2'26 is guided +19–23% QoQ. If edge-AI SoC re-rates the multiple even modestly toward a growth-SoC comp, you get paid 7% to wait for a mix-driven margin-and-multiple expansion — a high-quality, deeply-cash-generative optionality trade with a fat coupon.

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

1. DDIC is structurally commoditizing, and China wants it in-house. Novatek's core 65–70% of revenue is a mature product that Chinese domestic suppliers (Chipone, ESWIN) and Korean captives (LX Semicon, Samsung LSI) are strategically motivated to localize. With >50% of revenue in China, Novatek is the merchant most exposed to display-supply-chain de-globalization — a slow, permanent share-and-margin erosion, not a cyclical dip.
2. Margin has an input-cost ceiling Novatek doesn't control. FY2025 GM fell 269bps; management says 2026 memory/gold cost increases "will not fully offset" the pressure. Novatek is price-squeezed between an oligopoly foundry supply (UMC/TSMC/Samsung control HV-node capacity) and an oligopoly panel-maker customer base (BOE/CSOT) — structurally weak bargaining power at both ends.
3. The dividend is the tell. An ~86% payout on a fabless business = management signaling it has no high-return reinvestment for the cash. That's fine for a bond-proxy, but it caps the growth narrative — you're buying a yield instrument dressed as a tech stock.

Pre-mortem (it's Jan 2028, the thesis broke — what happened?): Edge-AI SoC turned out to be lower-margin and more competitive than hoped (everyone has a "machine-vision SoC"); China DDIC localization accelerated on geopolitics and took 3–4 points of share; a panel-inventory glut compressed ASPs; EPS drifted to ~NT$24 and the multiple de-rated to ~10x as the market re-classified Novatek as ex-growth. The 7% yield "protected" you into a stock that went nowhere for two years while the TWD/US$ and Taiwan-risk discount widened.

Are multiples too high? No — arguably too low on quality-adjusted metrics (half Himax's multiple at 3–5x the margins and ROE). The multiple is appropriately low for the growth. This is a value/yield name, not a momentum name.

Contrarian view (what the market refuses to see): The market is treating Novatek as a pure cyclical DDIC melting-ice-cube and ignoring that SoC crossed 50% of revenue in April 2026. If that mix genuinely holds, Novatek is quietly becoming an edge-AI SoC company with a DDIC cash cow attached — and it's priced at a DDIC-in-decline multiple. The disagreement: bulls say "it's already an AI-SoC re-rating story;" bears say "it's cyclical timing and DDIC gravity wins." The next 2–3 prints of the SoC mix number resolve it.

Devil's Advocate (short-seller)

Dismantling the bull case:

• What structurally breaks the money machine: the core DDIC product is a mature commodity whose entire cost advantage is foundry allocation Novatek doesn't own — and whose entire end-market (>50% China) has a national-strategic reason to design Novatek out. This is not a moat; it's a lease on incumbency that renews each product cycle and can be revoked by geopolitics.
• Revenue concentration → what if it shifts: if China panel makers accelerate qualification of domestic drivers even 3–4 points/year (entirely plausible under supply-sovereignty pressure), that's a permanent double-digit-percent revenue headwind against which the edge-AI SoC leg is too small to compensate for several years.
• Why the moat is weaker than bulls think: "switching costs" reset every panel generation; the "scale moat" is real only vs. tiny rivals (Himax, Sitronix) and irrelevant vs. captives (Samsung LSI) and subsidized Chinese entrants. The IP moat (TDDI/OLED) is a 1–2 year lead, not a decade.
• Most dangerous competitor bulls underestimate: Chinese domestic DDIC (Chipone/ESWIN) + Korean captive LX Semicon — not Himax. The threat is policy-driven localization, which is price-insensitive.
• Worst capital-allocation read: the ~86% payout isn't shareholder-friendly discipline — it's an admission there's nothing better to do with the money than hand it back, i.e. no organic growth engine large enough to matter.
• What must hold for today's price: that edge-AI SoC mix (44%→53.7%) is durable and margin-accretive, AND that China DDIC localization stays gradual, AND that mature-node input costs don't compress margin below ~37%. Break any one and EPS goes to ~NT$24 and the multiple to ~10x → ~30% downside to ~NT$240.
• If growth disappoints 20–30%: revenue back to ~NT$85–90B with GM ~36% → EPS ~NT$20–22 → at 10–11x, NT$210–240 — a real, non-catastrophic-but-painful de-rate on a name that already pays out most of its earnings.
• Single permanent-impairment scenario & plausibility: structural China DDIC in-sourcing that takes Novatek from #1-merchant to a squeezed-margin, share-losing legacy supplier while SoC stays sub-scale. Plausibility: moderate and rising — the direction is near-certain; only the pace is uncertain. That pace is the whole investment debate.

Management Questions (ordered by information value)

1. Is the SoC mix (44% Q1, 53.7% April) a durable structural shift, and what is the gross margin of the edge-AI/machine-vision SoC line vs. the corporate average? (The single number that resolves bull vs. bear.)
2. What share of your China revenue is on panel platforms where a domestic Chinese DDIC is already qualified or in qualification as a second source?
3. Quantify the FY2026 memory + gold cost headwind in gross-margin points, and how much you can recover via price vs. how much you eat.
4. What is your foundry-capacity allocation outlook on HV 40nm/28nm for 2026–27 — is capacity a growth constraint or is it freely available now?
5. At an ~86% payout, what internal reinvestment IRR bar are you clearing, and under what conditions would you retain more earnings for organic or M&A growth?
6. Which edge-AI SoC end-markets (machine vision, security, automotive, notebook) do you expect to be >NT$5B revenue lines by 2028, and against whom do you compete there?
7. How defensible is your OLED TDDI lead — what's the time-to-parity for Samsung LSI, LX Semicon, and the leading Chinese rival?
8. How dependent is your foundry cost position on the UMC affiliate relationship, and would that survive a change in UMC's capacity priorities or ownership?
9. What is your automotive-DDIC share trajectory vs. Himax (the auto #1), and is auto a real growth leg or a defensive one?
10. How should we think about the through-cycle gross-margin floor for legacy large-panel LCD DDIC as it commoditizes?
11. What is your current inventory and receivables exposure to the top Chinese panel customers, and how do you manage that credit/inventory risk?
12. What would make you pursue a transformational acquisition to accelerate the SoC pivot rather than dividend the cash?
13. How exposed is the business to TWD/US$ and to a Taiwan geopolitical-risk repricing, given >90% Asia revenue?
14. What is the realistic 2028 revenue split between DDIC and SoC you are managing toward?
15. If China consumer-electronics demand stays weak through 2026, which cost levers protect the ~17–18% operating margin?

---