Semiconductor Supply Chain — Deep-Dive Tracker #

Source map: resources/semiconductor-supply-chain.md Created: 2026-05-28 Goal: Push every candidate from the supply chain map through stock-data-analyst verification, then promote survivors to Stage 2 (Graham qualitative scoring + same-day red flag check per shr-020).

Status legend #

Batches #

Excluded ⛔ per supply-chain doc: WOLF, AMS.SW, SMIC, 1347.HK, 6967.T (JIC deal), 285A.T (lockup), CEVA, SLAB, MX, MU/AVGO/NVDA/etc (VWCE overlap).

Batch 1 — Priority shortlist (COMPLETE) #

Meta-finding (Batch 1) #

Zero green passes — the supply chain map was built on stale (Jan 2026 knowledge-cutoff) prices. The actual cycle bottom across the chain ran Aug 2025 (UMC, IMOS) → Jan 2026 (STM, Ibiden, SUMCO, Samsung, ARW). By 2026-05-28 every name had already rerated +91% to +595% off lows. Graham-trough setups existed; the entry window closed 3–10 months ago.

Verdict distribution: 5🔴 (rerated past Graham IV, F6/F7 fail not as cyclicality artifacts but pure pricing), 3🟡 (Samsung structurally best on forward but DEGIRO access via SMSN.L LSE GDR needs spread check; ARW & Screen quality but priced for growth).

Implications for Batch 2–5:

• Names with high AI-correlation (HBM, advanced packaging, leading-edge equipment) likely already rerated identically — expect more 🔴.
• Real value (if any) lives in (a) deep-cycle laggards that haven't followed the broad rally, (b) names with idiosyncratic problems (auto demand, balance sheet, dilution), or (c) the next leg if AI sentiment cracks.
• Watchlist entries (across all 🟡 + 🔴) now define a "second-chance basket" — set price alerts; revisit if drawdown materializes.

Detailed reports #

output/research/sectors/semi-screens/{IMOS,UMC,STMPA_PA,4062_T,3436_T,005930_KS,ARW,7735_T}.md

Batch 2 — L0/L1/L3 extended (COMPLETE) #

Meta-finding (Batch 2) #

Pattern reinforced: every name has rerated past Graham IV regardless of fundamental health. Even names with deteriorating fundamentals (WAF.DE: Q1 2026 gross profit negative, dividend cut 93%) have rallied 80%+ on sector sentiment. The supply chain bull thesis is consensus.

Critical structural finding: DEGIRO does not trade TWSE. Only Taiwan ADRs (IMOS on Nasdaq, UMC on NYSE) are accessible. The following candidates are eliminated by broker access alone: 6488.TW, 2454.TW, 5347.TW, 6239.TW, 2449.TW, 3037.TW, 2344.TW, 2337.TW, 2408.TW, 2379.TW, 8046.TW, 3036.TW (12 names total across Batches 2–5). Recommendation: don't burn compute on these.

Cumulative Batch 1+2 result: 0🟢 4🟡 12🔴 across 16 names. Only one structurally-attractive 🟡 is operationally accessible on DEGIRO (Nordic Semi NOK 145, ARW $150, Screen JPY 7,500-8,000). Samsung 🟡 requires SMSN.L LSE GDR verification.

Batch 3 — L3/L4/L5/L6 #

Batch 4 — L7/L8/L9 #

Batch 5 — L9/L10 tail #

Batch 3 — DEGIRO-accessible names (COMPLETE) #

Meta-finding (Batch 3) #

Pattern is now monolithic and dispositive. 14 of 14 ⛒ 🔴. Names with deteriorating fundamentals rerated (Renesas: loss year + Altium debt; ATS: -1.11 EPS + convertibles + dividend cut). Names with peak earnings rerated (TDK: Q4 EPS missed -39%, still 22% above PT). Names with C-suite selling rerated (POWI, DIOD, AVT all show systematic multi-officer selling at current prices with zero buys). Even VSH (+243% YTD) — explicitly identified by the DIOD agent as the cheaper auto-IDM expression — rerated equally hard.

The semiconductor sector is in late-cycle euphoria, not mid-cycle recovery. The supply chain map identified the right businesses for the wrong time.

Final cumulative result — 30 screened + 10 pre-excluded #

Stage 2 shortlist — RECOMMENDATION: NONE #

Do not deploy capital from the satellite into semiconductor names at current prices. The exhaustive 30-name screen returned zero Graham-defensive PASS results. The four 🟡 names are good businesses without sufficient margin of safety. Per shr-018 (EV/EUR comparison), the next EUR 250 monthly contribution is better deployed into:

1. AKE.PA T2 (trigger EUR 52, intact thesis, 5.7% yield)
2. Or VWCE per shr-019 + the satellite-funnel rule (feedback_satellite_to_vwce_funnel)
3. New pipeline candidate from next cycle's screen

Watchlist alerts — set these now (the "second-chance basket") #

For each name, the watchlist trigger represents either: (a) the Graham IV price at conservative growth, (b) the analyst consensus mean PT minus 15% MOS, or (c) the pre-rally base price. These are entry-prices NOT predictions. Most require 30–70% drawdowns that would only materialize on a sector de-rate / AI sentiment crack.

Highest priority (per Graham fit + accessibility):

Cyclical-laggard candidates if AI sentiment cracks:

Excluded / parked #