SILICON NEXUS

About

What this site is, who it's for, and how it works.

Think of Silicon Nexus as EDGAR for the global semiconductor industry. We pull filings from DART (Korea), MOPS / TWSE (Taiwan), SEC EDGAR (US/foreign filers), and TDnet (Japan), plus Chinese semiconductor news, translate non-English content with AI, summarize the takeaway, and map them to a unified framework so a global investor can read any market without learning a new regulatory vocabulary.

Built by

Silicon Nexus is a solo project born from the frustration of checking DART, MOPS, EDGAR, and TDnet separately while investing in semiconductor stocks.

5 years in software development, 4 years investing in semiconductors. Every line of code — from data pipelines to the frontend — is written and maintained by one person, with automated updates running 5 times daily.

AI is used as an analytical tool. All data is sourced from public filings (DART, EDGAR, MOPS, TDnet, etc.), and calculation logic and methodology are transparently documented below.

DART → EDGAR filing equivalents

Korean disclosure types and the US SEC filings they roughly correspond to. Mappings are approximate — the closest analogue for orientation, not a legal equivalence.

DART (Korean)EnglishEDGAR FORM
사업보고서Annual Report10-K
분기보고서Quarterly Report10-Q
주요사항보고Material Event Report8-K
단일판매·공급계약Sales / Supply Contract8-K
자기주식취득결정Treasury Buyback8-K
유상증자Rights Offering8-K / S-1
공개매수Tender OfferSC TO
합병Merger8-K / S-4
주주총회소집Annual Meeting NoticeDEF 14A
임원·주요주주 소유상황Insider Ownership ChangeForm 4
주식 대량보유Major Holder Report (5%+)Schedule 13D/G
정정공시Amendment / Correction/A
Who it's for
Foreign portfolio managers, analysts, traders, and anyone who needs to track Korean/Chinese/Taiwanese semiconductor stocks but can't read Korean, Chinese, or Japanese efficiently. We focus on chip and chip-adjacent names where filing flow is dense.
What we cover
110+ semiconductor and adjacent stocks across 5 countries. Korea (36): memory (Samsung, SK Hynix), foundry (DB HiTek), equipment (Wonik IPS, Hanmi Semi), materials (Soulbrain, Dongjin Semichem), OSAT (Hana Micron, Nepes). China: news coverage of SMIC, Huawei/HiSilicon, CXMT, and domestic semiconductor ecosystem. Taiwan (27): foundry (TSMC, UMC, VIS), fabless (MediaTek, Realtek, Novatek), packaging (ASE), wafers (GlobalWafers). US (31): NVIDIA, Broadcom, Micron, AMD, Intel, Marvell, Lam Research, Applied Materials, and AI infra/power names. Japan (15): Tokyo Electron, Advantest, Screen, Lasertec, SUMCO, Renesas, Kioxia. Plus ASML (NL).

Legal / disclaimer

For informational purposes only — not investment advice. AI-translated summaries are best-effort and may contain errors. Always refer to the original disclosure on DART/TWSE/MOPS/TDnet for the authoritative version.

Methodology

All Silicon Nexus indicators are grounded in public data and academic research. Full academic basis, honesty grades, and limitations for each indicator are documented on the Methodology page.

Semi Nexus Dual IndexTwo independent 0-100 indices separating real-economy data (6 indicators) from market signals (8 indicators). Cross-reading auto-detects market phase.
HBM Market TightnessWeighted 3-axis composite (0-100). Combines hyperscaler CAPEX, HBM supply allocation, and TSMC packaging load. Higher = tighter supply.
Semi PulseShort-term sentiment index (0-100). Fear-greed gauge combining market temperature, supply chain, and smart money layers.
Cycle Pressure IndexMeasures CapEx sustainability of 4 AI hyperscalers (Alphabet, Microsoft, Amazon, Meta). -50 to +50 scale.
AI Infrastructure TensionFact-axis composite (0-100) — weighted geometric mean of GPU rental prices and DC construction pipeline. CapEx news sentiment is shown separately as a narrative axis (not blended into the fact score), so the fact-vs-narrative gap stays visible. Geometric mean avoids one extreme axis being masked by another (compensability).
Supply Chain TensionWeighted average of customer-supplier temperature gap (T1) and price-volume divergence (T2).
Composite Score4-component weighted sum (0-100): CAN SLIM 40% + Trend 25% + Quality 25% − Risk 10%.
Piotroski F-ScoreBased on Piotroski (2000). 9 binary financial statement quality checks (0-9).
Mohanram G-ScoreBased on Mohanram (2005). 8-point growth stock checklist (0-8) vs sector medians.
CAN SLIMHeuristic application of O'Neil's framework. 7 factors (C·A·N·S·L·I·M) each scored 0-100.
Supply Chain MomentumMeasures momentum of 4 value chain segments (JP materials/equipment → TW foundry → KR memory) relative to their own history.
Narrative vs Fact DivergenceMeasures gap between news sentiment and actual data (exports, DRAM). Segmented by country and keyword.
Hiring SignalsJob postings across 5 chipmakers, dual-axis: market-awareness (leading/coincident vs news) + baseline state (new/stable/surge/drop via 28-day moving average). Observation-only.
DC Construction DemandAI datacenter pipeline (MW × build-stage) propagated to supply-chain nodes (memory/power/networking/packaging). Hyperscalers treated as demand source, not signal.
Technical Indicators (9 families)EMA, Bollinger, VWAP, Ichimoku, RSI, MACD, Stochastic, OBV, ATR. All industry-standard.
Data Sources
🇰🇷 Korea
DART, ECOS, Korea Customs, KRX
🇨🇳 China
Google News CN (Chinese semi news)
🇹🇼 Taiwan
MOPS, TWSE, TAIEX
🇺🇸 US
SEC EDGAR, SEC EDGAR Human Capital, yfinance, Vast.ai
🇯🇵 Japan
TDnet, Kabutan, e-Stat
Market
DRAMeXchange, Epoch AI, GitHub
Jobs
Workday (NVIDIA/Intel/Micron/Broadcom), Amazon Jobs
DC Infra
ai_data_centers (Epoch AI + GitHub + note-extracted MW)
Update
5x daily (KST 07/12/16/20/00)

Questions, feedback, or corrections? See our contact page.