Core philosophy: Why → What → How
Every section consists of three questions:
| Question | Role | Example |
|---|---|---|
| 왜 | Source of demand, reason for existing | "Why do customers pay?" |
| What | Specific business/product | "Which products generate revenue?" |
| How | Execution method, competitive advantage | "Why this company rather than a competitor?" |
❌ BAD: "NVIDIA makes AI GPUs, and data center revenue is 88%"
✅ GOOD: "Training AI models requires trillions of parallel operations. CPUs process sequentially, but GPUs compute simultaneously with thousands of cores. Training GPT-4 on a CPU would take decades, while on NVIDIA GPUs it takes weeks. This is why hyperscalers pour tens of billions of dollars into NVIDIA. Because falling behind in the AI race threatens their business."
Technical term explanation rules (required)
Three-tier explanation system:
| Tier | Location | Method | Example |
|---|---|---|---|
| Immediate explanation | First appearance of the term | Definition in parentheses | "PEG (P/E divided by earnings growth rate; below 1 may indicate undervaluation)" |
| Contextual explanation | When numbers are mentioned | Comparison/meaning | "75% is twice the industry average (40-50%), meaning strong pricing power" |
| Glossary | End of report | Alphabetical order | All technical terms + detailed explanations |
Required section structure
1. Executive Summary
- One-sentence position: "This company is the [role] of [industry]"
- Why should we pay attention now?
- Three-line summary of the core investment thesis
2. Source of demand: Why is this business necessary? ⭐ Core
Must begin with "Why." The most important section.
- What problem is the customer trying to solve?
- Why does this company's product solve that problem?
- Why this company rather than the alternatives?
예시 — NVIDIA:
고객의 문제: AI 모델 학습에 엄청난 연산량 필요
해결책: GPU의 병렬 연산 (CPU 대비 100배 빠름)
왜 NVIDIA: CUDA 생태계 20년 축적, 전환비용이 수백만 달러
3. Business model: What does it make money from?
3.1 Revenue structure — 🔷 Sankey diagram required (references/sankey_guide.md)
3.2 Revenue growth rate table — 🔷 Required
| Segment | FY23 | FY24 | FY25 | YoY% | Change in share |
|---|---|---|---|---|---|
| Data center | $150억 | $475억 | $1,152억 | +142% | 56%→88% |
| Gaming | $90억 | $101억 | $114억 | +13% | 34%→9% |
| Professional visualization | $15억 | $16억 | $19억 | +19% | 6%→1% |
| Automotive | $9억 | $11억 | $15억 | +36% | 3%→1% |
3.3 Customer base: Who pays? How stable are the relationships?
3.4 Margin Structure: How much is left over? Why more/less than others?
4. Position Within the Industry
4.1 Value Chain Position
[설계] → [제조] → [패키징] → [시스템] → [클라우드] → [앱]
│ │ │ │ │ │
NVIDIA TSMC Amkor Dell AWS OpenAI
4.2 Competitive Landscape: Who is after the same pie?
4.3 One-Sentence Position: "This company performs [role] in the [value chain], and is irreplaceable because of [reason]"
5. Economic Moat (Review All 4)
| Moat Type | Key Question |
|---|---|
| Switching Costs | Is it hard for customers to leave? Why? |
| Network Effects | Does value increase as users grow? |
| Intangible Assets | Brand/patent/regulatory barriers? |
| Cost Advantage | Can it produce more cheaply than competitors? |
5.5 Overall Assessment: Moat width, sustainability, key threats
6. Company History
- 6.1 Founding and Early Vision: What problem was it trying to solve?
- 6.2 Key Turning Points: Which decisions created today's position?
- 6.3 Recent Changes
7. Competitive Victory Story
- Which competitors were defeated in the past, and how?
- Do the factors behind those victories remain valid today?
- What are the new threats?
8. Future Outlook
- 8.1 Growth Drivers: What is raising revenue? 왜?
- 8.2 Key Risks: What could break the story?
9. Bull / Bear / Base Case — 🔷 Trendline graph required
Narrative-Based Forecast: Assumptions → Logical Chain → Outcome
Each scenario should be logically derived from the previously written narrative (sources of demand, moat, risks).
Bull Case 예시:
전제: AI 투자 사이클이 2027년까지 지속
논리: 하이퍼스케일러 설비투자 6,000억 달러 → Blackwell 2.2배 성능으로
교체 수요 → CUDA 해자가 경쟁 방어
결과: 데이터센터 매출 연 25% CAGR → 2030년 $3,500억
Trendline Graph: references/trendline_guide.md Reference
10. Valuation
- Current metrics (P/E, PEG, EV/EBITDA) + Contextual Explanation
- Historical Comparison: Is it expensive versus the past? Why?
- Investment Implications
11. Glossary — Required
Organize all technical terms in Korean alphabetical/English alphabetical order. references/glossary_template.md Reference
12. Disclaimer and Sources
Workflow
Phase 0: Research (Web Search) → Fact Base
Phase 1: Draft (내러티브 + 시각화)
Phase 2: Evaluate (Critical Issues)
Phase 3: Plan → Phase 4: Rewrite → Phase 2 (반복)
종료: Critical == 0 AND 필수 시각화 포함
Critical Issues
| ID | Violation |
|---|---|
| C1 | Hallucination — Figures not in the Fact Base |
| C2 | Missing "Why" — Business description without sources of demand |
| C3 | Undefined Terms — Use of technical terms without explanation |
| C4 | Numbers Without Context |
| C5 | Missing Visualizations (Sankey/trendline/growth rate table) |
| C6 | Insufficient Forecast Basis — Bull/Bear/Base disconnected from the narrative |
| C7 | Missing Glossary |
Reference Files
references/sankey_guide.md— Sankey Diagramreferences/trendline_guide.md— Bull/Bear/Base Trendlinereferences/glossary_template.md— Glossary Template