1. Executive Summary
1.1 The Intelligence Refinery Thesis
1.2 Key Metrics at a Glance
1.3 Core Findings & Investment Implications
1.4 Intelligence Margin vs SaaS Margin Framework
2. The AI Factory = New Oil Refinery
2.1 Why the Analogy is Structurally Accurate
2.2 AI Factory vs Oil Refinery: Economic Mapping
2.3 The “Crack Spread” → Gross Margin per Token
2.4 Throughput, Utilization & Revenue Ceiling
2.5 Investor Interpretation Framework
3. Cost per Token — Cross-Platform Model
3.1 Model Overview & Assumptions
3.2 Platform Benchmark Comparison (CPU, GPU, TPU, ASIC, LPU)
3.3 Cost per 1K Tokens: Full Stack Comparison
3.4 Hardware Stack Breakdown (Compute, Memory, Network)
3.5 Architecture-Level Efficiency Differences
4. Cost Waterfall — From Power to Profitability
4.1 Full Cost Stack Decomposition
4.2 Silicon vs Power vs Memory Cost Share
4.3 Power Cost: Volatility vs Structural Impact
4.4 Cost Evolution Across Architectures
4.5 TOPS/Watt vs FLOPS — The Real Efficiency Metric
4.6 Rising Software Cost Layer (CUDA, Middleware, Orchestration)
5. The Intelligence Frontier — Latency, Throughput & TCO
5.1 Latency vs Throughput Tradeoff Curve
5.2 Platform Positioning Across Workloads
5.3 Use Case → Infrastructure Matching Matrix
5.4 Real-Time vs Batch Economics
5.5 Token Value Differentiation by Use Case
6. TCO Sensitivity & Break-Even Analysis
6.1 Utilization vs Cost Heatmap
6.2 Power Price Sensitivity Modeling
6.3 Break-Even Utilization Thresholds
6.4 Cost Reduction per Utilization Increment
6.5 Operational Risk: Underutilized Infrastructure
7. Gross Margin per Intelligence Unit
7.1 Intelligence Margin Framework
7.2 Platform × Use Case Margin Matrix
7.3 Value Chain Margin Stack (Silicon → Application)
7.4 Margin Capture by Ecosystem Layer
7.5 Commoditization Risk & Bear Case Analysis
8. The Intelligence Economy — TAM (2023–2028E)
8.1 Market Sizing Methodology
8.2 Token Demand Growth Model
8.3 Use Case Adoption Curves
8.4 Revenue Forecast by Segment
8.5 $480B TAM Breakdown
8.6 Margin Allocation Across Value Chain
9. Portfolio Applications
9.1 Private Equity
- AI Diligence Framework
- Cost per Token Benchmarking
- Utilization Risk Assessment
9.2 Hedge Funds
- Long/Short Framework
- Pair Trades (Infrastructure vs Application Layer)
- Margin Expansion Plays
9.3 Enterprise CFO Strategy
- Build vs Buy vs API Decision Tree
- Cost Optimization Pathways
- Scaling Strategy by Token Volume
10. Appendix
10.1 Cost Model Architecture
10.2 Assumptions & Sensitivity Analysis
10.3 Platform Specifications Comparison
10.4 Data Sources & Methodology Notes
11. Disclaimers & Legal
11.1 Institutional Use Disclaimer
11.2 Forward-Looking Statements
11.3 Confidentiality Notice
