P2A-C1 · Valuation Basics (DCF / Multiples / Margin of Safety)¶
Core One-Liner
Price is what you pay, value is what you get. Always demand a margin of safety. — Buffett
Universal Investment Model — Works for Any Industry
P2A-C1 (Part 2.A Chapter 1 of 5). After this chapter, you can calculate a fair value range for any company, not just look at PE.
1. The Problem: You Only Look at PE — A Big Trap¶
Retail investors look at stocks: - "NVDA PE 30x, not cheap" - "AMD PE 50x, too expensive" - "OAI primary valuation $300B, how do you value that?"
Why only looking at PE is wrong: - PE only looks at 1 year of earnings, not the future - PE assumes stable earnings (a cyclical stock at PE 8x is actually most expensive) - PE ignores cost of capital (AI companies with high capex have artificially flattering PE)
→ Valuation is not a single number, it's a range. Use ≥ 3 valuation methods for cross-validation.
2. The Solution: 3 Valuation Methods + Margin of Safety¶
| Method | What It Looks At | Suitable For | Limitation |
|---|---|---|---|
| DCF (Discounted Cash Flow) | Future 10-year FCF + Terminal Value | Stable cash flow companies (KO / JNJ) | AI companies are unpredictable even 3 years out, not suitable |
| Multiples (PE / EV/Rev / P/B) | Compare to history / peers / itself | Most companies | Doesn't give absolute fair value |
| Reverse DCF | What growth assumption is implied by current price | High-growth companies | Requires reverse thinking |
Cross-validate with 3 methods → Get a fair value range, not a single point.
Margin of Safety (created by Graham in 1934): Within the fair value range, you want a 20-30% buffer before buying (e.g., fair value $100, you buy at $70-80).
3. How It Works: Detailed Explanation of 3 Methods¶
3.1 DCF (Discounted Cash Flow)¶
Core Formula (simplified):
Value = Σ (Annual FCF / (1 + WACC)^Year) + Terminal Value
WACC = Weighted Average Cost of Capital (typically 8-12%)
Terminal Value = Final Year FCF × (1 + g) / (WACC - g), g = long-term growth rate (typically 2-3%)
Example (simplified): Assume a company has Year 1 FCF $10B, 5-year growth 15%, then stable 3%, WACC 10%:
| Year | FCF | Present Value |
|---|---|---|
| 1 | $10B | $9.1B |
| 2 | $11.5B | $9.5B |
| 3 | $13.2B | $9.9B |
| 4 | $15.2B | $10.4B |
| 5 | $17.5B | $10.9B |
| Terminal Value | $17.5 × 1.03 / (0.10 - 0.03) = $257B | $159B |
| Total Value | $208B |
If current market cap is $150B → DCF is 38% above price, ample margin of safety, buy candidate. If current is $300B → DCF is 30% below price, don't buy.
Limitations of DCF for AI Companies: - Future 10-year FCF is unpredictable (e.g., NVDA's GPU demand 5 years from now is unknown) - Wrong growth assumption → DCF is 100% wrong
Using DCF for mature companies like KO / JNJ is fine, but not reliable for AI companies.
3.2 Multiples¶
Common Multiples:
| Multiple | Formula | How to Read |
|---|---|---|
| P/E (TTM) | Stock Price / Last 12 Months EPS | Compare to history / peers |
| fwd P/E | Stock Price / Expected Next 12 Months EPS | Key for growth stocks |
| PEG | P/E / EPS Growth Rate (%) | < 1 cheap, > 2 expensive |
| P/S | Stock Price / Revenue Per Share | For unprofitable companies |
| EV/Revenue | (Market Cap + Net Debt) / Revenue | SaaS / Neocloud |
| EV/EBITDA | (Market Cap + Net Debt) / EBITDA | Cross-industry comparable |
| P/B | Stock Price / Book Value Per Share | Financials / Cyclicals |
How to Use:
| Step | What to Do |
|---|---|
| 1 | Find your company's fwd PE |
| 2 | Find fwd PE for 3-5 true peers (not sector ETF) |
| 3 | Find your company's 5-year historical fwd PE high / median / low |
| 4 | See where your company currently sits in the 5-year range percentile |
| 5 | Synthesize — high percentile + high peers = expensive, low percentile + low peers = cheap |
Example: NVDA fwd PE 30-35x - 5-year range: 15-50x, currently at 70th percentile - Peers (AMD / AVGO) 25-30x — NVDA is 20-30% more expensive than peers - Judgment: Slightly expensive but growth (50%+) supports it. Buy on dips, don't chase.
3.3 Reverse DCF¶
Flip DCF: Don't calculate fair value, calculate "what growth is implied by the current price".
Example: NVDA currently $135, fwd PE 30x. What 5-year EPS growth is implied?
Simplified formula:
P/E_now = P/E_terminal × (1 + g)^5 / (1 + discount_rate)^5
Assume P/E_terminal = 15x (mature phase), discount = 10%
30 / 15 = (1 + g)^5 / 1.61
2 × 1.61 = (1 + g)^5
3.22 = (1 + g)^5
g = (3.22)^(1/5) - 1 ≈ 26% annual EPS growth (5 years)
→ Current $135 implies NVDA needs 26% annual EPS growth for 5 years. Do you think that's reasonable?
- NVDA's past 3 years EPS YoY +100%+ → 26% is a conservative assumption
- Reverse DCF shows the price isn't unreasonable, even with a 50% discount it's still OK
Reverse DCF is most useful for AI stocks: Because forward DCF has too many assumptions, reverse DCF simply asks "what is the market assuming", then you judge if that assumption is reasonable.
4. vs Retail Investor Approach¶
| Dimension | Retail Investor | What You Can Change |
|---|---|---|
| Looks at 1 PE number | Look at combined PE + PEG + EV/Rev | ✓ 5 minutes for 3 multiples |
| Doesn't know fair value range | Calculate DCF + Reverse DCF + Peers, range $X-Y | ✓ 1 hour to complete |
| Doesn't demand margin of safety | Wait for price 20-30% below range before buying | ✓ Mental discipline |
5. Try It: Do a 3-Method Valuation for Your Ticker¶
Task (~1 hr):
| Method | Your Stock Answer |
|---|---|
| fwd PE | ___ x, 5-year range ___ - ___ x, currently at ___ percentile |
| Peer fwd PE (3-5 companies) | , , ___ |
| Reverse DCF | Current price implies 5-year EPS growth ___ % |
| Combined fair value range | $ - $ |
| Margin of safety entry | $___ (low end of range -20%) |
Self-check (3 items checked → move to P2A-C2):
- You can calculate reverse DCF for a company (using PE / growth assumptions)
- You can list 3-5 true peers (not sector ETF)
- You can give a fair value range instead of a single point
6. What's Next¶
Valuation is one mental model. Buffett / Munger / Graham have more — Mental Models are a latticework.
→ P2A-C2 · Mental Models — The core 5 models from the Big Three: Buffett / Munger / Graham.
7. Deep Dive (optional): Damodaran Model Public + AI Company DCF Challenges¶
Click to expand deep content
Aswath Damodaran (NYU): - Free model excel: pages.stern.nyu.edu/~adamodar/ - He does quarterly DCFs on popular AI stocks (NVDA / TSLA / META) - Recommended: 2024 NVDA valuation post (free on Substack)
AI Company DCF Challenges: 1. Unpredictable Capex pace — NVDA capex grows $3-5B annually, what will it be in 5 years? 2. Customer concentration — NVDA's 4 hyperscalers account for 50%, one cut has massive impact 3. Margin sustainability — 75% gross margin isn't sustainable at 90% 4. Scaling laws assumptions — If a wall is hit, growth assumptions are upended
→ For AI companies, Reverse DCF is more useful than forward DCF — ask "what is the market assuming", not "what is the company worth".