Framework Convergence
Graham
SPEC.
P/E 198x avg (max 20x)
-1,488% MoS
Buffett
HOLD
Requires 20%+ CAGR 10yr
Wait for $130
Sowell
BULL
Natural monopoly, 87.4% share
Macro favorable
Reasoning Chain
How Each Framework Contributed
📐 Graham: 10-year average EPS ($1.02) × 20x max = $20.44 ceiling. Even TTM ($6.53) × 20 = $130.60. Current price 90% speculative component. Classification: SPECULATION by price.
🎯 Buffett: Projects EPS at 25% CAGR → $60.79 in 10yr. At 20x = $1,215. PASSES at 25% growth. At 15% CAGR → $475 = FAILS. The bet is on growth DURATION, not quality.
🌐 Sowell: Controls the scarcest resource in tech (AI compute). Natural monopoly with $18.5B R&D moat. All heavy-weight factors bullish. No systemic risk detected.
🔍 Synthesis: All three AGREE the business is extraordinary. Disagreement is purely about PRICE. Resolution: HOLD if owned, WAIT if not. Conditional BUY at ~$130 (20x TTM).
Action Levels
$202.81
Current (Spec. Premium)
≤$20.44
Graham Max Invest.
Actionable Signals
STRONG BUY
≤$82
12.5x TTM EPS
Aggressive accumulation
BUY
$82–$131
12.5–20x TTM
Measured buying
HOLD / WAIT
$131–$250
Current zone
Monitor EPS growth
TRIM
>$250
AND growth <30%
Reduce position
SELL
Decline
Earnings fall + price >$200
Exit position
Price Thermometer
Multi-Framework Price Zones
BARGAIN ≤$20GRAHAM ≤$82BUFFETT ≤$131SPEC. $200+
Bull Case
- AI capex cycle continues 5+ more years
- EPS CAGR sustains 25%+ (Scenario A)
- Blackwell/Rubin capture inference market
- Sovereign AI demand adds $10B+ TAM
- Software/CUDA licensing becomes recurring
Bear Case
- Earnings decelerate below 25% YoY
- Hyperscaler capex cuts (MSFT, GOOG, AMZN)
- Custom ASIC captures >30% inference
- AMD market share exceeds 15%
- TSMC disruption (geopolitical)
- Gross margin compression below 65%
Intrinsic Value Calculation
Buffett Valuation Method
STRONG BUY ≤$82BUY ≤$131HOLD ≤$250OVERVALUED $250+
Step 1: Current EPS (TTM) = $6.53
Step 2: Scenario A — 25% CAGR for 10yr: $6.53 × (1.25)^10 = $60.79
Step 3: Future price = $60.79 × 20 (mature tech P/E) = $1,215.80
Step 4: Required Return (10%/yr): $202.81 × (1.10)^10 = $526.05
Step 5: $1,215 > $526 → PASSES at 25% growth
Step 6: Scenario B — 15% CAGR: $6.53 × 4.05 × 18 = $476 → FAILS (< $526)
Step 7: Scenario C — 10% CAGR: $6.53 × 2.59 × 15 = $254 → FAILS significantly
"It's far better to buy a wonderful company at a fair price than a fair company at a wonderful price." — Warren Buffett
Durable Competitive Advantage
Economic Moat Assessment
NVIDIA is the TEXTBOOK natural monopoly in AI compute. CUDA ecosystem (4.7M+ developers, 18 years) creates impenetrable switching costs. $18.5B R&D exceeds AMD's entire revenue. 92% of revenue from data center — the scarcest resource in tech.
Moat Sources: CUDA lock-in, network effects, economies of scale, switching costs, pricing power (71% gross margins, Blackwell ~2x Hopper pricing).
Owner Earnings
Net Income: $120,067M
+ D&A: $2,843M
- Maint. CapEx: $3,021M (est. 50% of total)
= Owner Earnings: ~$119,889M ($4.89/share)
FCF Yield: 1.94% (below 4.5% Treasury)
Extraordinary cash generation on capital-light model (only $6B capex). FCF yield below risk-free rate — acceptable ONLY if growth continues at 30%+ rates.
Management Quality
Jensen Huang — co-founder (1993), visionary who pivoted from gaming to AI years before the market. Revenue 31x in 10 years. Skin in the game: ~$170B personal stake. $1 retained → $32 market value created.
10-Year Earnings Record
| FY | EPS | Growth | ROE | Net Margin |
|---|
| FY2017 | $0.06 | — | ~22% | 24% |
| FY2018 | $0.12 | +100% | ~35% | 31% |
| FY2019 | $0.17 | +42% | ~32% | 35% |
| FY2020 | $0.11 | -35% | ~18% | 26% |
| FY2021 | $0.17 | +55% | ~21% | 26% |
| FY2022 | $0.39 | +129% | ~37% | 36% |
| FY2023 | $0.17 | -56% | ~17% | 16% |
| FY2024 | $1.19 | +600% | ~69% | 49% |
| FY2025 | $2.94 | +147% | ~92% | 56% |
| FY2026 | $4.90 | +67% | ~76% | 56% |
10-Year EPS CAGR: 54.6%. Exceptional but volatile — two major declines (-35% FY2020, -56% FY2023) followed by explosive recoveries. EPS range: $0.06 to $4.90 (81x).
Buffett Scorecard
Consumer Monopoly
PASS
CUDA + 87% share
Yes
Gross Margin
PASS
71.1%
>40%
Debt Payoff
PASS
0.09 yr
<4 yr
CAPEX/Earnings
PASS
5.0%
<50%
Retained Earnings
PASS
$1 → $32 MV
>$1
Pricing Power
PASS
71% GPM, can raise
Above inflation
Owner-Oriented Mgmt
PASS
Founder-CEO 30yr
Aligned
Consistent Earnings
CAUTION
2 declines >35%
Stable growth
FCF Yield vs Bonds
FAIL
1.94% < 4.5%
>Bond yield
10yr Return ≥15%
DEPENDS
PASSES at 25% CAGR only
>15%
Score: 9 PASS / 1 FAIL / 1 CAUTION — Outstanding business; price is the ONLY question mark.
Buffett Would Like
- Widest moat in technology (CUDA + 87% share)
- Extraordinary ROE (81.7%)
- Capital-light model (FCF margin 44.8%)
- Founder-CEO with 30+ year track record
- Fortress balance sheet (net cash $51.5B)
- Massive & growing TAM ($1T+ AI infrastructure)
- Pricing power confirmed by 71% gross margins
Buffett Would Worry
- Technology businesses change fast — "durability?"
- Cyclical downturns (-55% NI drop in FY2023)
- FCF yield (1.94%) below Treasury (4.5%)
- Requires 20%+ growth for 10 years
- Customer concentration (hyperscalers 70%+)
- Those same customers building alternatives (TPU, Trainium)
- China export restrictions reduce TAM
- Semiconductor industry not "inherently stable"
Intrinsic Value — Graham Method
Security Analysis Valuation
BARGAIN ≤$8.52INVEST. ≤$20.44TTM MAX ≤$130OVERVALUED $130+
Step 1: Average EPS (10yr, FY2017–FY2026) = $1.022
Step 2: Normal Value = $1.022 × 12.5 = $12.78
Step 3: Bargain Price = $12.78 × 0.667 = $8.52
Step 4: Max Investment Price = $1.022 × 20 = $20.44
Step 5: Current Price = $202.81
Step 6: Margin of Safety = -1,487.6% (NEGATIVE)
Alt: Even TTM EPS ($6.53) × 20 = $130.60 — still 55% below current price
"An investment operation is one which, upon thorough analysis, promises safety of principal and a satisfactory return. Operations not meeting these requirements are speculative." — Graham & Dodd
Earning Power (10-Year Record)
| FY | Revenue ($M) | Net Income ($M) | EPS | NI Change |
|---|
| 2017 | 6,910 | 1,666 | $0.06 | — |
| 2018 | 9,714 | 3,047 | $0.12 | +83% |
| 2019 | 11,716 | 4,141 | $0.17 | +36% |
| 2020 | 10,918 | 2,796 | $0.11 | -32% |
| 2021 | 16,675 | 4,332 | $0.17 | +55% |
| 2022 | 26,914 | 9,752 | $0.39 | +125% |
| 2023 | 26,974 | 4,368 | $0.17 | -55% |
| 2024 | 60,922 | 29,760 | $1.19 | +581% |
| 2025 | 130,497 | 72,880 | $2.94 | +145% |
| 2026 | 215,938 | 120,067 | $4.90 | +65% |
EPS range: $0.06 to $4.90 (81x). Net income swung -55% then +581% in consecutive years. Stability ratio 5.9% (catastrophically below 50% threshold). This is the epitome of erratic.
Income Statement
Margins exceptional and improving. Gross margin recovered from 56.9% (FY2023 crypto bust) to 71-75%. Operating income primarily from core GPU/accelerator sales (92% data center). No accounting manipulation detected.
Balance Sheet
At $202.81, 97% of stock price is capitalized earning power with zero asset backing. However, financial position is a FORTRESS: net cash $51.5B, could repay ALL debt in 33 days of earnings.
Graham Scorecard
P/E vs 10yr Avg
FAIL
198.4x
≤20x
Earnings Stability
FAIL
5.9%
≥50%
Earnings Trend
CAUTION
Explosive/Erratic
Stable upward
Interest Coverage
PASS
503x
≥3x
Current Ratio
PASS
3.91x
≥2:1
LT Debt Payoff
PASS
0.09 yr
≤4 yr
Book Value vs Price
FAIL
31.6x P/B
≤1x
NCAV vs Price
FAIL
$3.14 vs $202
>Price
Dividend Record
PASS
12 years
≥10 yr
Revenue Stability
PASS
7% max decline
<33%
Margin of Safety
FAIL
-1,488%
≥33%
Financial Structure
PASS
Conservative
Conservative
Score: 6 PASS / 6 FAIL / 1 CAUTION — Business passes qualitative tests; PRICE fails quantitative tests catastrophically.
Economic Assessment
Sowell Verdict: ALCISTA (Bullish) — 80/100
Critical Tests (Heavy-Weight)
Scarce Resources
ALCISTA
Controls AI compute — scarcest in tech
15%
Price Signals
ALCISTA
71% GPM, rising prices, pass-through works
15%
Systemic Risk
NO RISK
Benign macro, AI has economic logic
15%
Competition
NATURAL
87.4% earned through innovation, not gov
15%
Supplier Mapping
NEUTRAL
TSMC sole manufacturer — risk managed
10%
Incentives
ALCISTA
Huang 3.5% ($170B), SBC $6.4B
10%
Government
NEUTRAL
China ban negative, CHIPS Act positive
10%
Scale & Efficiency
ALCISTA
$6.7M rev/employee, fabless
5%
Profit Diagnostics
ALCISTA
$120B NI signals genuine value creation
5%
Scarce Resources
| Resource | Control | Risk |
|---|
| GPU compute for AI training | Near-monopoly (87.4%) | LOW |
| CUDA ecosystem (18yr, 4.7M devs) | Exclusive IP | LOW |
| TSMC leading-edge capacity | Largest customer | MED |
| AI chip design talent (~32K eng.) | Concentrated | MED |
Market has correctly allocated resources to highest-value use (AI infrastructure) — willingness-to-pay $30K-$40K+ per GPU confirms genuine scarcity being communicated through prices.
Natural Monopoly Analysis
Why NVIDIA's Dominance is "Good" by Sowell's Standards
$18.5B
R&D (> AMD total rev)
NATURAL monopoly by Sowell's definition: Economies of scale in R&D create self-reinforcing advantage. Network effects in CUDA make it the Windows of AI. NOT artificial — no government granted this position. Earned through 30 years of engineering excellence. Entry is free; competitors simply can't match cumulative advantage.
High profits ($120B NI) attract competition (AMD, Intel, Google, Amazon) — exactly as Sowell predicts. Yet share remains 87.4%, suggesting moat too deep for rapid entry.
Evidence Integrity
All material conclusions rest on CONFIRMED data from primary sources (SEC filings, company IR, BLS, FRED). INFERRED claims are hedged and do not affect overall verdict direction.
Sowell Conclusion
NVIDIA exemplifies what Sowell describes as the price system functioning at its best. Resources flow to NVIDIA because it uses them most productively — generating $3.75M profit per employee. No central planner directed this allocation; billions of individual compute decisions, voted with real capital, determined that NVIDIA's products justify premium pricing.
Caveat on PRICE: The STOCK PRICE ($202.81, 31x TTM) reflects expectations of perpetual dominance. The BUSINESS is extraordinary; whether the STOCK at this price offers sufficient return is a separate question — that's for Graham and Buffett to answer.