NVIDIA CORP

NVIDIA CORP (NVDA) — Investment Memo

🐂 The Bull Case (Warren's voice)

• Moat is widening, not eroding. ROE from 28.9% to 76.3% while revenue grew 31×. That’s not luck — that’s CUDA lock-in turning hyperscalers into captive customers. Switching costs? Retraining a trillion-parameter model on competitor silicon takes years and billions. They won’t.

• Economics are exceptional — and improving. Net margin doubled from 24% to 55.6% while revenue doubled to $215.9B. That’s pricing power plus operating leverage. Free cash flow isn’t disclosed, but net income is $120.1B and capex is modest (no debt needed). The cash register is screaming.

• Attractive entry price? Only if the market panics. At the DCF estimate of $104.37 per share (implied ~$2.5T market cap), this is a 70% discount to today’s price. That would require a 60–70% drawdown — possible only if the AI narrative cracks. If you believe the moat holds, $140–$160 (25% margin of safety above DCF) would be a generational buy. Buffett’s ideal $78 (50% margin) is fantasy unless the world ends.

🐻 The Bear Case (Charlie inverts)

• Scenario #1: The CUDA compatibility crack. A hyperscaler (Amazon, Google, Microsoft) builds a chip that runs CUDA software natively — not emulated, not slower. If they undercut NVIDIA’s 55.6% margin by 30%, the switching cost evaporates overnight. This is the most likely structural threat. Timeframe: 3–5 years.

• Scenario #2: The customer concentration implosion. Three customers each >10% of receivables. If one (say, Microsoft) shifts to in-house silicon or simply cuts orders due to macro or overcapacity, revenue drops 20–30%. The fixed cost base doesn’t shrink. Margins collapse from 55.6% to 20%. The stock falls 60% and never recovers to today’s multiples.

• Scenario #3: The 2023 margin collapse repeats — permanently. Net margin cratered from 36.2% to 16.2% in one year, then bounced. The 10-K gives no explanation. If that was inventory writedowns or channel stuffing, the next downturn could be worse — and the bounce may not come. A 16% margin business at $200B revenue is worth $600B, not $3T.

Most likely threat: Scenario #1. The hyperscalers are already building chips. The question is when CUDA compatibility becomes a commodity.

💰 Valuation & Margin of Safety

• Intrinsic value estimate: $104.37 per share (DCF: 8% FCF growth, 10% discount, 3% terminal). This assumes the current revenue trajectory slows sharply — which it must.
• 25% margin of safety entry: $78.28 per share (conservative, assuming CAGR drops to 15% and margins normalize to 35%).
• 50% margin of safety entry: $52.19 per share (Buffett’s ideal — assumes the moat cracks and the business resets to a high-quality semiconductor company, not a monopoly).
• Current price: ~$880. Overvalued by 8.4× vs DCF. Even applying a generous 2× multiple to the DCF (say $209), the stock is 4× too expensive.

Verdict: PASS

At $880, NVIDIA is priced for perfection — zero margin for error on CUDA lock-in, customer concentration, or margin sustainability. The DCF says $104, and even a 25% margin of safety requires $78. This is a wonderful business at a terrible price. Wait for the moat to prove itself through a downturn, then buy the panic.