Sample note. Numbers are illustrative placeholders, not live data.
The setup
NVIDIA sells the only product in its category that ships at volume today. That gives it something rare in hardware: the ability to set price rather than accept it. The question I keep coming back to is not whether demand persists, but how long the margin structure survives while every large customer is funding an alternative.
What the price implies
| Metric | Value | Note |
|---|---|---|
| Forward P/E | 31.4x | Consensus, next twelve months |
| Gross margin | 73.1% | Trailing four quarters |
| Data center mix | 87% | Share of revenue |
| Net cash | $38.2B | Illustrative |
At roughly 31 times forward earnings, the market is paying for continued growth, but not for a permanent margin structure. My read is that the multiple already assumes some compression.
Three things that would break the thesis
- A credible second source. Custom silicon from hyperscalers reaching parity on total cost of ownership, not just on paper.
- A software moat that stops mattering. CUDA is the real lock-in. Watch adoption of open compiler stacks.
- Capex digestion. If the largest customers pause for two quarters, the operating leverage runs in reverse.
What would change my mind
A sustained decline in data center gross margin below 65% with no offsetting volume would tell me the pricing power story is over. Until then I treat every pullback as a chance to re-underwrite, not to panic.
Educational content only. Not investment advice.