NVIDIA Corporation
Fundamental — what it's worthbottom-up fair value
Built up from an explicit five-year discounted-cash-flow forecast (the primary lens), relative multiples and a through-cycle forward-earnings read — what is it worth?
Is it cheap or expensive right now?
Latest price vs our fair value. A comparison, not a recommendation.
What it's worth — three lenses
Three independent valuation methods, blended into one weighted central fair value.
| Valuation lens | Per share ($) | Weight |
|---|---|---|
| DCF (5-yr FCFF) — primary | 189 | 45% |
| Relative multiples (fwd EV/EBITDA, P/E) | 200 | 25% |
| Forward earnings power | 230 | 30% |
| Weighted central fair value | 204 | +3.9% vs spot |
The three lenses span $189 (DCF, the primary and most conservative read) to $230 (forward-earnings power, the ceiling) and straddle spot — cash-generative but priced for a long runway. In the DCF the terminal value is ~79% of enterprise value, disclosed and sensitized rather than buried; the swing factor is how many years AI data-center demand sustains super-normal growth. Full detail and all three lenses are in the study and the open model.
Technical & price structuretrend, momentum, key levels
Trend, momentum and key levels — what is the price doing now?
What the chart says
In plain terms: NVIDIA trades a few percent above our weighted central fair value (about $204) — roughly fairly valued, and priced on the durability of the AI data-center build-out rather than on today's earnings alone. It runs a net-cash balance sheet and earns a return on capital employed near 62% against a ~10.5% cost of capital, so the debate is not whether the returns are real but how many years the super-normal phase lasts. The bull case (~$287) is AI-infrastructure spend compounding toward ~$1tn a year while the runway extends and the multiple holds; the cautious case (~$147) is a broad AI-capex digestion or bubble scare with China Data Center still foreclosed by export controls. What moves it is the AI data-center capex path, the China export-control question, and customer concentration among a handful of hyperscalers.
Key levels
Volatility & where the case breaks
| Typical daily moveRoughly how far the price swings in an average day (based on a measure called ATR). Bigger means choppier. | ~2.2% |
Monte Carlo — where could the price go?near-term price paths
50,000 simulated futures — near-term price paths, independent of the fundamental value.
The 5th–95th and 25th–75th percentile bands from the 50,000-path simulation. Static and pre-computed — independent of the fundamental levers above.
The exact percentiles
The 5th / 25th / middle / 75th / 95th outcomesLine up all 50,000 outcomes low to high — these are the values at the 5%, 25%, 50%, 75% and 95% marks., from 50,000 paths.
| Timeframe | Rare low (5%) | 25% | Middle | 75% | Rare high (95%) |
|---|
What drives the odds
50,000 paths, 16 forces: the AI data-center capex cycle, hyperscaler capex guidance, semis/AI equity beta and Nasdaq flows, the gross-margin path as Blackwell gives way to Rubin, US 10-year real yields, USD and global liquidity, and a supply-chain regime (TSMC CoWoS / HBM) — plus event forces: a quarterly print versus the ~$91bn guide, a hyperscaler capex-guidance shift, a China export-control change, a Rubin ramp / supply-news event, an AI-capex-fatigue or bubble scare, a custom-silicon competitive win or loss, a major partnership or mega-order, a capital-return / buyback surprise, and an index / options-flow rebalancing. Details in the methodology.
Peer set & risks
Where NVIDIA sits in its markets
NVIDIA designs the accelerated-computing platform — GPUs, networking and the CUDA software stack — that trains and runs modern AI, and holds roughly four-fifths of the AI-accelerator market. Peers trade in different currencies and aren’t directly value-comparable, so this is a competitive map, not a price table.
| Arena | NVIDIA's position | Main rivals |
|---|---|---|
| AI training / inference accelerators | #1 (~81% share) | AMD (MI450) · custom ASICs |
| Data-center GPU compute | Dominant | AMD · Intel Gaudi |
| Custom AI silicon (hyperscaler) | Merchant leader | Google TPU · Amazon Trainium · Meta MTIA |
| AI software ecosystem | CUDA moat | AMD ROCm · open frameworks |
| Data-center networking | Strong (NVLink / InfiniBand) | Broadcom · Arista · Cisco |
| Discrete / gaming GPU | #1 | AMD Radeon · Intel Arc |
The investment debate turns less on any single arena than on how long AI data-center capex compounds and how much of that spend NVIDIA keeps as the merchant hyperscalers design their own silicon and AMD scales the alternative. Competitive map, not advice.
About this series & how we build these
Want the full study and the spreadsheet?
Full write-up plus the editable Excel model.
Found a flaw? Attack the model
The full study (PDF) and Excel model are available on a computer.
Edition: 6 Jul 2026. Older editions stay in the Library.