Valuation Workbench
A share price means nothing on its own. Eight methods for working out what a business is actually worth — a 20-year discounted cash-flow engine plus the mean-multiple methods you fall back to when cash flow is too cyclical to forecast. Live prices, editable assumptions, the whole AI-capex watchlist preloaded.
What the model is doing
Why a valuation is worth anything, and which method suits which business
What the model is doing
Why a valuation is worth anything, and which method suits which business
When you buy a share you buy a piece of a business, and a business is worth all the cash it will generate from here, discounted back to today. Divide by the share count and you have intrinsic value per share. Everything else is an argument about the forecast.
Which method for which company
| Business | Method | Why |
|---|---|---|
| Consistent compounder | Discounted free cash flow | Narrow range — the number means something. AVGO, TSM. |
| Compounder in a capex spike | DCF on normalised capex | Swap this year's capex for the multi-year average. MSFT, GOOGL, META. |
| Cyclical | Median P/E or median P/S | DCF produces nonsense on a cycle. MU, LRCX, KLAC. |
| Financial | Discounted net income | Cash-flow methods distort insurers and asset managers. |
| Bank | Median price-to-book | The balance sheet is the business. |
| Unprofitable / pre-scale | Do not value it | Negative cash flow, no track record — trade it, do not model it. |
The model
Every field is editable — presets are a starting point, not an answer
The model
Every field is editable — presets are a starting point, not an answer
Inputs
Valuation ladder — all eight methods against today's price
Cash-flow consistency — the test that decides which method you may use
Sensitivity — value per share
rows: discount rate · columns: years 1–5 growth| disc \ growth | 15% | 20% | 25% | 30% | 35% | 40% | 45% |
|---|---|---|---|---|---|---|---|
| 6% | $148.42 | $198.87 | $265.44 | $352.73 | $466.42 | $613.54 | $802.74 |
| 7% | $134.32 | $179.29 | $238.52 | $316.04 | $416.86 | $547.15 | $714.49 |
| 8% | $121.99 | $162.20 | $215.06 | $284.11 | $373.77 | $489.47 | $637.90 |
| 9% | $111.18 | $147.24 | $194.55 | $256.23 | $336.19 | $439.23 | $571.24 |
| 10% | $101.66 | $134.11 | $176.57 | $231.82 | $303.33 | $395.34 | $513.06 |
| 11% | $93.27 | $122.54 | $160.75 | $210.39 | $274.51 | $356.89 | $462.15 |
| 12% | $85.84 | $112.32 | $146.81 | $191.52 | $249.16 | $323.11 | $417.46 |
Market snapshot — NVDA
What the market currently pays and expects, live from Yahoo Finance
Market snapshot — NVDA
What the market currently pays and expects, live from Yahoo Finance
What the market currently pays and expects, live from Yahoo Finance. The valuation engine above does not read any of it — these are here to check your own assumptions against the consensus, and to show where the two disagree.
Loading live market data…
Glossary
Every term on this page, in the sense it is used here
Glossary
Every term on this page, in the sense it is used here
Every term on this page, in the sense it is used here. For the semiconductor vocabulary — HBM, CoWoS, WFE, EUV — see the site glossary.
The cash-flow engine
- TTM (trailing twelve months)
- The last four reported quarters added together, rather than a fiscal year. It keeps a company that reported two months ago comparable with one that reported last week.
- Operating cash flow (OCF)
- Cash actually generated by running the business, before anything is spent on new plant or equipment. Harder to flatter than net income, because it starts from cash received rather than revenue booked.
- Capex (capital expenditure)
- Cash spent on long-lived assets — fabs, tools, servers, data centres. A capex spike depresses free cash flow for years before the assets earn anything back, which is why this page can swap the latest figure for a five-year average.
- Free cash flow (FCF)
- Operating cash flow minus capex: what is left for shareholders after the business has paid to keep itself competitive. The default stream discounted here.
- DCF (discounted cash flow)
- Projecting a cash stream forward, then shrinking each future year back to what it is worth today. This engine runs 20 years in three growth stages and stops there.
- Discount rate
- The annual return you require to tie up money in this business rather than somewhere safer. Higher rate, lower value. It carries more weight than any other single input — the sensitivity grid exists to show you how much.
- Terminal value
- The lump sum normally bolted on to represent everything past the forecast horizon. This model deliberately has none, so any cash flow after year 20 is upside that the number does not claim.
- Growth stages
- Years 1 to 5, 6 to 10, and 11 to 20, each with its own growth rate. Splitting them stops a high near-term rate from compounding into an absurd figure two decades out.
The multiple methods
- EPS (earnings per share)
- Net profit divided by shares outstanding. Trailing and diluted here, so it counts profit already reported and shares that options and converts would create.
- P/E (price to earnings)
- Price divided by EPS — what you pay for a dollar of annual profit. Comparing today against a five-year average asks whether the market has re-rated the business or the business has actually grown.
- P/S (price to sales)
- Price divided by revenue per share. Useful where profit is distorted by write-downs or amortisation, because revenue is much harder to bend.
- P/B (price to book)
- Price divided by book value per share. Meaningful for banks, whose balance sheet is the business, and close to meaningless for a company that has bought back most of its equity.
- Book value per share
- Assets minus liabilities, per share. Relentless buybacks can drive it near zero without the business weakening at all, which is why several tickers here carry a warning against reading their P/B.
- PEG (price/earnings to growth)
- P/E divided by the growth rate, as a rough test of whether a high multiple is earned. Crude, and it collapses when growth is exceptional, so this page caps the growth input it will use.
- P/S-to-growth
- The same idea applied to sales, for businesses growing fast enough that earnings are not yet a fair measure of them.
- Mean versus median multiple
- A mean is dragged around by one freak year — a loss-making quarter can print a P/E of 130 and wreck the average. Where that happens the median is substituted, and the note on each ticker says so.
Reading the answer
- Intrinsic value
- What the business is worth on the assumptions you set, as opposed to what it currently trades at. The whole output of this page is one estimate of it.
- Margin of safety
- The discount to intrinsic value you insist on before buying, to absorb the fact that your forecast will be wrong. Sets the buy-below line.
- Bull / bear band
- The same model rerun with growth moved up and down by a few points. A wide band is the model admitting the answer is sensitive to a guess.
- Consistency test
- Whether cash flow rose in at least 70% of periods and never went negative. Fail it and a discounted valuation is arithmetic performed on noise, so the multiple methods take over.
- Compounder / cyclical
- A compounder grows cash flow through the cycle and can be forecast. A cyclical swings with capacity and pricing — its good year tells you nothing about its next one, so it is valued on where its multiple sits against its own history.
- Normalised capex
- Latest capex replaced by the five-year average, to stop a single heavy building year from making a healthy business look cash-poor. Flatters any company whose spending has risen permanently rather than temporarily.
Notes and limitations
What the model does not do, where the data comes from
Notes and limitations
What the model does not do, where the data comes from
What this model does. The three discounted methods project the chosen stream for 20 years across the growth stages you set, discount each year at your rate, and stop — no terminal value. The multiple methods apply a five-year average multiple to the current per-share figure.
What it does not do. No net cash or debt adjustment, no share-count drift, no buyback effect, no currency hedging on the ADRs. The multiple methods use the median of each reported year rather than the mean, because a single loss-making year prints a P/E in the hundreds and drags an average into nonsense; where the two diverge the per-ticker note says so, and where no year gives a usable figure the row reads "—" instead of guessing. The consensus figure is growth for the current fiscal year, not a five-year rate: for a cyclical coming off a trough it prints something like Micron's 785%, which measures a recovery rather than a rate anything compounds at. The quick-set buttons apply at most 40% for that reason — type a higher number if you actually want to assume it.
Method and independence. Discounted cash flow, mean-multiple comparison and growth-adjusted multiples are standard valuation techniques, long established and in the public domain. The engine, wording, presets and code on this page are original work. No third-party text, slides, images, datasets or branding are reproduced here, and this tool is not affiliated with, sponsored by or endorsed by any commentator, educator or publisher.
Data. Price and trailing financials refresh live from Yahoo Finance on each ticker change; multi-year history and five-year mean multiples are a snapshot taken 24 August 2026. Verify against filings before acting on any of it — this is a research tool, not investment advice.