A token has no cash flow unless somebody designed one. Four approaches are used and they differ enormously in how much of an assertion they make, so grade them on one standard: what would have to be observed for this approach to be shown wrong?
How does the cost-of-production model work?
The first approach computes the cost of production and treats it as a floor. At an illustrative 720 exahashes a second of network hashrate, a fleet averaging 22 joules per terahash and 5.5 cents a kilowatt hour, the network draws 15.84 gigawatts and 380,160,000 kilowatt hours a day, costing $20,908,800 against a daily issuance of 450 coins. That is $46,464 of electricity per coin. Every one of those inputs is perishable and two of them are unobservable in real time; widen the fleet efficiency to a 16 to 30 range and the power price to 3.5 to 7.5 cents and the answer runs from $21,504 to $86,400, a factor of 4.02. As a valuation this is close to useless.
As a statement about miner behaviour under stress it is genuinely informative, because it tells you where marginal supply switches off. Falsifiable in principle, and honest about being a cost rather than a value.
Advertisement
Does Metcalfe's law value Bitcoin?
The second approach reaches for a scaling law, and Metcalfe's is the usual one: a network's value scales with the square of its users. Fit the constant to a point where 42 million active addresses supported a $380 billion network value, apply it to 61 million addresses now, and the prediction is $801.6 billion against an observed $1,180 billion, an error of -32.07%. The exponent that fits both points exactly is 3.04, not 2.
That is the tell: an approach with a free constant refitted after each miss, and an exponent adjusted to taste, cannot be wrong and therefore says nothing. The network-value-to-transactions ratio is more useful because it is a comparable rather than a model; at $1,180 billion against $9.4 billion of daily on-chain transfer value it reads 125.53 days, which is a number you can put in a time series and argue about.
How does the addressable-market approach work?
The third approach names an addressable market and discounts an expected value out of it. State a market, a capture share, a horizon and a probability. If the addressable market is the $18 trillion of above-ground gold, the capture share is 20%, supply in ten years is 20.6 million coins and the discount rate is 12%, then the success state is worth $174,757 a coin, a 25% probability gives an expected terminal value of $43,689, and the present value is $14,067.
Now invert it against the observed price of $59,296, which is where the approach earns its keep: holding the capture share at 20% the market is pricing a probability of 105.4%, which is impossible, so at certainty the price requires a capture share of 21.08% of the entire gold market. Now the assumption is on the table and can be argued with, which is the whole point of the exercise.
The commonest error here is discounting at a venture rate and applying a probability. Both adjustments price the same risk. Doing both cuts the present value from $14,067 to $4,691, an understatement of 66.65%, and the mistake is invisible unless somebody writes the two adjustments down side by side.
How do token cash flows value a network?
The fourth approach discounts token cash flows, and it only exists where a protocol distributes fees to stakers. Take a 3.20% gross staking yield with 0.55% issuance and 0.40% fee burn: net supply growth is 0.15% and the yield net of dilution is 3.05%. At a 12% required return and 4% growth in fee demand the Gordon multiple is 12.50, so the justified value is 0.3813 of the current price, or equivalently the current price requires 8.95% growth rather than 4%, or a net yield of 8.00% rather than 3.05%. The current net yield falls short by a factor of 2.62.
This is the only one of the four that produces a claim in the same grammar as an equity valuation, and it is the only one that a future observation can straightforwardly refute.
How does stock-to-flow compare, and how do the four rank?
Grade all four against stock-to-flow, which divides the existing stock by annual new supply: 19.9 million coins over 164,250 issued a year gives 121.16, against gold's 59.35 at 216,265 tonnes over 3,644 tonnes of mine supply. As a scarcity comparison that is fine and mildly interesting.
As a price model it fails the falsifiability test twice over, because the regressor is a deterministic function of the issuance schedule known years in advance and because the fit is re-estimated after every deviation. A model whose independent variable is a calendar cannot be tested by the calendar. Rank the four approaches by what would refute them: token cash flows first, addressable market second because its assumptions are at least explicit, cost of production third as a supply statement rather than a value, and network scaling last. All four inputs above are perishable; treat every figure in this section as illustrative at September 2026 and verify against the network's own data before quoting it.