Bitcoin Valuation Models Explained: MVRV, Realized Price, NVT, Production Cost and What Each Says in July 2026

Every week someone tells you bitcoin is cheap, and someone else tells you it is expensive, and both of them point at a chart. This guide is about what those charts actually measure, which ones have a defensible logic behind them, which ones are numerology, and how to combine them into a view you can defend. Updated July 2026, with each model scored against the current tape — bitcoin around $64,200, eight weeks of ETF outflows just ended, and miners underwater.

The honest framing first: there is no accepted valuation model for bitcoin. It produces no cash flows, so discounted cash flow is unavailable. Everything below is a heuristic — a way of asking 'expensive relative to what?' The useful ones anchor price to something observable and external. The useless ones fit a curve to the past and extrapolate.

1. Realized Price and Realized Cap — the market's true cost basis

Realized Cap values every coin not at today's price but at the price it last moved on-chain. Divide by supply and you get Realized Price: the average price the market actually paid for its bitcoin. It is the closest thing bitcoin has to a book value.

Why it works: it is computed from real transactions, not from a model. It tells you the aggregate cost basis of holders, which is the level at which the median holder is underwater — historically a zone of strong support and, in deep bears, a level price has traded below only briefly.

How to use it: price above Realized Price means the average holder is in profit; price below means the average holder is at a loss. Prolonged trading below Realized Price has marked cyclical bottoms. Track it on Glassnode or CoinMetrics — do not eyeball it.

Limits: it is backward-looking and moves slowly. It tells you where the market's cost basis is, not where price is going.

2. MVRV and MVRV Z-Score — the workhorse

MVRV is simply Market Cap divided by Realized Cap. Above 1, the market holds unrealized profit; below 1, unrealized loss. The Z-Score refines it: (Market Cap − Realized Cap) ÷ standard deviation of Market Cap, which normalises the ratio so that extremes across different cycles are comparable.

Why it works: it measures aggregate unrealized profit, which is a real behavioural driver. People sell when they are up a lot; people capitulate when they are down a lot. High MVRV Z-Score has historically coincided with cycle tops and low readings with bottoms.

How to use it: treat it as a thermometer, not a trigger. It tells you where you are in the cycle's profit distribution, not what happens next week. Its useful variants are cohort-specific: Short-Term Holder MVRV (price ÷ STH realized price) is one of the better short-horizon sentiment gauges, because coins held under 155 days are the ones that actually panic.

Limits: the ETF era distorts it. When coins sit in a custodian's cold storage on behalf of an ETF, on-chain movement no longer maps cleanly to economic transfer of ownership. Shares change hands in a brokerage account while the coins never move. That means Realized Cap increasingly understates true turnover. Every on-chain valuation metric inherits this problem, and anyone who tells you otherwise is not paying attention.

3. Production cost — the model that just failed

The claim is that mining cost puts a floor under price. As of July 2026, it does not: JPMorgan estimates bitcoin is trading roughly 19% below production cost, with all-in cost for listed miners near $78,000 against a spot price around $64,200, and it has stayed below for months.

Why it fails: the causal arrow points the wrong way. Production cost is derived from hashprice — miner revenue per petahash per day, currently at a record-low ~$29 — which is itself derived from price and difficulty. When price falls, cost per coin rises mechanically. Nothing about a miner's cost basis creates a buyer. Sunk costs do not turn off ASICs; miners shut down when price drops below the cash cost of the next kilowatt-hour, which for efficient operators is nearer $32,000–$55,000 per coin.

What it is still good for: as a gauge of miner stress and therefore of potential forced supply. Read our full analysis of why the production-cost floor broke for the longer argument.

4. NVT and NVT Signal — bitcoin's P/E, sort of

Network Value to Transactions divides market cap by daily on-chain transaction volume. The analogy is a price-to-earnings ratio, where settlement volume stands in for earnings. NVT Signal smooths the denominator with a moving average to make it usable.

Why it worked, and why it works less well now: when almost all bitcoin economic activity settled on-chain, NVT was a reasonable measure of whether valuation had run ahead of usage. But settlement has migrated — to exchanges' internal ledgers, to the Lightning Network, and above all to the ETF wrapper, where enormous economic volume produces zero on-chain transactions. NVT now systematically flags bitcoin as 'overvalued' for a reason that has nothing to do with valuation. Use it with heavy scepticism, if at all.

5. Stock-to-Flow — retire it

S2F divides existing supply by annual new issuance and maps the result to price on a log scale. It produced a beautiful backtest and a series of failed forecasts. Its core problem is structural: it models supply only, ignores demand entirely, and its fit was achieved on a handful of halving cycles — a sample size too small to support the confidence it was sold with. Its own creator's price bands have been broken repeatedly. Include it here only so you can recognise it and set it aside.

6. Flow-based valuation — the ETF era's actual model

In practice, the model that has explained bitcoin's price best since 2024 is embarrassingly simple: net dollars in versus net dollars out. Spot ETF flows, CME basis, and exchange balances now dominate short-to-medium horizons. The week ending July 11, 2026 is a clean example — U.S. spot bitcoin ETFs took in $197.4 million, their first positive week since mid-May, ending an eight-week streak that removed $8.26 billion, and price rebounded from $57,750 toward $64,000 alongside it.

How to use it: track daily net flows (Farside, CoinGlass), check whether inflows are broad or concentrated in a single issuer, and cross-reference the CME basis to see whether the flow is directional buying or a delta-neutral carry trade. A flow that is really a basis trade is not a vote of confidence in bitcoin; it is a bet on a spread. We covered how to make that distinction in our basis trade guide.

7. Thermocap, Metcalfe and the rest of the long tail

Two more deserve a brief mention, mostly so you know why they are not in the checklist below.

Thermocap values the network by the cumulative dollar value of all miner revenue ever paid — an estimate of total capital burned to secure the chain. Market cap divided by Thermocap gives a multiple that has historically been high at tops and low at bottoms. It is respectable and it is slow, and it suffers from the same problem as production cost: it measures what was spent on security, not what anyone will pay for a coin.

Metcalfe-style models value the network as proportional to the square of active users, using active addresses as a proxy. The theory is sound for communication networks. The proxy is not: address counts are trivially inflated by exchange internal wallets and consolidation transactions, and they are simultaneously deflated by the ETF wrapper, which gives millions of people bitcoin exposure while generating no addresses at all. A metric that goes down when adoption goes up is not measuring adoption.

A note on what the models disagree about right now

It is worth being concrete about how these read against each other in July 2026, because the disagreement is unusually clean.

The flow-based read has just turned marginally constructive: the eight-week outflow streak ended, $197.4 million came back, and price recovered from $57,750 toward $64,000. The on-chain cost-basis read is neutral-to-constructive: long-term holders have flipped back to accumulation, adding 50,000–100,000 BTC on a 30-day basis, while the smallest and mid-sized wallet cohorts are accumulating aggressively. The production-cost read is outright distressed: hashprice at a record-low ~$29/PH/s/day, roughly 15–20% of the global fleet underwater, and JPMorgan flagging bitcoin at 19% below all-in cost.

And one indicator refuses to join in at all: the Coinbase Premium, the crude proxy for U.S. spot demand, has been negative for a record stretch of consecutive sessions since May 19 — through the crash and through the recovery. A market where flows are improving but the domestic spot bid is still absent is a market that has stopped falling without establishing why it should rise.

The lesson generalises. When models built on different foundations disagree, that is not noise to be averaged away — it is a map of where the uncertainty actually lives. In this case it is telling you that the July move is being driven by the mechanical unwinding of a positioning problem rather than by fresh conviction. Whether it becomes something more is a question the June CPI print on July 14 will answer long before any on-chain metric does.

Putting it together: a 10-minute valuation checklist

1. Where is price versus Realized Price? Above = average holder in profit. Below = deep-value territory, historically rare.
2. What is MVRV Z-Score doing? Extremes matter; the middle of the range tells you little.
3. What is Short-Term Holder MVRV? Below 1 means recent buyers are underwater — the cohort most likely to capitulate.
4. What are ETF flows doing, and are they broad or concentrated? One issuer is a weak signal.
5. Is the flow directional or basis-driven? Check CME basis and funding rates.
6. What is hashprice doing? Record lows mean miner stress and potential forced supply, not a floor.
7. Ignore Stock-to-Flow. Genuinely.
8. Ask what would prove you wrong. Write it down before you take the position, not after.

The final discipline is the most important and the least practised: no single model should carry a position. The models disagree right now — flow-based reads improving, on-chain cost-basis reads neutral, production-cost reads distressed — and that disagreement is itself information. It tells you the market is in transition, not in trend. Position sizes should reflect that.