Bitcoin Realized Price: How ETF Shares Freeze the On-Chain Cost Basis
Bitcoin realized price sits at $52,680 today. Spot is $63,866. Read the metric the way it is usually explained, and the conclusion is that the average holder is roughly 21 percent up.
That reading has been drifting out of date for about two years. Not because the number is wrong — it is calculated correctly. Because it now measures something other than what its name promises.
How the metric is built
Every unit of bitcoin carries a price stamp: the dollar rate at the moment it last moved on the blockchain. Sum every stamp and you get realized cap. Divide by circulating supply and you get realized price.
The metric rests on one assumption: that a movement on-chain represents a genuine change of ownership at market price.
Everything depends on that assumption, and it was never more than approximately true.
The version anyone can picture
Think of a gold vault. Each bar has a tag showing what was paid for it. As long as one person collects a bar and another carries one in, the tag gets rewritten.
Now someone launches a fund. It buys a thousand bars, puts them in the vault, and issues shares. The shares trade daily, in size, for years. The bars never move. Not one tag is ever rewritten.
That is what the bitcoin ETFs do.
What actually happens on-chain, and what does not
Creating new shares does move something. Until July 2025 the US structure was cash-only: authorized participants could deliver or receive dollars, never the bitcoin itself. So they bought on the market and the coins went into custody. That step is stamped correctly.
Since 29 July 2025 the SEC has also permitted in-kind creation and redemption, letting authorized participants transact directly in bitcoin. That normally involves an on-chain movement too.
Everything after that moves nothing. Each ETF share passing between two brokerage accounts is a change of ownership with no unit moving on the blockchain. The stamp stays frozen at the creation date, often for years.
Why this differs from exchanges
The obvious objection: exchanges do the same thing. Buying and selling on a venue moves nothing on-chain either.
True — but it works in the opposite direction, and that is the point.
Exchanges produce false stamps. Internal wallet reshuffles send coins across the chain with no sale behind them. That is noise, and there is a remedy: Glassnode computes an entity-adjusted realized cap that discards movements between addresses belonging to the same entity.
ETFs produce missing stamps. Here is the asymmetry: you can filter out movements that were not real. You cannot invent movements that never happened.
The exchange problem is a filtering problem. The ETF problem is lost information, and the existing corrections do not reach it.
Holding period compounds it. Exchange balances circulate. ETF balances age into the long-term holder cohort and stay there.
Two distortions pulling opposite ways
This is what makes it awkward: there is no single adjustment factor to apply.
Upward, one-off. The creation wave since 2024 re-stamped large volumes at high prices. An analysis from mid-July 2026 finds nearly half of supply last moved above $63,100. That is a break from the earlier distribution, and it is real money that genuinely flowed.
Downward, missing, ongoing. When an ETF investor sells into a decline, no realized loss registers on-chain. The metrics built to measure capitulation see none of it. They understate precisely in the phases where they are supposed to mark the floor.
Which metrics are affected, and how much
Not equally. This is the distinction that matters in practice.
Short-term holder metrics stay comparatively robust. By definition they track units that have recently moved, and anything that moves gets stamped.
The worst affected are aggregate realized price and the long-term holder cost basis. Which is unfortunate, because those two are the ones used as cycle anchors.
What the metric still tells you
It is not broken. It measures the cost basis of the supply still reachable on-chain, plus a frozen block that no longer moves.
That remains useful. It is simply not the same thing as the market's average cost basis, which is how the number is described almost everywhere.
What and when: the measurement that settles it
The custody addresses of the major ETFs are known and labelled. That turns the question into arithmetic instead of argument.
The measurement: compute realized cap twice — once in full, once excluding the identified ETF custody clusters — from January 2024 to today.
The finding would be: the gap between the two curves quantifies how much of the rise in realized price comes from balances that have not moved since.
The claim is falsified if the two curves run near-parallel. That would make the ETF share too small to change what the metric says, and this article would have described a problem that is not one.
That is the point where an observation becomes a measurement. We will publish it with the calculation alongside.
Limits of this piece
The US spot ETF share of supply is currently quoted at roughly 6.7 to 6.8 percent. That figure comes from secondary sources and has not been checked here against the custody addresses. Anyone using it as a load-bearing number should recompute it — it can be recomputed.
The size of the two distortions is also open. That they pull in opposite directions follows from the mechanism. How far they offset each other does not.
Sources: Glassnode Research, "The Foundational On-chain Metric: The Realized Cap" · Glassnode Studio, Entity-adjusted Realized Cap · Bitcoin Research Kit, realized_price series, as of 13 August 2026 · U.S. Securities and Exchange Commission, approval of in-kind creations and redemptions for crypto ETPs, 29 July 2025 · Katten Muchin Rosenman, note on the prior cash-only structure, 11 August 2025 · Coin Metrics, free-float market capitalisation adjustment.
Not investment advice.