The protocol has thousands of independent nodes and no operator who can be removed to sever it. A growing share of the bitcoin those nodes secure sits behind the key-management operation of a single company.
Bitcoin's mining and node decentralization are measured continuously by several public dashboards. Its custody concentration is measured by nobody. This page is our attempt at the missing half, built from fund SEC filings rather than from press coverage.
Share of 1,216,021 BTC held by US spot bitcoin ETFs, by the custodian each fund names as sole or primary in its SEC filings. As of July 31, 2026.
Coinbase Custody and Coinbase Prime are the same parent company. Taken together they are the named sole or primary custodian for 84.5% of ETF bitcoin.
The minimum number of entities whose failure or collusion controls a majority. Higher is more decentralized.
Foundry, AntPool and F2Pool together produced just over half of recent blocks. Independent research grouping pools by shared block templates argues the true figure may be 2.
mempool.space pool data, 0xB10C template-similarity research
One custodian is named as sole or primary for the large majority of US spot ETF bitcoin. No second entity is required.
Fund SEC filings, this dataset
Bitcoin's custody layer is more concentrated than its mining layer. Mining decentralization is measured continuously by several public dashboards. Custody concentration is measured by nobody, which is why this figure of 1 has not been published before.
The widely repeated version of this story says two funds are independent of Coinbase: Fidelity's FBTC and VanEck's HODL. That is wrong.
VanEck's FY2025 10-K names Gemini and Coinbase Custody as the bitcoin custodians who hold all of the trust's bitcoin. HODL is not independent. The only two funds with no Coinbase entity named anywhere in their filings are FBTC and DEFI, together 14.1% of ETF bitcoin.
Paul Baran drew these topologies in 1964 to study which networks survive attack. His argument is routinely inverted by people citing it.
Every holder depends on one operator. Its compromise, insolvency, or unilateral action reaches everyone at once.
Several hubs instead of one. Baran put this word in scare quotes and judged it fragile, because destroying a small number of hubs still severs the network. Most arrangements marketed as decentralized are this.
No node whose removal severs the whole. This is the topology bitcoin's protocol has, and the one its custody layer largely does not.
Baran described two poles, centralized and distributed. He put the word “decentralized” in scare quotes, attributed it to others, and treated it as a fragile middle case, noting that destroying a small number of nodes in such a network still destroys communication.
That middle panel, a federation of a few large hubs, is precisely what bitcoin custody looks like today. It is also what most products marketed as decentralized custody actually are. The distributed panel on the right, where no single node's removal severs anything, is what bitcoin's protocol achieved and what its custody layer has not.
This is the same measurement as our Custody Independence Standard, drawn as a picture: how many independent things must fail before the bitcoin is gone.
This measures control of keys. It does not measure beneficial ownership. A custodian holding coins for millions of ETF shareholders is a concentration of control and a dispersion of ownership at the same time. Both are true, and the distinction matters because the main counter-argument is correct as far as it goes: pooled custody does spread beneficial ownership across many more people than the address count suggests.
It does not rebut the control claim. Ownership dispersion tells you who benefits if nothing goes wrong. Control concentration tells you how many parties have to fail before it does. Our timeline of custody failures is fifteen years of the second question mattering more.
A second honest limit: 6 of the 13 funds name more than one custodian without disclosing how the bitcoin is split between them. Those are counted at their primary custodian, which is why the headline is a range rather than a single number, and why the conservative reading is lower than 84.5%.
Finally, this covers US spot ETFs only, because prospectus filings make the custodian a matter of public record. Corporate treasuries, private funds and institutional accounts hold a great deal more bitcoin, and almost none of them disclose who holds the keys. The true custody concentration across all custodied bitcoin is not publicly knowable, and is very unlikely to be lower than this.
Custodians as named in each fund's own filings. Holdings as of July 31, 2026.
| Fund | BTC | Primary custodian | Also named | Filing |
|---|---|---|---|---|
IBIT iShares Bitcoin Trust The FY2025 10-K states the sponsor has no plans to move any of the trust's bitcoin to the additional custodian. The diversification is documentary rather than operational. | 739,066 | Coinbase Custody | Anchorage Digital | FY2025 10-K |
FBTC Fidelity Wise Origin Bitcoin Fund Self-custodied by an affiliate of the sponsor. One of only two funds with no Coinbase dependency. | 171,561 | Fidelity Digital Assets | — | Fund disclosures |
GBTC Grayscale Bitcoin Trust | 133,439 | Coinbase Custody | — | Fund disclosures |
BTC Grayscale Bitcoin Mini Trust Anchorage holds a portion; the split is not disclosed. | 58,894 | Coinbase Custody | Anchorage Digital | 10-Q / FWP |
BITB Bitwise Bitcoin ETF | 36,999 | Coinbase Custody | — | Q1 2026 10-Q |
ARKB ARK 21Shares Bitcoin ETF Four custodians named as holding the trust's bitcoin. No allocation disclosed. | 33,802 | Coinbase Custody | BitGo Bank & Trust, BitGo NY Trust, Anchorage Digital | 10-Q |
HODL VanEck Bitcoin ETF Widely reported as Coinbase-independent. The FY2025 10-K names Gemini AND Coinbase Custody as the bitcoin custodians who hold all of the trust's bitcoin. It is not independent. | 16,616 | Gemini Trust Company | Coinbase Custody | FY2025 10-K |
MSBT Morgan Stanley Bitcoin ETF | 6,231 | Coinbase Prime | BNY | Fund disclosures |
BRRR CoinShares Bitcoin ETF Three custodians named. Allocation undisclosed. | 5,890 | Coinbase Custody | BitGo Trust, Komainu (Jersey) | FY2025 10-K |
EZBC Franklin Bitcoin ETF | 5,634 | Coinbase Custody | — | Fund disclosures |
BTCO Invesco Galaxy Bitcoin ETF | 5,433 | Coinbase Custody | — | Fund disclosures |
BTCW WisdomTree Bitcoin Fund | 2,230 | Coinbase Custody | — | Fund disclosures |
DEFI Hashdex Bitcoin ETF The only fund custodied solely by BitGo, and one of only two with no Coinbase dependency. | 226 | BitGo Trust Company | — | FY2025 10-K |
| Total | 1,216,021 | |||
Proof of Custody, Bitcoin ETF Custodian Concentration, proofofcustody.io/research/custody-concentrationOne company, Coinbase, is named as sole or primary custodian for about 84.5% of the bitcoin held by US spot bitcoin ETFs, and is named somewhere in the filings of funds holding about 85.9%. Only FBTC and DEFI name no Coinbase entity at all.
One. The Nakamoto coefficient is the minimum number of entities whose failure or collusion controls a majority. For US spot ETF custody, a single custodian exceeds a majority on its own. For mining pools the figure is three, so bitcoin's custody layer is more concentrated than its mining layer.
No. It is widely reported as one of two Coinbase-independent funds, but VanEck's FY2025 10-K names Gemini and Coinbase Custody as the bitcoin custodians who hold all of the trust's bitcoin.
Not at the protocol layer. Nodes and mining remain distributed, and holding coins confers no protocol power. The concentration is in control of keys: how many parties must fail before specific bitcoin is lost. That is a different question from who benefits from the bitcoin, and the historical record of custody failures is about the first one.