# Custody concentration, the hidden risk of crypto ETFs

> Coinbase holds 84.5% of the bitcoin in US spot ETFs. The prospectuses document a dependency that daily flows do not show.
> By José Maldonado · Bitcoin · 2026-10-19

An investor spreads their money across four bitcoin ETFs from four different asset managers and believes they have diversified. In most cases they have diversified nothing, because all four funds hand the keys to the same custodian. According to the tally that [Proof of Custody](https://proofofcustody.io/research/custody-concentration) published on August 3, 2026, based on the funds' own 10-K, 10-Q and 424B3 forms, Coinbase appears as sole or primary custodian of 84.5% of the bitcoin held by US spot ETFs: 1,027,618 coins out of a total of 1,216,021, as of July 31.

The SEC approved these vehicles in January 2024, more than two and a half years ago, after a decade of applications, and they solved a real problem: price exposure from any brokerage account without custodying anything. That same design shifted the risk from the twelve-word phrase on a piece of paper to a contract with a regulated entity, with its substitution clauses, its policies and its limits.

Public debate about these funds is measured in flows, and the first half of October 2026 provided plenty of material: on October 7, $484.9 million left in a single session, the largest withdrawal since June 25, according to Farside data. The figure changes every day. The structure that makes it possible does not change, and it rarely makes headlines.

## Four funds, a single guardian

A commodities ETF does not keep the bars in the asset manager's vault, but with an authorized, separate and audited custodian. Bitcoin replicated that scheme because US securities regulation requires registered funds to have a qualified custodian.

In January 2024, almost three years ago, very few companies had the fiduciary charter, the operating track record and the policies to pass a board of directors' review. Investor risk stopped being in the protocol and moved to whoever signs the transactions.

The breakdown can be reconstructed because each fund is required to name its custodian in its documentation. Aggregating by corporate group, with data as of July 31, 2026, the map looks like this:

Custodian group

Funds

Bitcoin

Share

Coinbase (Custody + Prime)

10 of 13

1.027.618

84,5%

Fidelity Digital Assets

1 of 13

171.561

14,1%

Gemini Trust Company

1 of 13

16.616

1,4%

BitGo Trust Company

1 of 13

226

marginal

Behind thirteen vehicles there are, therefore, four groups with their own custody. Coinbase Custody covers nine funds and 84.0% of the bitcoin, and Coinbase Prime adds half a point with Morgan Stanley's fund. Fidelity custodies its own vehicle, Gemini VanEck's and BitGo Hashdex's.

Six of the thirteen funds name more than one custodian without publishing the internal split, so the tally is done by primary custodian and the headline figure is a range, not an exact number. Using the strictest methodology, which excludes those mixed cases, [CryptoSlate](https://cryptoslate.com/over-80-of-bitcoin-etf-assets-hit-coinbase-custody-choke-point-with-74b-at-risk/) calculated 80.8% in April 2026. [SatsIntel](https://www.satsintel.io/en/indicators/concentration) puts the share at 84.0%.

## The fund that was believed to be independent and was not

The most repeated version of this story says that two funds are independent of Coinbase: Fidelity's FBTC and VanEck's HODL. The second is not, and the document that disproves it is signed by the issuer itself. VanEck's 10-K for fiscal year 2025 identifies Gemini Trust Company as custodian and Coinbase Custody Trust Company as additional custodian, and states that both hold all of the fund's bitcoin in its name.

The fund also reserves the right to add or terminate custodians and change the allocation with a simple prospectus supplement, without otherwise notifying unitholders. That discretion leaves real diversification in the hands of the issuer, not of the person buying the share.

The funds with no Coinbase entity named in their documents are therefore two: Fidelity's FBTC, self-custodied through its own subsidiary, and Hashdex's DEFI, custodied solely by BitGo. Together they account for 14.1% of the segment's bitcoin. The part of the market that does not depend on Coinbase at all is six times smaller than the part that does.

The BlackRock case illustrates the other extreme. Its 2025 10-K confirms Coinbase Custody as custodian of IBIT's bitcoin and Anchorage Digital as an additional custodian available, with a sentence worth reading in full: the sponsor has no plans to move bitcoin to that second custodian. Diversification exists on paper, not in practice.

## Why almost everyone signed with the same provider

The concentration was not born of an agreement, but of a race with a single prepared lane. Coinbase Custody Trust Company has operated under a New York fiduciary charter since 2018 and with SOC audits, so when the January 2024 approvals arrived it was the counterparty that already met all the requirements. Issuers who arrived later inherited the template: launching a new fund was cheaper and faster on the infrastructure the rest of the market was already using.

The trickle continued in 2026. The Morgan Stanley Bitcoin Trust began trading on April 8 on NYSE Arca with Coinbase Prime and BNY in the documents, according to [the asset manager's announcement](https://www.morganstanley.com/press-releases/msim-enters-with-launch-of-morgan-stanley-bitcoin-trust). It was the first US investment bank to launch its own fund and it used the same operational backbone as the rest of the market. For a fund already in operation, changing custodian means migrating keys and taking on new operational risk in the process, a cost almost no board accepts as long as the current provider has not failed.

## What breaks when the custodian fails?

The risk is not theft, and that is the part that is usually told badly. Keys are kept in cold storage, assets are segregated from the custodian's balance sheet and there is insurance coverage. What those protections do not cover is everything else.

A prolonged service outage, an incident in the signing infrastructure, a dispute with the regulator or an insolvency proceeding can block transfers for days. The damage that reaches the unitholder takes three forms documented in the prospectuses themselves: deviations between the share price and its net asset value, temporary suspensions of creations and redemptions and, in the extreme scenario, the liquidation of the fund.

Scale aggravates any incident. IBIT accounts for roughly two-thirds of the segment's assets, so a problem at its custodian does not stay within that fund: creation and redemption operations share a provider with eight other vehicles. Buying three different ETFs to spread risk only works if they custody in different places, and those combinations are a minority.

## What the 84.5% figure does not say

The central counterargument is correct and worth stating before going further: concentration of control is not the same as concentration of ownership. A custodian that holds coins for millions of unitholders does not own those coins and cannot dispose of them, and account segregation is designed to keep them out of an estate if the company goes bankrupt. Both things are true at once: control is concentrated and ownership is dispersed.

The second objection also carries weight. The segment has operated since 2024 without a known custody incident, with published audits and a commercial crime policy. The absence of accidents is not the same as the absence of exposure, but it is evidence and it would be dishonest to ignore it.

And mandating diversification by rule has its own costs: every key migration introduces operational risk and multiple custodians raise expenses, with the added danger of spreading bitcoin among entities less prepared than the only one that currently custodies it well. Standardized disclosure of the split and tested contingency plans are a more sensible starting point than a diversification mandate.

There is also a known gap within what is measurable and an ocean outside it. Within, those six funds that do not publish their internal allocation. Outside, everything that does not trade: corporate treasuries, private funds and institutional accounts hold far more bitcoin and almost never disclose who has the keys. The real concentration of all custodied bitcoin is not in any public source today.

## The door the flows leave through does not change

Every time a record outflow day is published, the conversation examines institutional demand and wholesale sentiment. Almost no one examines the other half of the structure: the contract that underpins each share and that, in four out of five cases, points to the same company.

Flows come and go through the same door, and the door does not change in January or in October. We will know whether 84.5% was the ceiling of this model or only its starting point when a large fund moves bitcoin to its secondary custodian or a regulator demands the full breakdown. Until then, the flow data tells the mood of the day and the custody data tells the risk of every day.
