Whoever holds the private key can move the asset. Everything else — the account statement, the fund structure, the monthly report — describes that fact rather than changes it. Custody is therefore not a records-keeping service bolted onto ownership; it is ownership delegated to somebody else, and the document governing the delegation is what a holder actually owns when something goes wrong.
Why is custody the binding constraint?
Because it decides the question before the investment question. In our survey of forty institutions, collected under anonymity and aggregated by institution type, the constraint named most frequently was custody eligibility rather than price or volatility. Allocator mandates we have described work the same way: custody requirements exclude most venues before anyone forms a view on the asset.
This is also why a regulatory change to custody moves more money than a change to any particular token's status. The practical blocker for the asset management side has been that a qualified custodian requirement written for securities and cash does not map onto an asset whose control is a key.
What the arrangements actually differ on
| Arrangement | Who can move the asset | What fails first | What a holder relies on |
|---|---|---|---|
| Self-custody | The holder | Key management: loss, theft, a single device | Its own operational discipline |
| Qualified custodian | The custodian, under a mandate | The custodian's operations, or its solvency | The custody agreement and the regime behind it |
| State trust company | The trust company | The same, under a state charter rather than a federal one | Whether its charter is recognised for the purpose |
| Exchange wallet | The exchange | Everything at once, since the venue is also the counterparty | Terms most holders have not read |
The difference between the rows is invisible on a dashboard and total in an insolvency. It is the first question an allocator asks and close to the last thing a retail interface mentions.
Why does diversifying across funds not diversify custody?
Because institutional requirements narrowed the field of acceptable custodians to a handful, and assets followed. The result is portfolio diversification sitting on top of infrastructure concentration: products that look unrelated to each other share the same operational dependency, and a failure at one propagates through all of them at once.
Most risk frameworks model price and liquidity in detail and treat custody as a yes-or-no field. It is not binary. A framework that cannot answer how much of a portfolio sits behind one custodian is not measuring the exposure that has historically done the damage.
What did the SEC propose on 1 October 2026?
Rules and amendments setting out how registered investment advisers and regulated funds — registered investment companies and business development companies — may hold crypto assets, under the Investment Advisers Act of 1940 and the Investment Company Act of 1940. Two provisions change what is practically possible: crypto assets could be held in self-custody in certain circumstances, and state trust companies could act as custodians for client and fund assets. The proposal also updates requirements for financial statement audits of registered advisers and for broker-dealer custodial services used by funds.
The Commission frames this as removing a barrier rather than adding a regime: its argument is that existing custody rules inhibit an adviser's ability to give crypto-related advice at all. Chairman Paul Atkins described the existing rules as crafted for a bygone era. Nothing is in force — the public comment period runs for 60 days after the proposing release is published in the Federal Register, and a rule this broad usually changes in exactly the operational details that matter.
A second strand is moving alongside it. The Federal Reserve's proposals for payment stablecoin issuers under the GENIUS Act include rules for the Board-supervised firms that safekeep the assets backing a stablecoin, which is the same question asked about reserves rather than about client holdings.
Custody does not end when the service does
An exit from a market closes the service and nothing else. Binance announced its withdrawal from Russia in 2023; in August 2026 it emerged that records it had collected before the exit — full name, date of birth, address, telephone number, passport number and document copies — had been handed to Russian authorities and used in a prosecution over donations totalling more than $700.
The general lesson is not about one firm. Market exit ends the service, not custody of the identity documents already collected, nor the jurisdiction that can compel their production. Any counterparty assessment that treats withdrawal from a market as the end of exposure in that market is measuring the wrong thing, and this applies to data custody exactly as it applies to asset custody.
What users actually choose
Where custody is offered as a choice rather than a requirement, the evidence runs against self-custody. In a Stripe pilot covered on our payments desk, 90% of workers took the platform's own wallet rather than setting one up independently. Demand, on that evidence, is for someone else to hold the keys — which makes the quality of that someone the thing worth regulating and the thing worth checking.
Questions worth asking of any custody arrangement
- Who can sign a transaction, and how many of them have to agree?
- Under which law, and in which insolvency regime, does the holder's claim sit?
- What share of the portfolio — across every fund and product — ends up behind this one custodian?
- What happens to collected identity documents if the service leaves the market?
- Is the arrangement eligible under the mandate as written, rather than under the mandate as hoped?
Questions
- What does a crypto custodian actually hold?
- The private keys, which in practice means the asset. Custody is ownership delegated to a third party, and the legal wrapper around it decides what happens if that party fails — which is why the custody agreement matters more than the account statement.
- Does holding several funds diversify custody risk?
- No, not if the funds share a custodian. Institutional requirements narrowed the field to a handful of names and assets followed, so products that look unrelated can carry the same operational dependency.
- Has the SEC allowed self-custody for investment advisers?
- It has proposed it, in certain circumstances, as part of a tailored custody framework announced on 1 October 2026 that would also let state trust companies act as custodians. The proposal is not in force; the comment period runs for 60 days after publication in the Federal Register.
- Why do allocators care about custody more than volatility?
- Because custody eligibility is a gate rather than a preference. In a survey of forty institutions it was the most frequently cited constraint, and mandates commonly exclude venues on custody grounds before any view on price is formed.