Crypto Banks and Custodians: Where a Private Holder Keeps Coins and Receives Fiat

A private holder choosing where to keep coins and where to receive fiat of crypto origin is answering two different questions with one decision. The first is legal: whose law decides what happens to the position if the holder of the keys fails. The second is operational: which account will accept the sale proceeds without a freeze. A bank licence answers the second question well and the first one only in some jurisdictions; a trust charter answers the first well and the second not at all.

This page compares eighteen providers across six jurisdictions on the axes that survive a bankruptcy or a compliance review: the licence and who supervises it, what happens to client crypto in the custodian's estate, whether a private individual or a trust can even be a client, the entry threshold, the custody model, the fiat rails and the published price. The structural context — keys, tax, succession — sits on the crypto for private wealth hub; the exit mechanics are on the OTC settlement page.

Exclusion criteria: what removes a provider before the comparison starts

Four filters cut the list faster than any feature table.

1. Client type. Most of the institutional segment does not open accounts for a natural person at all. BitGo, Copper, Komainu, Zodia, Fireblocks Trust Company and Anchorage are built for funds, registered advisers and corporate treasuries. A family reaches them through a trust, a fund or a private trust company — or through its own manager.
2. Residency perimeter. Xapo does not accept US persons; FV Bank does not serve US residents in retail mode; Dukascopy excludes residents of Russia, Belarus, the United States, Japan and a published list of further territories; Bank Frick's official perimeter is the EEA, Switzerland, the UK, Hong Kong and Singapore, with anything else "in very exceptional cases"; HashKey does not serve residents of mainland China.
3. Fiat or coins, not both. Copper, Komainu, Zodia, BitGo and Fireblocks hold coins and do not run a bank account. If the objective is to receive fiat proceeds, a custodian is the wrong instrument regardless of how good its segregation is.
4. Operating status. A conditional approval is a permission to finish building a bank, not a bank. Morgan Stanley Digital Trust holds preliminary conditional OCC approval and has not begun operations; of the federal crypto trust banks only Anchorage and Circle's vehicle actually operate, as the OCC trust charter page sets out.

Licence, supervision and the legal position of client assets

Three legal regimes do the real work in that last-but-one column. In Switzerland the test is statutory: Article 242a of the Debt Enforcement and Bankruptcy Act, inserted by the distributed-ledger statute of 25 September 2020 and in force since 1 August 2021, orders the surrender of crypto-based assets from the estate where the bankrupt undertook to hold them ready for the third party at all times and the assets are either individually assigned to that party or allocated to a community in which the party's share is clear. The second limb matters commercially: an omnibus wallet does not defeat segregation in Switzerland provided the share is determinable — the fatal defect is a custodian that cannot show which coins were held for whom.

In the United States the same outcome comes from charter type rather than statute. A national trust bank or a New York limited purpose trust company holds fiduciary assets off its own balance sheet, so they do not fall into the estate; what the charter never delivers is insurance, and the OCC route explicitly excludes deposits, lending and FDIC cover. Wyoming's SPDI adds a third mechanic — a statutory 100% reserve against fiat deposits with fractional lending prohibited, reported to the state on the SPDI call report.

Gibraltar shows the fault line most clearly because Xapo splits it across two companies: fiat sits in the bank and is covered to EUR 100,000, coins sit in Xapo VASP Limited and are covered by nothing but the custody architecture. Read every provider the same way. The question is never "is it regulated" but "which entity signs my agreement, and what does that entity's insolvency law say".

Access, assets and services

Published pricing

Custody pricing in this segment is mostly negotiated, and most providers publish nothing. The figures that do exist are worth holding in one place, because they set the anchor for any quote a private holder receives. The custodian benchmarks below were previously kept inside the Morgan Stanley profile; they belong here, alongside what the banks publish.

Two readings follow. First, the spread between a specialist custodian at 0.05–0.50% a year and a bank that prices the whole relationship is not a spread in the same units: the bank fee buys deposit rails, a compliance function that will accept the source-of-funds file, and often a Lombard line — the custodian fee buys safekeeping only. Second, a published price is itself a signal. Xapo, HashKey, Kraken and Dukascopy publish tariffs because they serve a retail-adjacent client; the private-banking names quote per mandate, which in practice means the comparison has to be run on identical assumptions of size, asset list and trading frequency, or it compares nothing.

Bank, custodian or self-custody

The three constructions fail in different places, which is why serious holders use more than one.

A bank is the only construction that closes the fiat loop. Sale proceeds land on a statement inside an institution that has already reviewed the origin of the coins, the source-of-funds file is a routine part of onboarding rather than an obstacle at the gate, and a Lombard line turns the position into liquidity without a disposal — which also avoids creating a taxable event, a point covered on the country tax comparison. The costs are price, a conservative asset list and, in the Gibraltar and Puerto Rico cases, the fact that the deposit guarantee stops at the fiat side or does not exist at all.

A qualified custodian is the only construction that satisfies a regulated manager. A fund or a registered adviser is required to place client assets with a qualified custodian, so for a trust or a fund the custodian is not a preference but a condition — a professional trustee will not hold private keys itself. The custodian brings segregation, audit and, sometimes, a disclosed insurance limit; it brings no account, no card and no bank reference letter.

Self-custody is the only construction with no counterparty risk at all, and it replaces that risk with operational risk, which has destroyed more crypto fortunes than volatility. At scale it means multisig or MPC rather than a single seed phrase, and it creates its own succession problem, handled on the inheritance of digital assets page. Its other cost is invisible until a sale: coins that have spent years in self-custody produce the hardest origin file to assemble, so the bank that eventually receives the proceeds asks the most questions.

Choice by profile

A worked example

A holder with 40 BTC bought on an exchange in 2017, moved into a 2-of-3 multisig in 2019, now resident in a jurisdiction that taxes disposals, wants USD 1m of liquidity without selling and eventually a clean exit for the rest.

Selling 40 BTC through an order book is the wrong first move, and a Lombard line is the cheaper one. At roughly 40% LTV a BTC-backed credit line at Xapo delivers the liquidity against part of the stack; at Sygnum the same line runs in CHF, EUR, USD or SGD against a token list that includes staked SOL, so the collateral keeps earning. Neither route creates a disposal. On price, the Lombard interest is compared not against zero but against the tax that an outright sale would trigger plus the custody fee on the remaining position — at 0.35–0.50% a year on a USD 4m position, custody is USD 14,000–20,000 annually, which is small next to a capital-gains bill and large next to BitGo's 0.05–0.15% ladder if safekeeping is all that is needed.

For the eventual exit, the sequence is fixed: the origin file first (2017 exchange KYC and trade reports, a signed message from the multisig address proving control, blockchain forensics on the address, returns for the years of ownership), then onboarding at the receiving bank, then the OTC trade, then settlement. Reversing steps two and four is what produces frozen proceeds.

Typical mistakes

• Reading the group brand instead of the contracting entity. Sygnum in Singapore is not a bank; Xapo's coins sit in a VASP, not in the bank; HashKey's exchange, custody, OTC and asset-management arms hold different licences. The protection regime attaches to the entity on the agreement.
• Treating a conditional approval as an operating bank. Morgan Stanley Digital Trust, BitGo's national charter, Coinbase's and Ripple's vehicles all hold approvals that permit them to finish organising. Assets today sit with the predecessor entity.
• Assuming a licence implies insurance. Only Coinbase publishes a figure (a USD 320m commercial crime policy). A trust charter, an SPDI and a Swiss banking licence all deliver segregation and none of them delivers cover for the coins.
• Confusing deposit protection with asset protection. esisuisse, the Gibraltar scheme and the Georgian scheme cover fiat; the coins are covered by segregation law. In the UAE no scheme with a published limit has been launched at all.
• Opening the account after the sale. Onboarding at a crypto-competent bank takes weeks with a complete file and fails without one. The evidence pack is prepared before the trade.
• Ignoring omnibus mechanics. Under Swiss law an omnibus wallet still yields surrender if the share is determinable; in a US prime product the trading portfolio may sit outside qualified custody entirely. Both facts live in the contract, not in the marketing.