pledgeDecide who can move your Bitcoin — before you need to.
11 custody providers scored on the v3.2 methodology. Custody is a control decision before it's a brand decision. Pick the model first — self, multisig, collaborative, institutional — then compare providers within it.
AnchorWatch
Unchained
Anchorage DigitalAnchorWatch changes the custody conversation by making insurance scope part of the product thesis.
Which custody model fits the way you actually hold?
The right answer depends on size, time horizon, and whether anyone besides you needs to recover the BTC. Five common profiles below, each with the model we'd recommend and the providers in that model worth looking at first.
Not a substitute for individual advice. Custody choice should match your operational reality — what happens if you lose a key, get sick, or die — not just your balance.
Single hardware wallet with a seed phrase backup is the right answer. Multisig adds complexity without meaningful protection at this size.
Single-sig becomes a single point of failure. Collaborative multisig — two keys with you, one with a service — survives a key loss and locks out the service from solo moves.
At this size, recovery becomes a real risk. Lloyd's-backed insured custody covers theft + key-loss scenarios self-custody can't. Trade-off: counterparty exists, but recoverable.
Qualified custodian status, SOC-2 audits, segregated cold storage, board-controlled withdrawals. Self-custody is logistically possible but operationally unworkable for an entity.
Self-custody fails its first test when you can't run it anymore. Multisig with a designated co-signer + a clear recovery doc is the floor; some services formalize the inheritance trigger.
Sorted by Pledge score — same 8 factors, applied to custody.
Research = active full review. Watch = monitoring before deeper work.
One score is a verdict. Eight show you the trade-offs.
The ranking above collapses eight safety factors into one number. This matrix unfolds them again: every scored provider as a row, every factor as a column, each cell shaded by its 0–10 sub-score and weighted by the methodology. Read across a row to see a single provider's shape; read down a column to see where the entire field is strong — or, in the case of insurance, conspicuously soft.
Where every provider is strong — and where the field is soft.
How to read it.Each cell is one provider's 0–10 score on one safety factor; the column header carries that factor's weight in the overall score. The Weighted column is the weighted average of the eight cells in its row (the same number that ranks the providers above). Reading down a column shows where the whole field is strong or soft — insuranceis the category's softest factor (field average 7.0 / 10), because most custody products rely on architecture and segregation rather than a named, fully-disclosed insurance policy. AnchorWatch tops the matrix at 8.2 / 10.
Source: Pledge custody research, methodology v3.2 (8 factors, weights summing to 100%), snapshot 2026-05-24. Sub-scores are best-effort grades from public disclosure — verified or estimated-confidence per provider — and the weighted total is recomputed from the published weights, so it cannot drift from the ranking. Scores are editorial judgments, not a guarantee of safety. Not investment advice.
11 provider front doors. Fewer hands on the keys.
A custody brand is a wrapper; the bitcoin moves only when the underlying key agent signs. Mapping each provider to the institution(s) that actually hold a key, the 11 tracked providers resolve to 15 underlying holders — and 3 of them sit behind more than one provider. BitGo alone is a key-holder behind 4 of the 11: an operational failure or court order at BitGo would reach across products that look independent from the outside.
Same shape as the spot-ETF story below the surface: diversifying across brands does not diversify the institution holding the keys.
The honest counterweight: in 4 of the 11 (the self-custody and collaborative-multisig models), you hold a signing key too — so no single institution above can move the coins alone.
How to read it. Each bar is one institution that holds or co-controls a signing key, sized by how many of the 11tracked providers depend on it. A bar in orange sits behind more than one brand — that is where the apparent diversification quietly collapses. BitGo is the most concentrated dependency at 4 of 11.
Source: Pledge custody research. Each provider's key-holders are derived strictly from that provider's own published custody model and per-claim citations (collaborative-multisig key splits and named key agents / qualified custodians) — not inferred. The end client's own key in self-custody and collaborative models is counted as honest context but excluded from the bars, since each client is a distinct person rather than a shared institution. Editorial mapping, not a guarantee; not investment advice.
How long they've held the line — and what's on the record.
A high safety score is a snapshot; tenure is the part you can only earn with time. Every provider sits on one shared year axis, oldest first — from BitGo (2013) down to the newest entrant. The bar runs from founding to today; the markers are the events the provider has actually disclosed. 6 of 10 carry no disclosed incident; 4 carry at least one.
A long clean bar is the strongest signal here — but read the markers, not just the length. A disclosed incident is not automatically disqualifying; an undisclosed one is the real risk.
How to read it.Each bar starts at the provider's founding (or first live custody operation) and runs to today, so the length is roughly “years operating.” A diamond is a regulatory or charter milestone; a red triangle is an event the provider has disclosed. BitGo, founded 2013, is the longest-running here. Note the disclosed incidents that did nottouch custodied bitcoin — Coinbase's 2025 data breach and Gemini's 2022 Earn halt both hit other parts of the business, not the custody product — which is exactly the distinction a track record should make visible.
Source: Pledge custody research. Founding years, charter milestones, and incidents are taken strictly from each provider's own published record (loss-history summaries, regulatory filings, and per-claim citations) and dated to the year stated there — not inferred. The axis ends at the 2026-05-24 research date. Onramp is tracked but omitted from this chart because its founding year is not stated in itsrecord — we don't guess a date. A disclosed incident is not a verdict; absence of a bar marker means no incident is stated, not a guarantee none exists. Not investment advice.
The ETF custodians are the retail custody menu.
Two verticals, one short list of institutions. Every one of the 5 qualified custodians that hold the bitcoin behind the U.S. spot Bitcoin ETF complex is also a scored Pledge custody provider or the key-holder behind one. That join covers 100% of the $78.3B in cohort AUM across 12 funds: Coinbase alone backs $63.9B of ETF assets and is a 7.8/10 Pledge custody provider at the same time.
The concentration in the ETF wrapper is the same concentration behind the retail custody menu. Diversifying across ETF issuers and across retail custody brands can still route everything to one of these 5 names.
How to read it. Each row is one institution that shows up in both verticals. The left bar is the ETF AUM it custodies; the chip in the middle is its Pledge custody score; the pips on the right count how many tracked retail custody brands route a signing key to it. Read down the rows and the two concentrations rhyme: Coinbase dominates the ETF dollars (82% of cohort AUM), while BitGo is the busiest underlying holder on the retail side, sitting behind 4 brands.
Source: Pledge ETF research (each fund attributed to its primary disclosed custodian; AUM as last verified) joined to Pledge custody research (each provider's key-holders derived strictly from its own published custody model). The two datasets name the same institutions slightly differently (e.g. “Fidelity Digital” vs “Fidelity Digital Assets”); the bridge reconciles them. AUM is an illustrative snapshot, not a live feed. Editorial mapping, not a guarantee; not investment advice.
Why these 3 specifically?
AnchorWatchUnchainedAnchorage DigitalHonest about what we've verified— and what we haven't.
All 11providers scored against v3.2's 8 factors. Launch scores carry a verified or estimated-confidence flag (entity and core facts verified; estimated means some sub-factors are inferred from public disclosure), openly revised as deeper verification completes. Pledge's third tier — pending verification, where a core claim can't be confirmed and the product is kept but never headlined — applies elsewhere on the site; no tracked custody provider is pending today.
About → How Pledge makes money ↗Across collaborative, self-custody, and institutional models.
Custody model, key management, insurance, regulatory, recovery, transparency, track record, loss history.
Best-effort grades from public disclosure; refined as deeper verification completes.
Published. No hidden formulas.