The Architecture of Failure

Download the full report →

Key Takeaways

Prime Trust did not fail from bad luck. It failed because of a custody architecture that let a single entity hold the private keys, take title to client assets in its own name, and commingle them until they could no longer be traced. The largest clawback action alone seeks roughly 970 million dollars from a single firm.

The clawback suits are stronger than most observers assume. A July 2025 ruling in the Prime Core case established that the commingled client assets held at the time of filing were property of the bankruptcy estate, so transfers made on the way out can be recovered under Section 547. The trust is applying that mechanism methodically against firms such as Swan, Strike, Compass Mining, Fold and Galaxy Digital.

The risk is not confined to direct custody customers. It reaches anyone holding bitcoin through an intermediary, including holders of bitcoin treasury-company securities and spot bitcoin exchange-traded funds, most of whom cannot see, evaluate, or choose the custody arrangement sitting beneath their position.

Contractual language describing ownership is not the same as ownership. A regulated charter is not the same as physical segregation, and a qualified custodian is not necessarily one that holds client assets apart from its own.

The only durable protection is architectural. Multi-institution custody splits signing authority across independent regulated institutions in a two-of-three quorum, so that no single party can move, lose, or take title to client assets. Onramp's implementation adds a per-incident insurance backstop underwritten at Lloyd's of London for the residual risk.

Executive Summary

The lawyers handling the Prime Trust bankruptcy are going after the businesses that pulled their bitcoin out in the months before it collapsed, trying to claw that money back so it can be split proportionately across everyone who was left short. The largest of these actions seeks roughly 970 million dollars from a single firm.

Here is the part that should make anyone holding bitcoin pay attention: a judge has already ruled this is allowed, and the same weakness that made it possible exists in most of the places that hold bitcoin for other people today.

This report makes three arguments.

First, Prime Trust did not fail because it was unlucky or uniquely badly run, although both of those are also true. It failed because of a custody architecture that let a single entity hold the private keys, take title to client assets in its own name, and commingle those assets until they could no longer be traced.

Second, the claims now being pursued against firms such as Swan, Strike, Compass Mining, Fold and Galaxy Digital are stronger than most observers assume, because a July 2025 ruling has already established that commingled client assets held by the failed custodian became property of the bankruptcy estate.

Third, this is not a problem confined to direct custody customers. It reaches anyone who holds bitcoin through an intermediary, including holders of bitcoin treasury-company securities and spot bitcoin exchange-traded funds, most of whom have no visibility into the custody arrangements sitting beneath their position.

Our conclusions are as follows. Contractual language describing ownership is not the same as ownership. A regulated charter is not the same as physical segregation. The only durable protection against the pattern that produced the Prime Trust collapse is an architecture in which no single party can move, lose or take title to client assets. That principle, and the custody model that implements it, is the subject of the second half of this report.

What's Inside

The report is organized in the following sections:

• The Anatomy of the Failure — How an infrastructure custodian that sat beneath firms millions of people used lost access to legacy wallets, drew on pooled customer funds for roughly eighteen months, and ran an on-chain shortfall that peaked near 1,400 bitcoin, alongside the single-page clause that both granted and denied customer ownership.
• Why the Claimants Have a Stronger Case Than People Think — The Section 547 preference mechanism, the July 2025 ruling that made commingled client assets property of the estate, and the drafting choices that make the largest complaint hard to defend.
• The Layers Most People Do Not See — Why the same structure reaches direct custody customers, bitcoin treasury-company shareholders, and spot bitcoin ETF holders alike.
• The Structural Answer — Multi-institution custody: a two-of-three quorum across independent regulated institutions, why it breaks the pipeline from custodian failure to asset capture, and the measurable reduction in catastrophic-loss probability.
• What Every Bitcoin Holder Should Be Asking — The specific questions to put to a custodian, a treasury company, or an ETF, most of which holders cannot currently answer.

Download the full report →

Related Reading

• What was the Prime Trust bankruptcy?
• Can bitcoin be clawed back in a bankruptcy?
• Can a custodian take title to your bitcoin?
• Who custodies bitcoin ETFs?
• Preference (Section 547) (Glossary)
• Commingling (Glossary)
• Omnibus account (Glossary)
• Qualified custodian (Glossary)
• Bankruptcy remoteness (Glossary)
• Multi-Institution Custody (Glossary)