The 2022 collapses are usually told as separate stories. Celsius froze withdrawals in June. Voyager filed in July. BlockFi unwound into the fall. FTX detonated in November. Different companies, different balance sheets, different proximate causes — a hedge fund here, an exchange run there.
Told that way, each looks like its own accident. A specific bad bet, a specific bad actor, a specific piece of bad luck.
That framing is comforting and wrong. The companies were different. The failures were the same. Underneath the particular triggers, every one of these blowups shared the same three structural defects — and each defect is a violation of what collateral is actually supposed to do.
Collateral exists to protect principal at the moment the borrower cannot: present, liquid, and enforceable under stress. Each of the three failure modes below is a way of quietly removing one of those properties while the system still appears to be functioning.
Failure mode one: the same collateral backed more than one claim
In a sound collateral system, a unit of collateral secures exactly one obligation. It is segregated, it is identifiable, and the lender who is relying on it has an exclusive call on it.
In every 2022 blowup, that exclusivity was gone. Customer assets were not held in segregated accounts. They were treated as assets of the platform, commingled across operations, then lent to third parties, deployed into yield strategies, and pledged as security for the platform’s own borrowing. The same asset supported multiple claims at once.
Under normal conditions this is invisible. Every claimant looks fully covered, because no one is asking to be paid at the same time. The structure only reveals itself under stress, when multiple parties reach for the same asset simultaneously and discover it can satisfy one of them, not all of them.
This is the cleanest example of collateral that satisfies the description — the asset exists, it’s on a statement — while failing the function. It was never present in the sense that matters. It was present for one claimant at a time, in a system that had sold it to several.
The technical name is rehypothecation. The plain version: they lent out the thing they told you was set aside for you.
Failure mode two: short-term promises funded by long-term bets
The second defect is a maturity mismatch, and it is the one that turns a liquidity problem into an immediate collapse.
These platforms offered depositors withdrawals on demand. That is a short-term liability — the money can be called at any moment. To generate the yields they had promised, they deployed those same deposits into strategies that were not liquid on demand: loans to counterparties with fixed terms, positions in protocols that couldn’t be exited quickly, bets with longer horizons than the deposits funding them.
An asset that is long-term or illiquid cannot satisfy a liability that is short-term and callable. The two are structurally incompatible. A system built this way cannot survive a period when withdrawals exceed the rate at which its assets can be turned into cash — and that period always arrives eventually, because the entire arrangement depends on depositors not asking for their money at the same time.
This is the second property of real collateral being stripped out: liquidity. Collateral has to be convertible to cash on the timeline the obligation requires. An asset you can only sell “eventually,” or only at a fire-sale price once everyone moves at once, is not liquid in the way the obligation needs. It looks liquid right up until it’s tested.
The mismatch was always fatal. It simply took stress to reveal it.
Failure mode three: the lender was never actually holding the collateral
The third defect is counterparty dependency — every layer of intermediation between the lender and the collateral is a point where the chain can break.
In these failures, depositors believed they had exposure to Bitcoin-backed loans. What they actually had was exposure to the entity managing those loans. The Bitcoin sat somewhere downstream — with a third-party custodian, a trading counterparty, an exchange, another platform. When one of those intermediaries failed, the collateral became unreachable, regardless of whether it still existed.
This is the third property gone: enforceability. Collateral has to be something the lender can actually reach and act on without depending on a counterparty’s solvency or cooperation. The moment your access to the collateral runs through someone else’s balance sheet, you don’t hold collateral. You hold a claim against that someone else, and you’re only as protected as they are solvent.
When custody and credit are bundled into the same entity, or strung across a chain of intermediaries, the depositor’s protection is only as strong as the weakest party in the chain. In 2022, the chain broke in a different place each time. That it broke at all was the constant.
Celsius ran all three at once
Celsius is the clearest case because every mode operated simultaneously and in public.
At its peak it held roughly $25 billion in customer assets and offered yields up to 18% — returns with no relationship to what conservative, collateral-backed lending against borrower interest could actually produce. That number was the tell. The yield could only be paid by violating the structure.
Customer assets were commingled on Celsius’s balance sheet, not segregated (mode one). They were deployed into external strategies that were not liquid on demand, while depositors retained the right to withdraw on demand (mode two). And a significant share of the risk ran through external counterparties — most notably Three Arrows Capital, whose collapse in June 2022 directly impaired assets that were supposedly backing depositor balances (mode three).
On June 12, 2022, Celsius froze all withdrawals. Roughly $8 billion in customer assets were locked. Six weeks later it filed for bankruptcy with a hole of about $1.2 billion.
The important question the Celsius collapse raises is not why the company failed. It is why Bitcoin collateral did not protect depositors. The answer is that Celsius was not, in any structural sense, a Bitcoin-backed lender. The Bitcoin did not fail. The structure failed — because it had abandoned every property that makes collateral protective in the first place.
Why this matters more than the triggers
It is tempting to read 2022 as a run of bad luck — a hedge fund here, an exchange there. But the triggers were interchangeable. Any one of these platforms would have failed on a different trigger, because the defects were already in the structure before the trigger arrived. The trigger only set the timing.
That is the difference between a system that holds collateral and a system that says it does. The three failure modes are not exotic. They are the three specific ways an operator can keep the appearance of collateral while removing its substance: sell the same unit twice, fund short promises with long bets, and put a counterparty between the lender and the asset.
A system that refuses all three — segregated collateral, matched liquidity, and custody the lender can actually reach — does not survive 2022 by being lucky. It survives by never having taken on the defects that made 2022 fatal.
The blowups were not a verdict on Bitcoin as collateral. They were a verdict on structures that treated the word “collateral” as a label rather than a function. The asset was never the problem. The structure always was.