Collateral Liquidation: The Crash's Second Order

Three Bitcoin (BTC) crashes in eight months drove the liquidations, clearing north of $2bn of collateral off the books of Aave (AAVE) V3, Morpho Blue and SparkLend combined. The machinery worked – but the data, drawn from onchain analytics provider Datum Labs, raises sharper questions about what kind of risk each venue was actually running.

Lending without banks rests on one substitute for trust – collateral worth more than the loan it secures. A borrower posts assets, draws against a fraction of their value, and what sits between the two is the only thing protecting the protocol from a loan it cannot recover. When the price falls faster than a position can be unwound, the code sells the collateral, repays the debt, and hands a bonus to whichever liquidator clears it first.

Aave V3 is a decentralized lending protocol where users can borrow and lend crypto assets without intermediaries; Morpho Blue is a peer-to-peer lending protocol that isolates risk within individual lending markets; SparkLend is a lending platform built on MakerDAO's infrastructure. Some 575 addresses – unique identifiers on the blockchain, analogous to account numbers – competed for liquidation work across all three, but 5 of them captured roughly 87% of the take. The backstop that keeps these protocols solvent is, in practice, a handful of well-capitalized bots that have so far always chosen to show up.

The concentration cuts both ways. It makes liquidations fast and dependable on the events that have happened, yet it leaves the floor under solvency resting on a few balance sheets. How much of that work each venue generates is not a matter of luck. It is a matter of what collateral each one agreed to hold – and on Ethereum, where the bulk of all liquidations occur, the three largest desks made very different choices.

June hit thinner book harder

Measured in raw collateral, the three events sit closer together than their headlines suggest. The weekly print following the early February crash was the largest, clearing roughly $231mn across the three venues, with early October and June landing almost on top of one another near $187mn each. Magnitude is not what separates them. What separated them was which collateral broke and how much capital was still standing to absorb the hit.

October, with Bitcoin sliding from a $126k record towards $103k alongside the heavily bid broader market, cleared the most speculative collateral – long-tail assets outside the major pairs made up 46% of what Morpho liquidated and 19% of Aave's. February, the $90k-to-$60k unwind, was a Bitcoin event through and through, and BTC-correlated collateral dominated at 90% of Morpho's liquidations and 34% of Aave's, force-selling $14.8mn and $72mn respectively.

(Source: Datum Labs, Ethereum Mainnet across Aave V3, Morpho Blue and SparkLend)

June was different in kind. Bitcoin only retested the ~$60k area February had already carved out, holding a level rather than breaking new ground. Ether did the breaking, marking a fresh low and revisiting the $1,500 zone for the first time in a year. The collateral that makes the new low is the collateral that gets sold, so the mix inverted – ETH-correlated positions were 85% of Aave's June liquidations at $142mn of a $168mn week and effectively all of Spark's $3.4mn, while Bitcoin's share of Aave's book more than halved from February's 34% to 13%.

And the feedback runs one way, with each forced sale feeding the next, ETH collateral hitting a bidless market and dragging the price into the following tranche of liquidation thresholds.

The venue detail is where the risk actually lives. Roughly 37% of Morpho's Bitcoin liquidations were cbBTC – Coinbase's (COIN) wrapped Bitcoin token, redeemable 1:1 for BTC and designed to bring Bitcoin liquidity into DeFi – much of it posted by retail users of consumer apps like Coinbase built on top. These are borrowers who tend to read a liquidation not as a risk they accepted but as a reason to never show up again. June opened a second Morpho fault line, where 41% of that week's liquidations were stablecoins – cryptocurrencies designed to maintain a fixed value, typically pegged to the US dollar – concentrated in Pendle principal tokens backed by STRC-linked stablecoins that slipped their peg during recent turmoil around Strategy (MSTR).

Pendle (PENDLE) is a DeFi yield-trading protocol that splits yield-bearing assets into two tradeable components: a Principal Token (PT), which represents the underlying asset redeemable at maturity – similar to a zero-coupon bond – and a Yield Token (YT), which represents the future interest the asset will generate. When the underlying stablecoin loses its peg, the PT trades at a discount, and if that PT is posted as collateral, it can fall below liquidation thresholds rapidly.

STRC is a variable-rate perpetual preferred share issued by Strategy (MSTR) – the publicly listed Bitcoin treasury company formerly known as MicroStrategy – which raised $5.58bn through the instrument. In recent weeks, STRC severely depegged, hitting a closing price of $88.9 against its $100 par value, as pressure on Strategy's stock coincided with broader crypto market stress. Stablecoins backed or collateralized by STRC were caught in that dislocation.

Spark sat at the opposite pole, with almost no stablecoin exposure and no long-tail collateral at all. A venue never chooses how much volatility arrives, only which collateral it agreed to hold when it does – and that single choice is most of the risk.

Rising share of shrinking book

Scaled against each loan book, the hierarchy is clean. Averaged across the three events, liquidations ran 1.41% of Aave's active loans, 0.78% of Morpho's, and just 0.33% of Spark's and the ordering neatly tracks each protocol's design.

Spark accepts the narrowest collateral set under the tightest framework, so little breaks. Morpho keeps each lending market separate, so problems in one market stay contained and don’t affect lenders in other markets. Aave pools deposits and lets any supplied asset be borrowed against the full menu of approved collateral, looped and leveraged positions included – the widest surface and the most to seize.

(Source: Datum Labs, Token Terminal, Ethereum Mainnet across Aave V3, Morpho Blue and SparkLend)

The trend is the sharper signal, and it is what makes June the violent event the dollar totals disguise. Aave's liquidation intensity climbed from 0.65% of its book in October to 1.35% in February to 2.23% in June, a clean escalation in which each crash bit deeper than the last. The denominator did the heavy lifting, but the numerator variable is still relatively significant. Aave's active-loan book on Ethereum collapsed from $25.6bn in September to $7.5bn by the June crash, a 70% contraction in borrowing demand. October liquidated $145mn against a $22.3bn book and barely registered; June liquidated more – $168mn – against a book a third the size, so the same machinery cut more than three times as deep. As the conservative borrowers withdrew, what remained was smaller and riskier, and each event took a larger bite of it. A book that shrinks while liquidating an ever-greater share of itself is the textbook print of adverse selection – the phenomenon where safer participants exit first, leaving a riskier residual pool – the safe money leaves first, and the risk concentrates on whoever stays.

When the yield leaves, the cascade pays

Lay the liquidation weeks over protocol revenue and they sit almost exactly on the peaks. Aave's 2 highest fee weeks from the Ethereum mainnet in the past year were the October and February crash weeks, near $27mn apiece, and Spark's single best week was the October cascade. October's print was partly flattered by a market still trading near its highs, but the relationship held as conditions soured. By spring, weekly Aave fees that once cleared $20mn had compressed towards $5–8mn as capital drained out of decentralized finance (DeFi) – the ecosystem of financial services built on public blockchains, operating without banks or brokers – and yields stopped paying for the risk.

(Source: Token Terminal, Ethereum Mainnet across Aave V3, Morpho Blue and SparkLend)

Against that decaying baseline, the episodic spikes matter more, not less. The June crash week alone generated close to $14mn in Aave fees, several times the surrounding run-rate. Aggregate the three venues and the signature is unmistakable – the year's fattest weekly fee prints all sit directly on top of a liquidation cluster. As organic borrowing drained away, the cascades stopped reading as a tail risk and started carrying the revenue line – and how each protocol was built decided who absorbed the hit.

The data point towards an uncomfortable conclusion: DeFi's largest lending desks may have become, without ever deciding to, structurally long volatility – underwritten less by the demand to borrow than by the periodic violence that empties their books.

The drain has not fallen evenly. Aave's pool did almost all the bleeding while capital rotated rather than simply vanished – Morpho's active loans bottomed in February and climbed 68% back to $3.1bn, above where they began the period, and Spark clawed 26% off the same low to $1.6bn, though still short of its autumn level as collateral value shrank.

The recent series of protocol exploits and stablecoin depeg has made the case that onchain credit is neither as safe nor as well-paid as its headline yields implied, and deposits are now pulled or repriced faster than borrowing demand can replace them.