DeFi Liquidation Cascades: How the Spiral Drains Your Collateral

You deposited ETH, borrowed USDC, and felt safe. Then the price moved 15% in an hour and your health factor flashed red. Here's why cascades accelerate, the real events that broke major protocols, and the buffers that separate survivors from statistics.

By DifiCalc Research Team · Published Sep 17, 2026 · Reviewed Sep 17, 2026 · 9 min read

It's 3 AM and your Aave health factor just flashed red. You put down $50,000 of ETH as collateral and borrowed $30,000 of USDC against it — a comfortable 1.67 ratio when ETH was steady. Then ETH dropped 18% in twenty minutes. Your health factor slid from 1.4 to 1.05 to 0.97. Liquidator bots, which monitor the chain for exactly this, repaid part of your USDC debt in exchange for your ETH plus a 5% bonus — automated, no warning, no negotiation. By the time you opened the app, you'd lost $5,200 of collateral, and the dump had pushed ETH down another 3%, putting the next tier of borrowers underwater. The cycle was now feeding itself.

If you've ever held a leveraged DeFi position during a crash, that scenario is you — or it will be the next time. This piece breaks down what a liquidation cascade actually is, why it accelerates instead of self-correcting, the real events that broke major protocols in 2020–2022, and the concrete buffers that keep you on the right side of the spiral.

TL;DR. A liquidation cascade is a self-reinforcing cycle where falling collateral prices trigger automated liquidations, whose forced selling pushes prices lower, triggering more liquidations. The cycle ends when either buyers step in or positions are exhausted. Borrowers survive it by maintaining 150%+ collateral buffer on Aave/Compound, monitoring health factor daily, and avoiding correlated collateral.

How a liquidation cascade actually starts

Every cascade begins with an external price shock. ETH falls 15% in an hour, or a stablecoin depegs by 8%, and the borrowers sitting at the highest loan-to-value ratios suddenly find their health factor dipping below 1.0. On Aave and Compound, anyone whose collateral value × liquidation_threshold falls below their debt value becomes a target. Health factor = collateral value × liquidation_threshold ÷ debt value — safe above 1, liquidatable below.

Liquidator bots — purpose-built MEV searchers — race to repay a slice of that debt in exchange for the borrower's collateral plus a liquidation bonus, typically 5–10% on Aave and varying by asset. The bots don't hold the collateral; they dump it on Uniswap or Curve within the same block to lock in the bonus as profit. That forced selling pushes the asset's price down further, putting the next tier of borrowers — those who were a little more conservatively positioned — into the danger zone. The cycle continues until either buyers step in to absorb the selling, or every over-leveraged position has been flushed.

The five phases of a cascade:

  1. External shock — a market-wide price drop, depeg, or stale oracle update hits.
  2. First-wave liquidations — highest-LTV positions cross under health factor 1.0.
  3. Liquidator bonus capture — bots repay debt, seize collateral + bonus, dump on DEX.
  4. Price impact propagates — DEX selling pushes the asset lower; oracles catch up.
  5. Second-wave liquidations — previously safe positions breach threshold; the loop closes.

Once the loop is running, the only question is how much damage thin liquidity will amplify.

The math that makes cascades accelerate

The mechanics above would be a minor cleanup if markets were infinitely deep. They aren't. The same $5M of forced selling produces very different damage depending on the liquidity waiting to absorb it. On a Uniswap pool with $20M of depth, a $5M market sell moves price 20–25%. On a pool with $200M of depth, the same sell moves price 2–3%. The protocol doesn't change; the venue does.

This is why cascades are non-linear. The first liquidations hit thin overnight liquidity, push price down 10%, and that 10% is enough to push the next safety tier underwater. Those liquidations are larger, hit slightly thicker (but still shaken) liquidity, push price down another 8%, and so on. The cascade is a multiplier, not a sum.

Collateral dump size Pool depth Approx price impact Next-tier positions affected
$1M$20M~5%Marginal
$5M$20M~20–25%High
$5M$200M~2–3%Low
$20M$200M~8–10%Moderate–High

Black Thursday, March 12, 2020, is the textbook case. ETH fell roughly 43% in 24 hours, MakerDAO's collateral auctions failed to find bids (the "0-bid auction" problem), and the protocol absorbed about $4.5M of bad debt that MKR holders ultimately had to recapitalize through minting. Even well-designed protocols accumulate bad debt during severe cascades — the question is who pays for it.

The lesson from the math: thin overnight liquidity, not protocol design, is what turns a 15% drop into a 40% cascade.

Real cascade events — what actually broke

Cascades aren't theoretical. Three of the largest DeFi blowups of the last cycle were cascade mechanics playing out on different collateral types:

Event Trigger asset Peak price drop Protocols affected Bad debt generated
Black Thursday (Mar 2020)ETH~43% in 24hMakerDAO~$4.5M
Terra collapse (May 2022)UST~70%+Anchor, Venus, Abracadabra~$400M+ (Anchor)
stETH depeg (Jun 2022)stETH~7% vs ETHCurve, leveraged stETH loopsLimited (peg recovered)
FTX collapse (Nov 2022)FTT~90%FTX, Alameda-exposedN/A (CeFi failure)

Notice the pattern: each cascade was triggered by a single asset class — ETH, UST, stETH, FTT — but spread through the protocols that had accepted it as collateral. The protocol didn't break first; the collateral did.

How to protect your positions

The strategies that survive cascades are boring and asymmetric. None of them eliminates risk — the goal is to be the last position standing when the cascade runs out of fuel.

Honest risk framing: no strategy eliminates cascade risk. Even well-capitalized positions can be hit if the cascade is severe enough and liquidity is thin enough. The goal is being the last position standing — not the one immune to the spiral.

Frequently asked questions

What is a liquidation cascade in DeFi?

A liquidation cascade is a self-reinforcing cycle where falling collateral prices trigger automated liquidations, whose forced selling pushes prices lower, triggering more liquidations. The cycle ends only when buyers step in to absorb the selling, or when over-leveraged positions are exhausted. It's distinct from a normal price drop because the selling pressure is mechanical, not discretionary.

How does a liquidator bot profit from my liquidation?

Liquidator bots monitor the chain for positions whose health factor has dropped below 1.0. They repay a portion of your debt in exchange for your collateral plus a liquidation bonus (typically 5–10% on Aave, varying by asset). The bot immediately sells the seized collateral on a DEX to lock in the bonus as profit. The bonus is the protocol's incentive for keeping the system solvent.

What's a safe collateral ratio on Aave or Compound?

The protocol-minimum health factor is 1.0, but a safe buffer is 1.5 or higher — equivalent to roughly 30% of headroom before liquidation triggers. Below 1.5, a single-day 20% collateral price drop can put you at risk. Above 1.7, you survive most historical cascades including Black Thursday and the stETH depeg without being touched.

Can a liquidation cascade drain my collateral completely?

On most modern protocols, no — liquidations are partial, not full. Aave and Compound liquidate up to a capped percentage of your position (close factor) per liquidation, leaving the rest. However, in a cascade the capped percentage can be hit repeatedly across consecutive liquidations as your health factor keeps dropping. The practical result is that an under-buffered position can lose 30–60% of its collateral during a fast cascade.

How is a liquidation cascade different from a normal market correction?

In a normal correction, sellers choose when to sell and the market absorbs it. In a cascade, the selling is mechanical — triggered by automated liquidator bots executing protocol rules, not human discretion. The selling pressure is also timed: bots dump seized collateral in the same block to lock in profit. This makes cascade selling concentrated, non-linear, and self-reinforcing in a way ordinary corrections are not.

Grade your portfolio's liquidation risk

Run your DeFi positions through the DifiCalc Yield Risk Grader to see your health factor, correlated-exposure heatmap, and the cascade scenarios that could reach you.

Open the Yield Risk Grader

Related reading: Stablecoin Depeg Risk, DeFi Yield Traps: Red Flags, Aave vs Compound, and the Aave protocol review. More guides in the DifiCalc blog.

Sources and further reading