Aave Releases Post-Mortem on Major Swap Incident of User Losing $50M to Swap

Aave has released an official post-mortem on a major swap incident from March 12, 2026, where a user lost approximately $50 million specifically around $50.4 million while attempting to swap aEthUSDT (wrapped USDT) for aEthAAVE via the CoW Swap router integrated into the Aave interface (aave.com).

What Happened

The user tried to exchange roughly $50.43 million worth of aEthUSDT for AAVE tokens. Due to extremely low liquidity in the relevant markets and routing issues, the trade executed with a massive ~99.9% price impact. The user ended up receiving only about 324–327 aEthAAVE tokens, valued at roughly $36,000–$36,500 at the time—a near-total loss of over $50 million.

This wasn’t a hack or exploit of the Aave core lending protocol which remained unaffected. Instead, it stemmed from: Executing a very large order in an illiquid market/pool. Routing through CoW Swap (a third-party aggregator), which provided a poor quote under the circumstances.

A sandwich attack by an MEV bot that captured around $10 million in profit from the trade. The user explicitly confirming multiple high-risk warnings in the interface including a checkbox acknowledging potential extreme slippage or total loss. Aave emphasized that the interface displayed strong warnings, and the trade required user confirmation.

CoW Swap published its own separate post-mortem, highlighting infrastructure and auction failures that compounded the poor execution. Aave plans to contact the affected user and refund approximately $110,000–$600,000 in fees collected from the transaction via the 0.25% swap fee on the interface, pending verification.

In direct response, Aave is deploying a new protective feature called Aave Shield. This will: Automatically block any token swaps on the Aave interface (aave.com) that would result in a price impact greater than 25%. Act as a default high-friction guardrail to prevent similar catastrophic executions in low-liquidity scenarios.

Allow advanced users to disable it manually in settings if they accept the higher risk. The goal is to enhance user protections without altering the underlying permissionless nature of DeFi, while reducing the chance of users accidentally or carelessly approving ruinous trades. This incident has sparked broader discussions on DeFi UX, liquidity fragmentation, MEV risks, and the need for better frontend safeguards—especially for large trades.

Interestingly, despite the negative event, AAVE token price has seen some upward movement in reports, possibly tied to perceived proactive response or broader market factors.

MEV sandwich attacks are one of the most common and notorious forms of Maximal Extractable Value (MEV) exploitation in DeFi, particularly on automated market makers (AMMs) like Uniswap, or when trades route through aggregators.

MEV refers to the additional profit that block producers (miners in pre-merge Ethereum, validators post-merge), searchers, or bots can extract by reordering, including, including, or censoring transactions within a block they control or influence. A sandwich attack specifically targets a user’s large swap (often on a DEX) by “sandwiching” their transaction between two of the attacker’s own transactions.

This manipulates the price in the liquidity pool to the attacker’s advantage, forcing the victim to get a worse execution price while the attacker pockets the difference. A user broadcasts a transaction to swap a significant amount of Token A for Token B; swapping millions in USDT for another token on a low-liquidity pool.

This pending tx sits in the public mempool visible to everyone, including MEV bots. Attacker detects the opportunity. Sophisticated bots constantly scan the mempool for large trades that will cause meaningful price impact / slippage due to the AMM’s constant product formula (x * y = k).

The attacker submits their own buy transaction just before the victim’s tx. If the victim is buying Token B, the attacker buys Token B first ? this pushes the price of Token B up in the pool. The victim’s swap now executes against a worse price for Token B, receiving fewer tokens than expected.

The user’s trade goes through at the now-inflated price ? they suffer extra slippage and get rekt (worse rate). Immediately after the victim’s tx, the attacker submits a sell transaction. They sell the Token B they just bought back into the pool at the now-higher price created by the victim’s large buy.

This captures the profit from the temporary price spike. The three transactions end up in the same block in this order: The attacker risks almost no capital often using flash loans for zero-risk execution and extracts profit purely from the victim’s slippage. Imagine a low-liquidity Uniswap pool with USDC-TOKEN: Without attack: Your $1M USDC buy might get you 10,000 TOKEN at an average ~$100 each.

In extreme cases like very illiquid pairs or huge orders, victims can lose massive portions of value — as seen in incidents where users lost tens of millions due to near-total slippage amplified by sandwiches. Transactions in the mempool are public ? bots see them instantly.

Validators / builders can reorder txs within a block post-Merge via PBS — proposer-builder separation — this has evolved but sandwiches persist. AMMs are deterministic and permissionless ? predictable price impact from order size. Trade via CoW Swap or other intent-based / batch-auction protocols.

Set tight slippage tolerance. Avoid huge trades in low-liquidity pools. Newer frontend features like the Aave Shield’s 25% price impact block add guardrails. Some chains or encrypted mempools reduce visibility. Sandwich attacks remain one of the biggest “taxes” on regular DeFi users — generating millions in weekly profits for searchers — but awareness, better routing, and protocol-level fixes continue to chip away at their dominance.