The Mango Squeeze
How $5 million in USDC became $110 million in withdrawals
Some DeFi blowups need a long postmortem. Mango Markets is almost rude in how easy it is to understand once you see the trick: get Mango to believe your position is worth a fortune. Borrow everything against it before it notices the number was nonsense.
On October 11, 2022, Avraham Eisenberg put about $5 million of USDC into Mango. A few minutes later, he had borrowed and withdrawn more than $100 million of the platform's crypto. The SEC put the figure at roughly $116 million. The criminal case used roughly $110 million. The exact number is less interesting than the shape of it.
Mango let a temporary price become permission to withdraw real assets.
The setup
Mango was a Solana-native cross-margin trading and lending protocol. Users could trade spot markets and perpetuals, deposit assets as collateral, and borrow from the same account. That second part is normal enough. Put up assets, borrow against them, withdraw what you borrow. The whole arrangement works as long as the thing you put up can actually be sold for roughly the value the protocol gave it.
Mango's own token was called MNGO. It was thinly traded, which meant a large buyer could move its price a lot. Mango also had a perpetual market on MNGO. The platform counted those positions toward a user's collateral.
This is where the setup became dangerous. Mango did not make up a MNGO price itself. It read an average from three outside exchanges: FTX, AscendEX, and Serum. That sounds safer than trusting one exchange. It is not much safer if somebody can move all three at once.
Eisenberg used two Mango accounts. One was positioned to win if MNGO went up. The other was positioned to win if it went down. That meant he could borrow after pushing MNGO up, then borrow again after pushing it back down.
The trade
First, he bought MNGO across the exchanges Mango watched. The price went up. Mango saw the higher price, marked the long position higher, and treated that paper gain as something he could borrow against.
That was enough. He borrowed against the long position and withdrew real assets from Mango.
Then he sold MNGO back into those same markets. The price fell. Now the short position looked valuable, so he borrowed more against that one too. The court's description of the sequence is unusually plain: deposit the USDC, take opposite positions, push the oracle price up, borrow, push it down, borrow again.
The squeeze was not about keeping MNGO expensive forever. It only had to be expensive at the exact moment Mango asked how much his account was worth.
The visual below compresses that idea into one direction. Keep pumping the price until the withdrawal line opens. The real trade was uglier because it worked in both directions.
The Mango squeeze
Push the MNGO market price up. Mango then values your position higher, until its assets unlock.
Mango Markets
Your account
MNGO market
$0.04
Value of your MNGO position
Withdraw hereMango's assets
Your wallet
Waiting
MNGO starts at $0.04.
What broke
It is tempting to say Mango had an oracle problem. It did, but that is only half the story.
The bigger mistake was letting a mark on a thin market become a withdrawal right immediately. A price is not the same thing as exit liquidity. If a position says it is worth $100 million, the important question is not whether a feed printed that number for a second. It is whether someone could sell it for anything close to that while trying to take $100 million out of the protocol.
Mango's answer was effectively yes. The smart contract did what it had been told to do. Its collateral number rose, so its borrowing limit rose. It did not stop to ask whether the price was being manufactured by the person about to drain it.
That is the uncomfortable bit about DeFi. The code can behave perfectly and still be attached to a terrible economic assumption.
The aftermath
The aftermath became its own mess. Eisenberg described the trade as a legal, highly profitable strategy. U.S. authorities disagreed. A jury convicted him in 2024 of commodities fraud, commodities manipulation, and wire fraud. The Justice Department called it its first cryptocurrency open-market manipulation case.
Then, in May 2025, a federal judge vacated the two commodities convictions because the government had not proved venue in New York and entered a judgment of acquittal on the wire-fraud count. That ruling was not a happy ending for Mango, or a declaration that the trade had been harmless. The order itself says the venue reversal did not resolve the underlying question of culpability.
The legal argument was always going to be interesting. The protocol had no person at a desk approving a loan. It had an automated rule reading an automated price. But the users whose assets left Mango did not experience an interesting legal question. They experienced an exchange with no money left in it.
What it left behind
Mango is a good reminder that the most dangerous part of a money market is sometimes the one number nobody thinks to challenge.
You can have an oracle. You can have collateral ratios. You can have liquidation code and a very serious-looking dashboard. None of it helps if an attacker can make the thing you accept as collateral briefly look valuable enough to open the vault.
For a few minutes, MNGO was very expensive.