MAI: the bridge that broke the peg

MAI is a CDP stablecoin issued by QiDao: a borrower locks crypto collateral in a vault and mints MAI against it, overcollateralized, and recovers the collateral by repaying. The system is supposed to be fully backed because every MAI in circulation traces to a vault holding more value than it minted. That invariant held chain by chain — as long as the collateral in each chain's vaults remained worth what the oracle said.

On July 6, 2023, Multichain's Fantom bridge was drained of roughly $125M. The bridged assets that sat as collateral inside QiDao's Fantom vaults were suddenly worthless, but the MAI those vaults had minted did not evaporate with them — it had already been bridged to Polygon and other chains and was circulating freely. The result was that the protocol now had more MAI outstanding than collateral backing it: bad debt, and crucially, bad debt that lived on chains other than the one where the collateral failed.

QiDao reacted fast, shutting the Multichain route and isolating Fantom to stop the bleed. But isolation came after the tokens had already escaped. The undercollateralization translated directly into price: MAI slipped a few percent in July, and as the market absorbed that the shortfall was structural rather than transient, the deviation widened through the autumn to a low near $0.72 in mid-October.

The instructive part of this case is the causal chain. MAI did not depeg because crypto collateral prices fell — the usual way a CDP stablecoin breaks. It depegged because a bridge failed. The collateral inside the Fantom vaults was destroyed not by a market move but by the disappearance of the bridge that held the assets backing those collateral tokens. The CDP mechanism behaved correctly given its inputs; the inputs were corrupted by infrastructure QiDao did not run.

This is the distinction worth holding against the wrapped-asset bridge story. When Multichain collapsed, the Fantom-bridged USDC.m — a wrapper that was nothing but a claim on the bridge — fell to roughly $0.22 and died, because it had no backing of its own once the bridge emptied. MAI is a different animal: a sovereign CDP stablecoin with vaults and collateral on many chains. The bridge failure did not zero it. It punched a hole in one chain's collateral and let the multichain footprint carry the damage outward, producing a sustained sub-peg rather than a wrapper's total collapse.

The lesson is that for a multichain CDP coin, every bridge it relies on to keep collateral and liquidity coherent is part of its risk surface, indistinguishable in effect from a collateral oracle. A bridge that strands or destroys backing on one chain mints bad debt the same way a bad price feed would — and the broader the deployment, the more places that bad debt can hide and the longer it takes to find and burn.

Bad debt confined to one chain is a contained accounting problem; bad debt smeared across a dozen is a slow-motion one. The MAI minted on Fantom had been bridged elsewhere, so the chains that imported it were holding tokens with no collateral standing behind them, and there was no single ledger where the shortfall sat to be cleared in one motion. QiDao had to find the unbacked supply and retire it chain by chain.

A bank-run dynamic made it worse before it got better. On Fantom, where nearly everything else had become worthless, MAI was briefly the only liquid escape — so holders drained the remaining MAI pools to sell the tokens on Polygon and elsewhere, exporting the sell pressure to exactly the chains that still had functioning markets. The deviation propagated along the same rails the protocol had used to grow.

Clearing it was a treasury exercise stretched over months. QiDao bought MAI back and burned it, funded by a treasury holding over $7M in crypto, shrinking the circulating supply on the affected chains until outstanding MAI again matched real collateral. The Polygon supply fell sharply over the period, and the protocol added a peg-stability module in January 2024 to firm up redemption. The final tranche — roughly $668K of bad debt on Polygon — was settled in February 2024, more than half a year after the bridge failed.

A sustained sub-peg is not only a problem for holders; it is a problem for every lending market that listed the coin as collateral or as a borrowable asset. MAI had been integrated into Aave v3 across several chains. As the deviation persisted and deepened, the risk managers who parameterize those markets moved to limit the protocol's exposure rather than wait for a recovery that kept not arriving.

On October 13, 2023, with MAI down near $0.72 and unable to hold peg for months, Gauntlet formally recommended that Aave begin deprecating the MAI/MIMATIC listings on Polygon, Avalanche, Arbitrum, and Optimism. The mechanism was the standard wind-down: drive loan-to-value toward zero and raise borrow rates so positions are pushed to repay and the market can be retired, an estimated ~$70K in forced liquidations to unwind. Earlier in the year, after the first July wobble, risk managers had already tightened MAI's supply and borrow caps; the October step escalated containment to deprecation.

Deprecation pressure is its own kind of scar. Even after a coin claws back toward par, the integrations it lost do not automatically return — a market that was wound down and a risk team that flagged the asset both leave a coin smaller and less trusted than it was. That is much of what 'wounded' means here: the peg recovered, but the standing did not fully.

First, a CDP stablecoin's collateral risk is not only price risk. Anything that can corrupt the value of vault collateral — a bridge that holds the underlying assets, a wrapper that can be drained, an oracle that can be manipulated — is a depeg vector, even when crypto prices never move against the borrower. For MAI, the failure was an infrastructure dependency, not a market one, and a collateral-price lens alone would have missed it entirely.

Second, a multichain deployment is a double-edged footprint. The same breadth that grows a stablecoin spreads its bad debt: tokens minted against failed collateral on one chain were already loose on others, so the shortfall could not be quarantined and the cleanup ran chain by chain over months. Broad deployment widens both the addressable market and the surface where damage can hide.

Third, mind the difference between dying and being wounded. The wrapped USDC.m on the same bridge had no backing of its own and went to roughly $0.22 and zero recovery; MAI, a sovereign CDP coin, took a sustained sub-peg and a wave of deprecation but bought back its bad debt and recovered toward par. Survival is the better outcome — but the lost integrations, the months below peg, and the fragmented chain-by-chain liquidity are the durable cost of depending on infrastructure you do not control.