Iron Finance’s TITAN collapse exposed a fatal stablecoin feedback loop

On June 17, 2021, Iron Finance published a postmortem after its Polygon-based IRON stablecoin broke its intended $1 peg and the linked TITAN token collapsed. The episode mattered because it showed how a token marketed as stable could transmit a run through its own redemption machinery even without a conventional custodian failing or an outside attacker being identified.

The project called the event the first large-scale crypto bank run. That description was Iron Finance’s contemporaneous characterization, not an independently established industry first. What the public record did establish was a rapid failure of the two-token design: selling pressure hit TITAN, IRON traded below its target, redemptions created still more TITAN, and that newly issued TITAN was sold into a falling market.

How IRON was supposed to hold one dollar

IRON was not fully backed by dollars or dollar-equivalent reserves. Its design combined USDC collateral with TITAN, the protocol’s volatile share token. At the effective collateral ratio around the run, redeeming one IRON was intended to return roughly $0.75 of USDC plus about $0.25 worth of TITAN. The TITAN amount was calculated from an oracle price rather than guaranteed in dollar terms.

That distinction was central. A conventional redemption promise depends on the reserve asset remaining available. IRON’s promise also depended on a market price for TITAN and on enough liquidity to sell it. When TITAN fell quickly, a lagging price input could value freshly minted TITAN above what buyers would pay in the spot market.

The feedback loop

Iron Finance’s June 17 account said large holders began removing liquidity from the IRON/USDC pool at about 10:00 UTC on June 16, then sold TITAN and IRON through liquidity pools rather than redeeming IRON. The team said the first loss of the peg briefly corrected, but another wave of selling pushed TITAN down and IRON below $1 again.

The crucial mechanism followed. Holders could buy discounted IRON and redeem it for USDC plus newly minted TITAN. Because TITAN’s calculated redemption price lagged its collapsing spot price, each redemption could require the protocol to mint an expanding number of TITAN tokens. Redeemers then sold those tokens, adding pressure to the same market used to price the non-USDC portion of IRON. The process turned an attempted arbitrage back toward the peg into a reflexive dilution loop.

CoinDesk reported on June 17 that TITAN had fallen from roughly $65 to near zero and that redemptions, automatically halted during the disruption, were scheduled to resume at 17:00 UTC. The price description was a contemporaneous market report, not an official closing price. Crypto traded continuously across decentralized pools, and thin liquidity, slippage and different observation times made a single definitive daily close inappropriate.

Why the failure mattered

The collapse separated stablecoin branding from reserve quality. IRON holders were exposed not only to USDC but also to TITAN’s market depth, oracle behavior and the protocol’s mint-and-redeem rules. The event also demonstrated that smart-contract execution can operate as coded while producing an economically unstable result. No verified exploit was necessary for the design to fail under concentrated selling and redemptions.

For Polygon’s growing decentralized-finance market, the incident was an immediate stress test of liquidity incentives and partially collateralized tokens. It also supplied regulators and risk managers with a concrete example of why the word stablecoin covered materially different structures: cash-backed claims, overcollateralized crypto loans and reflexive algorithmic designs did not carry the same failure modes.

Later analytical context

A Federal Reserve staff note published on June 2, 2022, later reconstructed Polygon transactions and found that TITAN moved from about $60 to effectively zero within hours on June 16, while IRON fell below $0.75 before stabilizing around the USDC-backed portion. That later study supports the mechanism described above, but it was not available on June 17, 2021, and should not be read as part of the event-day information set.