Several stablecoins have failed. The record matters because the mechanics of each failure are distinct and the market's response — in product design and regulation — followed directly from what each collapse revealed. What follows is a factual account of the three most significant events, followed by the concrete changes they drove.
TerraUSD (UST) — May 2022
Loss: approximately $40 billion in UST/LUNA market capitalisation
TerraUSD was an algorithmic stablecoin issued on the Terra blockchain by Terraform Labs, led by Do Kwon. Its dollar peg was maintained not by held reserves but by an arbitrage loop with a sister token, LUNA. When UST traded below $1, users could burn UST to mint LUNA, earning the difference; when it traded above $1, they could burn LUNA to mint UST. The mechanism depended on LUNA retaining independent market value.
In early May 2022, a large coordinated withdrawal from Anchor Protocol — Terra's lending platform, which was offering 19–20% yields on UST deposits — began. Two large entities converted over $185 million of UST into other stablecoins in rapid succession, triggering a liquidity cascade. UST slipped below $1. The arbitrage loop activated: users burned UST to mint LUNA, flooding the market with new LUNA supply, which drove LUNA's price down, which further undermined the loop's economics. Within days, UST had fallen to near zero and LUNA from approximately $80 to fractions of a cent.
The Federal Reserve documented the contagion effects: the collapse contributed to the failure of crypto hedge fund Three Arrows Capital, lender Celsius, and broker Voyager Digital over the following months, spreading losses well beyond direct UST holders.
Do Kwon fled to Serbia after arrest warrants were issued, was extradited to the United States, and in 2025 pleaded guilty to fraud charges. He was sentenced to 15 years in federal prison.
What it proved: An algorithmic peg mechanism has no external capital to absorb a confidence shock. When the mechanism is tested at scale, the confidence that sustains it is precisely what vanishes. The architecture is inherently reflexive — a falling price makes the mechanism perform worse, which further reduces confidence, which further reduces the price.
Iron Finance / IRON-TITAN — June 2021
Loss: approximately $2 billion
Iron Finance operated on Polygon with a partially-collateralised model. IRON was backed partly by USDC (a real reserve asset) and partly by TITAN, the protocol's own governance token. This hybrid model was capital-efficient: it required less than $1 of external collateral per $1 of IRON.
On June 16, 2021, a wave of large withdrawals began, possibly triggered by whale positions reducing exposure. As TITAN's price fell under redemption pressure, the effective collateral backing each IRON fell below face value, triggering further redemptions, further price declines in TITAN, and further collateral deterioration — the same death spiral as UST but on a smaller scale. IRON fell from $1 to under $0.75 within hours. TITAN went to near zero.
The Federal Reserve cited the Iron Finance case alongside UST in its June 2022 analysis of algorithmic stablecoin runs, noting that the run dynamics were structurally similar to traditional bank runs: rational individual behaviour (withdrawing) produced collectively destructive outcomes.
Investor Mark Cuban, who had publicly invested in the ecosystem, disclosed losses and subsequently called for regulation of stablecoins.
What it proved: Hybrid designs that use a protocol's own token as partial collateral are subject to the same reflexive collapse as pure algorithmic designs. The USDC portion of the backing was safe; the TITAN portion went to zero, destroying the peg.
Basis Cash — late 2020 to early 2021
Loss: market cap peaked around $30 million before collapse
Basis Cash was a three-token algorithmic stablecoin system — BAC (the stablecoin), Basis Bonds, and Basis Shares — inspired by the Basis project that had shut down in 2018 after regulatory concerns. BAC launched in November 2020 with an algorithmic seigniorage model: Bonds were sold at a discount when BAC fell below $1, to be redeemed later at par; Shares captured protocol revenue when BAC exceeded $1.
BAC never regained sustained dollar parity after falling below $1 in January 2021. The bond redemption mechanism required BAC to trade above $1 to be effective; when it didn't, bondholders could not be repaid, removing the incentive to hold bonds, which removed the stabilising mechanism. BAC settled permanently around $0.03. While the dollar scale of losses was small compared to UST, the mechanism's failure was a clear precursor and a documented warning that the market did not heed at scale.
USDC / Silicon Valley Bank — March 2023
Loss: zero (full recovery); brief depeg to $0.87
USDC is fiat-backed, not algorithmic. Its near-failure in March 2023 was not a design failure but a custodian failure — and the resolution demonstrated both the strength and the limits of reserve-backed designs.
On March 10, 2023, US regulators closed Silicon Valley Bank. Circle disclosed that $3.3 billion of USDC's approximately $40 billion in reserves — roughly 8% — was held in SVB cash deposits and potentially inaccessible. The disclosure, made late on a Friday, triggered immediate redemption demand on secondary markets. USDC fell to $0.87 by Saturday morning.
Over the weekend, US federal regulators announced that all SVB depositors — including institutional deposits above FDIC limits — would be made whole. USDC recovered its peg by Monday, March 13. The underlying token never incurred a loss; only secondary market prices depressed temporarily.
What it proved: Even a well-managed, transparent fiat-backed stablecoin carries custodian risk. The resolution required federal intervention. Without that guarantee, a run dynamic was already underway. The event reinforced the case for reserve diversification across multiple custodians and accelerated Circle's move toward holding the bulk of reserves in a government money market fund rather than bank deposits.
What changed after these failures
| Failure | Regulatory response | Industry response |
|---|---|---|
| TerraUSD (2022) | GENIUS Act (2025) prohibits algorithmic designs for payment stablecoins; MiCA (2024) requires 1:1 liquid backing | Market effectively abandoned new large-scale algorithmic stablecoin projects |
| Iron Finance (2021) | Cited in Fed research; contributed to broader algorithmic concern | Hybrid partial-reserve designs largely abandoned |
| USDC/SVB (2023) | GENIUS Act requires reserve diversification and monthly disclosure | Circle shifted to predominantly money market fund reserves; industry increased disclosure frequency |
The US GENIUS Act, signed July 18, 2025, codified the clearest lesson: a dollar-pegged payment stablecoin must be backed 1:1 by cash and short-term US Treasuries. Algorithmic designs — regardless of complexity or claimed stability mechanisms — are excluded from the payment stablecoin category. The EU's MiCA regulation reached the same conclusion through a parallel path, with stablecoin provisions applying from June 2024.
The stablecoin market that emerged from these failures is more concentrated around fiat-backed designs from regulated issuers, and more oriented toward transparency — monthly attestations by recognised accounting firms, disclosed reserve composition, and legal segregation of reserves from issuer assets.
For current risk analysis by design type, see stablecoin risks explained. For the full chronological arc from 2014, see the history of stablecoins.