A stablecoin promises one thing (a stable unit value, usually $1 or its home-currency equivalent) and there are three structurally different ways to try to deliver it, and they do not carry the same risk.
Fiat-collateralized stablecoins (Tether's USDT, Circle's USDC) hold reserves (cash, short-dated government securities, and similar instruments) roughly matching the tokens in circulation, redeemable (for eligible holders) for the underlying currency. The primary-source discipline here matters precisely: both issuers publish periodic attestations of reserve composition (Tether's transparency page; Circle's monthly attestation reports), and an attestation is a real, useful disclosure, but it is not the same assurance level as a full financial-statement audit; know the difference before citing either as proof of anything beyond what it actually certifies.
How does a crypto-collateralised stablecoin work?
Crypto-collateralized stablecoins (MakerDAO's DAI, now under the Sky Protocol brand) are backed by over-collateralized crypto deposits locked in a smart contract, per the mechanism described in Rune Christensen's 2017 Dai whitepaper. Work the numbers once, because they show exactly what "over-collateralized" buys you and where it stops buying protection. Deposit 10 ETH at $3,000 = $30,000 of collateral and mint 15,000 DAI against it, an initial collateralization ratio of 200%, comfortably above a historically typical 150% minimum (liquidation) ratio for Maker's standard ETH vault type (governance-set, and periodically adjusted, check the current parameter before relying on it). The liquidation price (the ETH price at which the position becomes eligible for forced liquidation) is (debt × liquidation ratio) ÷ collateral = (15,000 × 1.50) ÷ 10 = $2,250, a 25% drop from the entry price. At liquidation, a historically typical 13% liquidation penalty (also governance-set) applies: collateral worth $22,500 at that price must cover the $15,000 debt plus a $1,950 penalty ($16,950 total), leaving the vault owner roughly $5,550 (about 18.5% of the original collateral value) before any further auction slippage. Push the stress further and the arithmetic flips sign: a price that gaps a further 10% below the liquidation price ($2,025) still leaves a $3,300 surplus over the debt-plus-penalty (the buffer holds) but a gap of 30% below the liquidation price ($1,575) leaves collateral worth only $15,750 against a $16,950 obligation, a $1,200 shortfall: real, uncollateralized bad debt.
Advertisement
What happened to DAI on Black Thursday?
That shortfall branch is not hypothetical, it happened, once, in public, and the case is worth knowing by name. On 12–13 March 2020 ("Black Thursday"), ETH's price fell roughly 50% in about 24 hours; Ethereum network congestion drove gas prices so high that competing liquidation "keepers" dropped out, letting roughly $8.3 million of collateral get auctioned off for zero DAI to a single keeper who bid zero and faced no competition, and the protocol's own post-mortem and subsequent MKR debt auction record a net shortfall of roughly $4 million that the system had to cover by minting and selling new MKR tokens to make depositors whole (source: MakerDAO's own governance-forum post-mortem, and independent accounts including Messari's "The Legacy of MakerDAO's Failed Collateral Auctions"). The lesson is not "DeFi collateral mechanisms are fake" (the mechanism is real, published, and mostly worked as designed even that day) the lesson is that "over-collateralized" is a statement about normal conditions, not a guarantee about auction mechanics under network stress, and a serious analyst reads the stress-scenario fine print, not just the headline ratio.
How did Terra's algorithmic stablecoin collapse?
Algorithmic stablecoins carry no collateral at all in the fiat/crypto sense, they hold their peg through a mint-and-burn arbitrage mechanism between two tokens, and the canonical failure is worth knowing in real, sourced detail rather than as vague crypto-Twitter lore. Terra's UST, per Do Kwon and Nicholas Platias's 2019 whitepaper Terra Money: Stability and Adoption, held its dollar peg by letting arbitrageurs burn $1 of UST for $1 of the sister token LUNA (and vice versa), a design with no real assets behind it at all, propped up further by Terra's Anchor Protocol, which paid depositors an advertised, unsustainably high ≈19.5% yield on UST deposits, funded by a subsidy pool rather than by any organic lending spread (per the NBER working paper Anatomy of a Run: The Terra Luna Crash, Liu et al., 2023). When large holders began redeeming UST starting 7 May 2022, the arbitrage mechanism that was supposed to defend the peg instead amplified the collapse, minting ever more LUNA to absorb the redemptions crashed LUNA's price, which broke the arbitrage's own economics, in a self-reinforcing "death spiral." By 13 May 2022 the system had largely imploded, wiping out, per that same NBER paper, roughly $50 billion in valuation, with the paper's own finding that wealthier, more sophisticated holders ran first and lost proportionally less, while less sophisticated holders ran later and absorbed larger losses. Hold the comparison across all three designs firmly: a fiat-collateralized stablecoin's risk is counterparty and reserve-quality risk; a crypto-collateralized stablecoin's risk is collateral-value and auction-mechanics risk (Black Thursday); an algorithmic stablecoin's risk is that the "backing" is a second volatile token and a arbitrage story, which is not backing at all under stress, it is a bet that the story keeps being believed, and Terra is the primary-sourced proof of what happens when it stops being believed.