Traders treat stablecoins like USDT and USDC as cash, but they are not cash — they are liabilities of a private issuer, and the quality of that liability depends entirely on how it's backed. Understanding the three main collateral models is essential risk management, not academic trivia.
Three Collateral Models
| Model | Example | Key risk |
|---|---|---|
| Fiat-collateralized | USDT, USDC | Reserve quality and transparency; banking counterparty risk |
| Crypto-collateralized (over-collateralized) | DAI | Collateral value can crash faster than liquidations can process |
| Algorithmic (uncollateralized/undercollateralized) | UST (collapsed 2022) | Death spiral: loss of confidence breaks the peg mechanism itself |
The UST Collapse: A Case Study
In May 2022, TerraUSD (UST) — an algorithmic stablecoin that maintained its peg via a mint-and-burn arbitrage with its sister token LUNA — lost its peg after large withdrawals from the Anchor lending protocol triggered selling pressure. The arbitrage mechanism, designed to restore the peg, instead hyperinflated LUNA's supply as it was minted to absorb UST sell pressure, destroying roughly $40 billion in combined value within days.
Unlike collateralized stablecoins, an algorithmic peg has no hard asset backstop. Once holders doubt the mechanism will hold, the resulting sell pressure can become self-fulfilling and irreversible within hours.
Assessing Fiat-Backed Stablecoin Risk
- ▸Check reserve composition — cash and short-term T-bills are far safer than commercial paper or corporate debt.
- ▸Attestations are not audits — they confirm a snapshot, not ongoing solvency or asset quality over time.
- ▸Redemption access matters — a stablecoin only institutional clients can redeem 1:1 trades at a slight discount on secondary markets during stress.
Practical Risk Management
Diversify stablecoin holdings across issuers if you park significant capital in them, monitor on-chain reserve dashboards where available, and treat any stablecoin yield above the risk-free rate as compensation for real risk, not free money.