Lesson 1 — The three structural designs and where each fails

**Fiat-backed stablecoins** are issued 1:1 against a reserve of fiat assets held by the issuer. USDC, USDT, USDP, and PYUSD all work this way. You give the issuer a dollar (sometimes literally, more often through an intermediary), the issuer holds that dollar (or an equivalent — typically a mix of cash, US Treasuries, and money-market instruments), and the issuer mints you one token. The token is a claim against the reserve, redeemable for the underlying asset. The peg holds because the redemption claim is real and the arbitrage path exists: if the token trades below $1, someone redeems at $1 from the issuer and pockets the difference. If it trades above $1, someone deposits fiat with the issuer to mint new tokens and sells them.

The load-bearing assumptions of fiat-backed designs are: (a) the reserves actually exist and match the circulating supply; (b) the assets in the reserve are themselves stable (not exposed to bank failure, not concentrated in long-duration bonds during a rate-hike cycle, not subject to operational freeze); and (c) the redemption mechanism actually works at scale. The failure modes follow directly. Reserve shortfall (the issuer hasn't held what it claimed) — this is the historic concern with Tether, where reserve composition has been a regulatory and journalistic focus for years. Asset-quality risk (the reserves themselves are unstable) — this is what happened to USDC on the SVB weekend, covered in Lesson 3. Redemption-channel breakage (the issuer can't process redemptions) — this is what risks any fiat-backed when its bank partners freeze.

**Crypto-collateralized stablecoins** are issued against deposits of *other* crypto, locked in smart contracts. DAI is the canonical example. To mint $100 of DAI, you deposit, say, $150 of ETH into a vault. The DAI is minted as a loan against that collateral. If your collateral's value drops below the protocol's required ratio, your position is liquidated automatically by the smart contract — the collateral is sold to repay the DAI, and you keep whatever is left. The peg holds because anyone who wants $1 can either redeem DAI by paying back the loan and reclaiming the collateral, or buy DAI on a market and use it the same way.

The load-bearing assumptions are: (a) the collateral can be reliably valued (requires honest oracles, which we cover in the DeFi Safety course); (b) the over-collateralization buffer is large enough to survive a fast price drop; (c) the liquidation mechanism actually clears under stress, including during the most volatile minutes when many positions are being liquidated simultaneously. Failure modes: oracle manipulation makes the collateral appear to be worth more than it is, the system mints too much DAI, and the peg breaks downward. Cascading liquidations during a crash can leave the protocol under-collateralized — this nearly happened to MakerDAO on Black Thursday (March 12, 2020), when ETH dropped 50 percent in 24 hours and liquidations failed to clear cleanly; the protocol ended up with bad debt and had to recapitalize.

**Algorithmic stablecoins** maintain their peg without holding fiat or crypto collateral of equivalent value to the supply. The most prominent example was TerraUSD (UST), which was paired with the volatile sister token LUNA. UST was minted by burning $1 worth of LUNA, and burned to mint $1 worth of LUNA. The peg theory: if UST trades at $0.98, you burn 0.98 UST to mint $1 of LUNA, which you sell for the profit — restoring the peg. If UST trades at $1.02, you burn $1 of LUNA to mint 1 UST, which you sell — restoring the peg.

The load-bearing assumption: LUNA's market cap always exceeds the value of circulating UST, and demand for LUNA is robust enough that mint-burn arbitrage works at scale. This assumption fails catastrophically when UST starts trading meaningfully below $1 because the arbitrage becomes self-reinforcing in the wrong direction. To restore the peg, more and more LUNA must be minted, which dilutes LUNA, which drops LUNA's price, which requires even more LUNA to be minted to absorb the same dollar of UST — a 'death spiral.' UST went from $1.00 to $0.01 in days in May 2022. The collapse was structural, not accidental: every algorithmic stablecoin with no external collateral that has ever been launched at scale has ultimately failed for the same reason. Iron Finance, USDR, USDA, NUSD — the list is long.

There is a fourth category sometimes claimed — 'hybrid' or 'partially-collateralized' — but these are best understood as crypto-collateralized stablecoins with a reduced collateral ratio, which means they sit at a specific point on the same spectrum, with proportionally elevated risk. FRAX's V1 design was the most prominent example; over time the protocol moved toward higher collateralization precisely because the partial-collateral model is structurally fragile in stressed markets.

When you see a new stablecoin announced, the first question is: which of these three families does it actually belong to? The branding may obscure the answer — 'algorithmic' is sometimes used loosely, 'asset-backed' is sometimes used for designs with no fiat reserves at all — but the mechanism is always there in the documentation if you read past the marketing. The family tells you the failure mode. The failure mode tells you what to monitor.