Stablecoins are not one thing

Every stablecoin displays the same number. That shared price is the source of an enormous amount of confusion, because it makes instruments with completely different risk structures look interchangeable on a portfolio screen. They are not interchangeable, and the differences are invisible right up until the moment they are the only thing that matters.

Three broad designs dominate, and each fails in its own characteristic way.

Fiat-backed: you are holding a claim on a company

The most common design holds reserves — cash, short-term government debt, sometimes other assets — and issues tokens against them. The peg is maintained by the promise that tokens can be redeemed for the underlying at par.

The risk here is not really cryptographic. It is ordinary counterparty and banking risk wearing an unfamiliar interface. What are the reserves actually composed of? Who has audited them, how recently, and was it an audit or the weaker “attestation”? Which banks hold the cash, and what happens if one of them fails? Who can redeem directly — everyone, or only large institutional partners, with everyone else dependent on secondary markets?

That last question is more consequential than it appears. If direct redemption is restricted to a small set of counterparties, then for most holders the peg is maintained by arbitrage rather than by a right they personally hold. Arbitrage works well in calm conditions and less well when the arbitrageurs themselves are constrained.

Overcollateralised: transparent, but reflexive

The second design locks volatile crypto assets in contracts and issues stablecoins against them at a ratio well above one to one. The reserves are visible on-chain, which removes the trust-us problem almost entirely — anyone can verify the collateral exists.

What replaces it is reflexivity. The collateral is volatile, so a sharp fall in crypto prices reduces the backing at the same time as it triggers liquidations, and those liquidations sell into the falling market. The system is designed for this and generally handles ordinary volatility well. The stress case is a fast, deep move where liquidations cannot clear at expected prices and the network is congested precisely because everyone is transacting at once.

There is a second wrinkle worth knowing: many overcollateralised systems now hold a meaningful share of their backing in fiat-backed stablecoins, which quietly reimports the counterparty risk they were designed to avoid.

Algorithmic: the design that keeps failing

The third design maintains the peg through mechanism rather than assets — minting and burning a companion token, or offering yields that attract capital to defend the price. The appeal is obvious, since it requires no reserves and scales without a balance sheet.

The recurring failure is a confidence loop. The mechanism depends on people being willing to hold the companion asset, and that willingness depends on the peg holding. When the peg slips far enough, the mechanism issues more of an asset nobody wants, which accelerates the slide it was supposed to arrest. This has now happened at multiple scales, including one collapse large enough to take a substantial portion of the market down with it.

Newer designs are more careful, often backing the mechanism with real reserves or hedged positions. Those are meaningfully different from the pure version, but the label rarely distinguishes them, and a design that is partly mechanical still inherits the reflexive failure mode in proportion to how mechanical it is.

Why this matters for a holder

Someone who moves to stablecoins during a drawdown has not moved to cash. They have moved to a specific instrument with a specific issuer, a specific reserve composition, and a specific redemption mechanism — and quite possibly to a chain where a bridge sits between them and the thing they think they hold.

The practical steps are unglamorous. Know which design you are holding. Read the most recent reserve disclosure and check its date. Know whether you personally can redeem or are relying on someone else’s arbitrage. Consider splitting across issuers, since the failure modes are largely independent. And treat a peg that has been perfect for years as evidence about calm conditions only — every stablecoin that has ever broken had an unblemished record right up until it did not.