Algorithmic vs Fiat-Backed Stablecoins: Understanding the Risks

The Taxonomy of Stablecoin Designs

Stablecoins are broadly categorised by the mechanism they use to maintain a stable value. The most widely adopted design is the fiat-backed model: for every token in circulation, the issuer holds one unit of the reference fiat currency — or a portfolio of assets denominated in that currency — in reserve. This is the model used by Tether (USDT) and USD Coin (USDC), which together account for the large majority of the stablecoin market by market capitalisation.

A second design is the crypto-collateralised model, in which the peg is maintained not by fiat currency in a bank account but by other crypto assets locked in smart contracts. Because crypto assets are themselves volatile, these systems typically require significant over-collateralisation — a user might deposit $150 of ether to mint $100 of a stablecoin, providing a buffer against price declines in the collateral. DAI, issued by the MakerDAO protocol, is the longest-running example of this approach, though it has evolved considerably over its history and now incorporates real-world asset collateral alongside crypto.

The third design is the algorithmic model, in which there is no external collateral at all. Instead, the peg is maintained through algorithmic mechanisms — typically a combination of token supply expansion and contraction managed by smart contracts, sometimes supported by a companion token that absorbs volatility. It is in this category that the most spectacular failures in stablecoin history have occurred.

How Algorithmic Designs Attempt to Work

The core concept behind algorithmic stablecoins is that supply can be managed automatically to maintain price stability, in the way that a central bank manages money supply to maintain the value of a currency. If the stablecoin trades above its peg, the protocol mints new tokens to increase supply and reduce the price. If it trades below its peg, the protocol contracts supply through mechanisms designed to reduce the number of tokens in circulation and push the price back up.

The contraction mechanism is where designs have most commonly failed. Most algorithmic systems have relied on a companion token — sometimes called a governance token or a reserve token — to absorb the volatility. When the stablecoin falls below its peg, holders are offered the ability to burn the stablecoin in exchange for newly minted companion tokens at a rate that should make the exchange economically attractive, thereby reducing stablecoin supply. The companion token acts as a shock absorber: it falls in price when the stablecoin is under stress, accepting the volatility that the stablecoin is designed to avoid.

The fundamental vulnerability of this design is that the shock absorber can run out of capacity. If confidence in the stablecoin falls sharply, users rush to exit, the contraction mechanism requires minting large quantities of the companion token, which dilutes its value, which further reduces confidence in the entire system, which triggers more exits. This reflexive dynamic — in which the mechanism designed to restore the peg accelerates its breakdown — is what economists and crypto analysts now commonly call a death spiral.

The TerraUSD Collapse: What Happened

The May 2022 collapse of TerraUSD (UST), an algorithmic stablecoin issued on the Terra blockchain, and its companion token LUNA is the defining case study in algorithmic stablecoin risk. At its peak, UST had a market capitalisation approaching $18 billion, making it one of the largest stablecoins in existence.

UST maintained its peg through a mint-and-burn mechanism with LUNA. A user could always redeem one UST for one dollar’s worth of LUNA at the protocol level, and vice versa, creating an arbitrage opportunity that was supposed to enforce the peg. The Anchor Protocol, a lending platform on Terra, offered yields of approximately twenty percent annually on UST deposits, which attracted substantial capital and created high demand for the token.

In early May 2022, large outflows from Anchor combined with substantial selling pressure on UST caused it to break its peg. The response — minting large quantities of LUNA to absorb the selling — worked at first but quickly became self-defeating. As LUNA’s price fell due to dilution, the amount that needed to be minted to compensate each redeemed UST increased, accelerating the supply expansion and the price decline. Within days, LUNA had lost more than ninety-nine percent of its value and UST was trading at a small fraction of its one-dollar target. Tens of billions of dollars of value were effectively destroyed.

The Terra collapse triggered broader contagion across crypto markets, contributing to the failures of several other entities — including the hedge fund Three Arrows Capital and the lending platforms Celsius and Voyager Digital — that had significant exposure to the Terra ecosystem or were caught in the resulting market downturn.

Fiat-Backed Models: More Resilient, Not Risk-Free

The comparison between algorithmic and fiat-backed stablecoins is not simply a matter of one being safe and the other dangerous. Fiat-backed stablecoins carry their own risk set, which is less existential in normal conditions but can materialise in specific stress scenarios.

The primary risks for fiat-backed stablecoins are counterparty risk (the custodians holding the reserves may fail), composition risk (the reserve assets may not be as liquid or safe as claimed), and regulatory risk (an issuer may be unable to continue operating following regulatory action). The March 2023 episode in which USDC temporarily depegged after Silicon Valley Bank’s failure — where Circle held a portion of its cash reserves — illustrates how even a well-managed reserve-backed stablecoin can experience temporary peg stress when its counterparties are under pressure.

The critical difference is that fiat-backed stablecoins have an external anchor: the actual dollars in the reserve. A death spiral in the algorithmic sense is not possible because the mechanism for restoring the peg involves selling real assets, not minting new tokens of declining value. Losses in a reserve-backed model are bounded by the size of the reserve shortfall; losses in an algorithmic model are bounded only by the time it takes for confidence to collapse entirely.

Regulatory frameworks globally — including MiCA in the EU and proposed stablecoin legislation in the United States — have largely converged on the view that algorithmic stablecoins require either prohibition or very stringent capital and reserve requirements that would effectively require them to hold external backing. The regulatory direction of travel suggests that the pure algorithmic model — no collateral, no fiat backing, peg maintained entirely through token economics — faces an increasingly difficult legal environment in major jurisdictions.