What Is an Algorithmic Stablecoin? How They Work and Why They Fail

Stablecoins are the plumbing of crypto markets, letting traders move in and out of volatile assets without touching a bank. Most of the money sits in fully reserved, fiat-backed coins such as USDT and USDC. But a distinct and far riskier category has repeatedly captured attention: the algorithmic stablecoin, which aims to stay pegged using software and economic incentives instead of a dollar-for-dollar reserve. The idea is elegant on paper and dangerous in practice. This guide explains how algorithmic stablecoins work, walks through the mechanisms that power them, dissects the TerraUSD collapse that erased tens of billions of dollars, and explains why so few of these designs survive.

What is an algorithmic stablecoin?

A stablecoin is a token designed to track the price of a reference asset, almost always the US dollar. There are several ways to achieve that. Fiat-backed coins hold cash and short-term government debt in reserve. Crypto-collateralized coins lock up volatile assets worth more than the coins they issue. An algorithmic stablecoin takes a different route: it uses an automated protocol to manage the token’s supply, minting new coins when the price drifts above the peg and removing coins when it drifts below, in the hope that these supply changes push the price back to target.

The crucial distinction is backing. A purely algorithmic stablecoin is not fully backed by reserves; it leans on arbitrage incentives and market confidence to hold its value. That makes it capital-efficient, since it does not require a dollar in the bank for every token issued, but it also means the peg is only as strong as traders’ willingness to keep playing along. As one explainer from Chainlink notes, algorithmic stablecoins attempt to maintain their peg through on-chain mechanisms rather than off-chain collateral, which is precisely what makes them both innovative and fragile.

Types of algorithmic stablecoins and how they work

Algorithmic designs are usually grouped into three families, described in beginner guides from outlets such as Cointelegraph and Decrypt. Each manages supply differently.

1. Rebase stablecoins

Rebase models adjust the token balance in every holder’s wallet. If the price trades above the target, the protocol increases everyone’s balance proportionally; if it trades below, it shrinks every balance. The number of tokens you hold changes automatically, while the value of each token is meant to revert to $1. Ampleforth (AMPL) is the best-known rebasing design. The model relies heavily on accurate price oracles to trigger each “rebase” at the right moment, and critics point out that it shifts volatility from price to quantity rather than removing it.

2. Seigniorage (multi-token) stablecoins

Seigniorage-style systems use two or three linked tokens: the stablecoin itself, a “share” or governance token that absorbs volatility, and sometimes a “bond” token. When the stablecoin trades above $1, the protocol mints new coins and hands them to share holders, expanding supply to push the price down. When it trades below $1, the protocol offers bonds at a discount, letting users burn stablecoins now in exchange for a promise of more coins later, contracting supply to lift the price. Basis Cash was an early example, and TerraUSD was a seigniorage-style coin paired with LUNA. The fatal weakness is that the “buy the dip” incentive only works if people believe the system will recover; in a panic, nobody wants the bond or the share token, and the mechanism stalls.

3. Fractional-algorithmic stablecoins

Fractional-algorithmic designs are hybrids. They hold partial collateral and rely on an algorithm for the remainder, trying to blend the capital efficiency of a pure algorithm with the stability of real backing. Frax Finance pioneered this approach, launching with a collateral ratio that started at 100% and was lowered over time as confidence grew. The collateral cushion is meant to provide a floor that a purely algorithmic coin lacks, though, as we will see, even Frax ultimately retreated from the algorithmic portion of its model.

Case study: the TerraUSD (UST) and LUNA collapse

No event defines the risks of algorithmic stablecoins more than the collapse of TerraUSD in May 2022. UST was a seigniorage-style stablecoin tied to a sister token, LUNA, through a mint-and-burn arbitrage loop. The promise was simple: you could always burn $1 worth of LUNA to create one UST, or burn one UST to mint $1 worth of LUNA. In theory, that two-way swap would keep UST at a dollar, because arbitrageurs would profit whenever it drifted off peg.

The system grew explosively, fueled largely by the Anchor Protocol, a lending platform that offered depositors a yield of roughly 20% on UST. Analysts widely viewed that rate as unsustainable, but it drove enormous demand. At its peak, the Terra ecosystem was worth tens of billions of dollars.

The unwinding began in early May 2022. Around May 7, large withdrawals and heavy UST selling pushed the coin slightly below its peg. That should have been self-correcting, but the scale of selling overwhelmed the arbitrage mechanism. As holders rushed to redeem UST for LUNA, the protocol minted staggering quantities of new LUNA. According to Wikipedia’s account of the collapse, LUNA’s supply ballooned from roughly 725 million tokens to several trillion within days, destroying its price. LUNA, which had traded near an all-time high above $100, fell toward zero, and UST plunged far below its dollar peg, eventually trading in the pennies.

The Luna Foundation Guard tried to defend the peg with a reserve of bitcoin and other assets, but the reserves were a fraction of what was needed against billions of UST in circulation, and the defense failed. The Terra blockchain was halted on May 13, 2022. The scale of the damage is captured in the US Securities and Exchange Commission’s complaint against Terraform Labs and founder Do Kwon, which stated that UST, LUNA and related tokens lost more than $40 billion in combined market value by the end of May 2022. Estimates vary depending on whether you count peak market capitalization, total value wiped out, or realized investor losses, but every measure runs into the tens of billions.

Why algorithmic stablecoins fail: the death spiral explained

The Terra collapse is the textbook example of a death spiral, the structural flaw that haunts purely algorithmic designs. The mechanism works like this:

• Reflexivity. The stablecoin’s peg depends on the value of a second, volatile token (LUNA, in Terra’s case). But that second token derives much of its value from the demand for the stablecoin. The two are circular, so confidence in one props up the other, and a loss of confidence drags both down together.
• The trigger. A large sell-off or redemption pushes the stablecoin slightly below peg. The protocol responds by minting more of the volatile token to absorb the redemptions.
• Dilution. The new supply of the volatile token crashes its price. Because the stablecoin is “backed” by that token, the backing evaporates exactly when it is needed most.
• The run. Seeing the backing collapse, more holders rush to exit, which forces still more minting, which crushes the price further. This feedback loop accelerates until both tokens are nearly worthless.

In short, algorithmic stablecoins are vulnerable to classic bank-run dynamics with no deposit insurance and no hard asset to stop the fall. A fiat-backed coin can meet redemptions from reserves; a purely algorithmic coin can only print more of a token that is already collapsing. Once the market stops believing the peg will hold, the incentives that were supposed to restore it instead amplify the damage. This is why regulators and many economists treat uncollateralized algorithmic stablecoins as inherently unstable.

What survives, and why

The aftermath of Terra reshaped the stablecoin landscape. Several coins that are sometimes lumped in with “algorithmic” stablecoins are, on closer inspection, something sturdier.

DAI is crypto-collateralized, not purely algorithmic

DAI, issued by MakerDAO (now operating under the Sky brand), is often miscategorized as algorithmic because its peg is managed by a protocol rather than a company. In reality, DAI is overcollateralized: users mint it by locking crypto assets worth more than the DAI they create, through what the protocol calls vaults or collateralized debt positions. If collateral values fall too far, the position is liquidated to protect the peg. There is real backing behind every DAI, which is the decisive difference from Terra. Notably, much of that backing now consists of other stablecoins such as USDC, which makes DAI more robust but less decentralized than purists prefer.

Frax abandoned its algorithmic portion

Frax began as the flagship fractional-algorithmic stablecoin, at one point backing each FRAX with collateral covering only around 80–90% of its value and relying on its FXS token for the rest. After Terra, the community voted through proposal FIP-188 in February 2023 to move FRAX to full collateralization, effectively retiring the algorithmic component. Frax has since leaned into fully backed designs. The lesson its own governance drew is telling: even the most successful fractional-algorithmic project concluded that the algorithmic portion was not worth the tail risk.

Ethena’s USDe is a synthetic dollar, not a seigniorage coin

Ethena’s USDe is sometimes grouped with algorithmic coins, but its mechanism is different again. USDe is a “synthetic dollar” backed by a delta-neutral position: Ethena holds spot crypto collateral (such as staked ether) and simultaneously opens an equal and opposite short position in perpetual futures, so that price moves in the collateral are offset by the hedge. Its yield comes from staking rewards plus the funding payments on the short futures. This is not a supply-expansion algorithm trying to conjure stability from confidence; it is a hedged, collateral-backed strategy. It carries its own distinct risks, chiefly that futures funding rates can turn negative and that it depends on centralized exchanges for the hedge, but it does not share Terra’s reflexive death-spiral design.

The common thread among survivors is simple: they hold real assets. Whether it is overcollateralized crypto (DAI), full collateral (post-2023 Frax), or a hedged collateral position (USDe), the coins that endured all stepped away from the idea that software incentives alone can hold a peg. The purely algorithmic dream, a stablecoin that needs almost no backing, has yet to survive a serious stress test.

Frequently asked questions

Are algorithmic stablecoins safe to hold?

Purely algorithmic stablecoins carry significantly higher risk than fully reserved, fiat-backed coins. History shows they can lose their peg suddenly and irreversibly, as TerraUSD did in 2022. Anyone considering them should treat the peg as conditional rather than guaranteed, and understand that in a crisis there may be no reserves to redeem against. This article is educational and not financial advice.

What is the difference between an algorithmic and a collateralized stablecoin?

A collateralized stablecoin is backed by reserves, either fiat and cash equivalents (USDC, USDT) or overcollateralized crypto (DAI). An algorithmic stablecoin relies primarily on code that expands and contracts supply, with little or no full backing. The presence of real, redeemable assets is the key dividing line.

Is DAI an algorithmic stablecoin?

No. DAI is best described as a crypto-collateralized, overcollateralized stablecoin. Its peg is managed by the Maker/Sky protocol, but every DAI is backed by locked collateral worth more than the coins issued, which is fundamentally different from a purely algorithmic model.

How much money was lost in the TerraUSD collapse?

According to the SEC’s complaint against Terraform Labs, UST, LUNA and related tokens lost more than $40 billion in combined market value by the end of May 2022. Some estimates of peak-to-trough destruction run closer to $45 billion, while Chainalysis estimated roughly $20.5 billion in realized investor losses. The exact figure depends on the measure used.

Why do algorithmic stablecoins enter a “death spiral”?

Because their peg depends on a second, volatile token whose value is itself tied to demand for the stablecoin. When confidence drops, the protocol mints more of that token to defend the peg, crashing its price and the backing simultaneously. That feeds more selling in a self-reinforcing loop until both tokens collapse.

Do any algorithmic stablecoins still exist?

Some rebase and partially algorithmic projects still operate, but as of late 2026 none has established itself as a large, consistently stable, purely uncollateralized dollar coin. The projects that grew after Terra, including Frax and Ethena’s USDe, did so by moving toward real collateral or hedged positions rather than relying on supply algorithms alone.