An algorithmic stablecoin tries to hold a $1 price with code that changes the token supply, instead of with reserves in a bank. This guide explains the three algorithmic designs and why the best-known algorithmic stablecoins collapsed. It also shows how a decentralized stablecoin backed by crypto collateral holds its peg.
Stablecoins that an issuer backs with bank reserves, such as USDC, are covered in the stablecoin development company guide, along with the GENIUS Act and MiCA rules for those tokens. This page covers stablecoins where code, not an issuer, manages the supply or the collateral.
What an algorithmic stablecoin is
An algorithmic stablecoin has no reserves, or only part of the reserves it needs. A set of smart contracts expands the supply when the price is above $1 and shrinks it when the price is below $1. The design only works while people want to hold the token or the system's second token.
| Design | How it defends $1 | Examples |
|---|---|---|
| Two-token (seigniorage) | Holders can swap one stablecoin for $1 of a second token, and the reverse, and the system mints the second token to absorb selling. | TerraUSD (UST) with LUNA; Basis, which never launched |
| Rebase | The contract changes the balance in every wallet at once. Each holder keeps the same share of the supply, but the token count changes. | Ampleforth (AMPL), which targets a price but does not hold $1 day to day |
| Fractional | Part of each token is backed by a stablecoin such as USDC, and the rest relies on a second token that the system mints and burns. | Iron Finance (IRON with TITAN) |
Why algorithmic stablecoins fail
The backing of a two-token design is a token that the same system creates. When holders sell the stablecoin, the system mints more of the second token to pay them. The new supply pushes the second token's price down, so each redemption needs even more new tokens. This kind of loop is called reflexive, and it can run until the stablecoin and the second token are both worth almost nothing.
TerraUSD and LUNA, May 2022
TerraUSD held its peg through a swap with LUNA. In May 2022, large withdrawals and sales pushed UST below $1, and redemptions minted LUNA in huge amounts. Within about a week, UST fell to around 10 cents and LUNA fell to near zero. Terraform Labs filed for bankruptcy in January 2024. Co-founder Do Kwon pleaded guilty to two fraud charges in a US court in August 2025, and in December 2025 the court sentenced him to 15 years in prison.
Iron Finance, June 2021
Iron Finance backed IRON partly with USDC and partly with its own TITAN token. Large holders sold TITAN in June 2021, and redemptions of IRON minted more TITAN. TITAN's price fell to almost zero within a day, and IRON lost its peg.
Basis, December 2018
Basis raised $133 million for a two-token design, but the team shut the project down before launch and cited US securities law.
Regulation of algorithmic stablecoins
- United States: the GENIUS Act of July 2025 lets only permitted issuers with full reserves issue payment stablecoins. Section 14 of the Act ordered the Treasury to study "endogenously collateralized" stablecoins within one year, defining them as tokens that keep a fixed price using only another digital asset made by the same originator.
- European Union: under MiCA, an unbacked algorithmic token does not qualify as an e-money token or an asset-referenced token, so it falls under the general crypto-asset rules and cannot be marketed as a regulated stablecoin.
- Other countries: Bahrain's 2025 stablecoin framework prohibits algorithmic stablecoins.
Decentralized stablecoins backed by collateral
Most decentralized stablecoin projects now use crypto collateral instead of a pure algorithm. MakerDAO started this model with DAI, then renamed itself Sky in 2024 and added USDS as an upgraded stablecoin, though DAI is still in circulation. Most collateral designs follow these steps:
- A user deposits collateral, such as ETH or a liquid staking token, into a vault.
- The user mints stablecoins up to a limit set by the collateral ratio; a 150% ratio allows $100 of debt for each $150 of collateral.
- The debt pays a stability fee, which is interest. The protocol raises the fee to shrink supply when the coin trades below $1 and lowers it to grow supply when the coin trades above $1.
- If the collateral value falls below the liquidation level, anyone can start a liquidation, and the protocol sells the collateral, by auction or at a discount, to repay the debt plus a penalty.
- To close the vault, the user repays the stablecoins, which the protocol burns before it returns the collateral.
Design choices in a collateral-backed stablecoin
| Component | Options | What can go wrong |
|---|---|---|
| Collateral types | ETH, liquid staking tokens, wrapped BTC, tokenized Treasury bills, other stablecoins | Collateral prices crash together. Wrapped or staked assets lose their peg. Custodial assets add a central point of control. |
| Price oracles | Chainlink-style feeds, prices from several sources, time delays | A stale or manipulated price triggers wrong liquidations or blocks liquidations that should happen. |
| Liquidation engine | Dutch auctions, fixed-discount sales, stability pools | Liquidation bots stop working when the network is congested. Auctions sell collateral for almost nothing. |
| Peg tools | Stability fees, savings rates, modules that swap 1:1 with other stablecoins, redemption against collateral | Swap modules make the peg depend on centralized stablecoins. |
| Governance | Token voting, fixed parameters, rates set by an algorithm | An attacker buys or borrows voting power. Votes are slow in a crisis. |
| Backstop | A surplus buffer, or minting governance tokens to cover bad debt | Bad debt grows larger than the buffer. |
Liquidations under stress
On March 12, 2020, ETH fell sharply and the Ethereum network was congested, so some Maker auctions sold collateral for bids near zero. The protocol was left with bad debt and covered it by minting and selling its MKR governance token. A liquidation system must keep working when gas is expensive, prices jump, and liquidation bots are overloaded.
Liquity's answer to this problem was a stability pool. Depositors put stablecoins into the pool in advance, and the pool absorbs liquidated debt at once in exchange for the collateral.
Redemptions as a price floor
Some protocols let anyone redeem one stablecoin for $1 of collateral, taken from specific vaults. Traders buy the coin below $1 and redeem it, so the price cannot stay far below $1. Liquity v1 used this rule for LUSD, with contracts that nobody can change and no governance, while Liquity v2 issues BOLD and lets borrowers set their own interest rate.
Hedged synthetic dollars
Ethena's USDe holds crypto and sells an equal amount of perpetual futures, and the two positions offset each other, so the total value stays close to the dollar. USDe earns the funding payments that futures traders pay while those payments are positive. The risks are long periods of negative funding, the failure of an exchange, and custody of assets off the blockchain. During a market crash in October 2025, USDe traded near $0.65 on Binance while its price on other venues stayed near $1. USDe depends on exchanges and custodians, so it is not decentralized in the way a vault design is.
Engineering and security work
- Model the economics before you write contracts. Simulate price crashes of 30% to 60% within hours, oracle delays and network congestion; agent-based simulations show which parameter sets fail.
- Keep the core contracts small and well tested. Formal verification can prove key rules, such as "total debt equals the sum of all vault debts".
- Treat oracles as critical systems. Use several price sources, sanity limits and circuit breakers, and never rely on the spot price of a single DEX pool. The DeFi lending platform guide covers the same problem for loans.
- Run several independent audits and a bug bounty. A system that holds large amounts of collateral needs more than one smart contract audit.
- Plan for governance attacks, such as voting power borrowed through a flash loan. A timelock on parameter changes gives users time to exit.
A minimal single-collateral vault protocol takes three to five experienced Solidity engineers several months, plus an economist or risk analyst. That figure is an estimate based on the scope above, and audits and simulations add time and cost on top. The DeFi development guide covers related protocol work.
Frequently asked questions
Is an algorithmic stablecoin the same as a decentralized stablecoin?
No. An algorithmic stablecoin relies on supply changes or a second token instead of full collateral. A decentralized stablecoin such as DAI or LUSD uses smart contracts and is backed by more crypto collateral than the coins issued.
What caused TerraUSD to fail?
TerraUSD had no outside collateral. When holders rushed to sell, the system minted more and more LUNA, and LUNA's price and the peg collapsed together.
Why must a decentralized stablecoin hold extra collateral?
Crypto prices move fast, and a liquidation takes time. The extra collateral covers the price drop between the start and the end of a liquidation, so each stablecoin stays backed.
Is a stablecoin backed by USDC in a swap module decentralized?
Only in part, because the peg then depends on USDC and on Circle's power to freeze funds. Many protocols accept that dependence to keep the peg stable.
Who runs liquidations?
Independent bots called keepers run liquidations and earn the penalty or discount. A design must keep keepers profitable even when gas prices are extreme.