Stablecoins look simple from the outside. One token, one dollar, always. Underneath, four different peg mechanisms compete to deliver that promise, and each one breaks in a different way when stress hits. This guide walks through how each design holds the peg, where reserves sit, how mint and burn flows work, and what happens when the system fails.
What keeps a stablecoin pegged to $1?
A stablecoin holds its peg through a combination of redeemable backing (reserves or collateral), arbitrage incentives (mint/burn at par), and market liquidity. When the price drifts, authorized participants or smart contracts step in to mint cheap tokens or redeem expensive ones until the price returns to $1.
The peg is not a law of physics. It is a behavioral equilibrium. As long as redemption is credible and reserves are liquid, traders will arbitrage any deviation. When either of those assumptions cracks, the peg cracks with it.
The four peg mechanisms compared
Every stablecoin in circulation today fits into one of four designs: fiat-backed, crypto-collateralized, algorithmic, or hybrid. The differences matter because they fail in different ways under stress.
Mechanism | Example | Pros | Cons | Failure mode |
Fiat-backed (reserve-backed) | USDC, USDT, PYUSD | Simple, scalable, deep liquidity | Centralized, custodial risk, regulatory dependence | Reserve loss or banking failure (USDC, March 2023) |
Crypto-collateralized | DAI, USDS | Onchain transparency, censorship resistant | Capital inefficient, oracle risk | Cascading liquidations during crashes (Black Thursday, 2020) |
Algorithmic | UST (defunct) | No collateral required, capital efficient | Reflexive, depends on demand for sister token | Death spiral when arbitrage incentive inverts (UST, May 2022) |
Hybrid | FRAX (historical), USDe | Combines collateral with delta-neutral hedges or partial backing | Complex, dependent on funding rates or external venues | Funding-rate inversion, hedge counterparty failure |
Fiat-backed stablecoins: USDC, USDT, and the reserve question
Fiat-backed stablecoins are the largest category by supply. Circle issues USDC. Tether issues USDT. Each token represents a claim on a dollar (or near-dollar instrument) held by the issuer. The simplicity is the appeal, and the reserve composition is the entire risk story.
Reserves typically sit in four buckets. Cash and bank deposits provide instant liquidity. Short-duration U.S. Treasury bills earn yield while staying liquid. Overnight repos backed by Treasuries add another short-duration tier. Money-market funds (sometimes the issuer's own, sometimes BlackRock's BUIDL) round out the mix. Circle's monthly attestations and Tether's quarterly reports break down the percentages.
Mint and burn flows run through authorized participants. A market maker wires dollars to Circle, Circle mints USDC at par. To redeem, the participant burns USDC and Circle wires dollars back. Retail users almost never touch this primary market. They buy and sell on exchanges, where arbitrageurs keep the secondary price close to $1 by exploiting any gap against the redemption window.
Crypto-collateralized: how DAI and USDS stay overcollateralized
Crypto-collateralized stablecoins replace bank reserves with onchain collateral. Sky (formerly MakerDAO) is the canonical example. Users lock ETH, staked ETH, real-world assets, or other approved collateral into vaults, then mint DAI or USDS against it at a loan-to-value ratio below 100%. If collateral value drops, the vault is liquidated and the position is closed.
Two mechanisms keep the peg. First, the stability fee (interest on minted DAI) and DAI Savings Rate adjust supply by changing the cost of borrowing and the reward for holding. Second, the Peg Stability Module lets users swap USDC for DAI 1:1 with no slippage, capping deviations within a tight band. Sky publishes the collateral mix, debt ceilings, and PSM balances on its dashboard.
Algorithmic stablecoins: the UST lesson
Algorithmic stablecoins try to maintain a peg without holding equivalent collateral. Terra's UST is the cautionary tale. UST was redeemable for $1 of LUNA at any time. If UST traded below $1, arbitrageurs were supposed to burn UST and mint LUNA at a profit, contracting supply until the peg returned.
The design works in calm markets and breaks in panicked ones. In May 2022, large UST withdrawals from the Anchor protocol triggered a sell-off. As UST dropped, arbitrageurs minted billions of LUNA tokens, collapsing LUNA's price. Lower LUNA price meant less backing per UST, which accelerated redemptions. The death spiral wiped out roughly $40 billion in market value across the Terra ecosystem in under a week. Pure algorithmic designs have not recovered as a category since.
Hybrid models: USDe and the delta-neutral approach
Hybrid stablecoins blend collateral with derivatives or partial algorithmic mechanics. Ethena's USDe is the most prominent live example. Users deposit staked ETH or BTC, and Ethena opens an equal-and-opposite short perpetual position on a centralized exchange. The long spot and short futures positions cancel out price exposure, leaving a synthetic dollar backed by the collateral plus funding-rate yield.
The design is elegant when funding rates are positive. It strains when funding inverts (longs pay shorts) for extended periods, when exchange counterparties get stressed, or when collateral liquidity dries up. Ethena publishes reserve attestations and exchange exposure, but the model is younger than the fiat-backed or crypto-collateralized cohorts and has not been tested through a deep bear market.
What does a reserve actually hold?
For fiat-backed issuers, reserve composition has converged on a similar template since 2023. Cash deposits across diversified banks (often Bank of New York Mellon, Customers Bank, and others post-Silicon Valley Bank). U.S. Treasury bills with maturities under three months. Overnight reverse repurchase agreements backed by Treasuries. Money-market funds with same-day or next-day liquidity.
Circle's published attestations show USDC reserves split roughly 80 to 90 percent across short Treasuries and repos, with the remainder in cash. Tether's quarterly reports break down USDT reserves into Treasuries, secured loans, bitcoin, gold, and other assets. The proportion of non-cash, non-Treasury assets in Tether's reserve is one of the longest-running debates in the category. Reserve attestations (covered in our reserve attestation guide) explain the audit limits and what an attestation does and does not certify.
How does redemption actually work?
Two paths exist. Primary redemption goes through the issuer. A KYC-approved counterparty (usually an exchange, market maker, or large treasury) sends tokens to the issuer's redemption address and receives a wire in return. Minimum sizes typically start at $100,000 for Circle and run higher for Tether. Settlement is same-day or next-day depending on banking hours.
Secondary redemption is what retail uses. You sell USDC or USDT on Binance, Coinbase, Kraken, or a DEX, and a market maker on the other side eventually rolls that inventory back to the issuer through the primary window. The arbitrage between secondary price and primary redemption is what keeps a fiat-backed token within fractions of a cent of $1 in normal conditions.
When the peg breaks: USDC March 2023 and what we learned
On March 10, 2023, Silicon Valley Bank failed. Circle disclosed that $3.3 billion of USDC reserves (roughly 8 percent of the float) sat in SVB. Over the weekend, USDC dropped below $0.90 on secondary markets as holders raced for the exits and primary redemption was paused due to closed banks. When the FDIC announced on Sunday that depositors would be made whole, USDC repegged within hours.
Two lessons stuck. First, even fully backed stablecoins can depeg sharply when redemption is friction-blocked, even temporarily. The reserves were there. The settlement rails were not. Second, banking diversification matters. Issuers now spread reserves across multiple custodians and weight more toward Treasuries and repos that settle through the Federal Reserve rather than commercial banks alone.
UST's collapse ten months earlier taught the harder lesson. A peg that depends on a sister token's market cap is reflexive. When confidence drops, the mechanism that should restore the peg destroys the asset backing it. Algorithmic-only designs are functionally extinct in 2026, replaced by overcollateralized or delta-neutral models.
How do these designs stack up for users in 2026?
For most enterprise treasury use cases, fiat-backed stablecoins (USDC, USDT, PYUSD, USDG) remain the default. They scale, they settle through known regulatory perimeters, and their reserves are auditable. For users who prioritize censorship resistance and onchain transparency, DAI and USDS offer overcollateralized exposure with no dependence on commercial banks. Hybrid models like USDe attract yield-seekers willing to underwrite funding-rate risk. Pure algorithmic stablecoins are no longer a serious category.
The choice is not just about which design feels safest in the abstract. It is about which failure mode you can tolerate and monitor. Reserve risk, oracle risk, funding-rate risk, and counterparty risk all show up differently in your stack.
Methodology and sources
This guide draws on Circle's monthly reserve attestations (circle.com/transparency), Tether's quarterly reserve reports (tether.to/transparency), Sky's collateral dashboard (sky.money), and Ethena's transparency page. The USDC depeg analysis references Circle's March 2023 disclosures and FDIC press releases. The UST collapse references Terra's onchain mint/burn data and post-mortem reports from Nansen and Chainalysis published in mid-2022.

