A stablecoin is a token that holds a target price (almost always one US dollar) by pairing each unit with a redeemable claim on something else. The "something else" splits the category into four groups: fiat-backed stablecoins like USDC and USDT sit on cash and Treasury bills, crypto-collateralized stablecoins like DAI / USDS sit on overcollateralized onchain assets, algorithmic and synthetic-dollar stablecoins like USDe sit on derivatives or pure code, and commodity-backed stablecoins like PAXG sit on bullion in a vault. Total supply across the four buckets reached roughly $321 billion in April 2026 per the DeFiLlama stablecoins dashboard, with Artemis recording about $33 trillion in onchain transfer volume during 2025 per Bloomberg. This article walks through what each category backs, how the peg is held, what has gone wrong
​

(TerraUSD), and how supply is distributed today.
​

A category is defined by two design choices: what backs the token, and what mechanism enforces redemption at par. Backing is the asset that settles a redemption claim. The mechanism is the procedure that keeps the secondary-market price within a few basis points of par when balance flows in either direction. Backing alone does not enforce a peg: a fiat-backed token with reserves but no public redemption channel will trade off-peg whenever an arbitrageur cannot close the loop, and a crypto-collateralized token without a working liquidation engine will accumulate undercollateralized debt during a drawdown.
​

Fiat-backed stablecoins back each token with cash and short-duration government securities held by a regulated custodian. Crypto-collateralized stablecoins back each token with onchain assets locked in vaults at a collateralization ratio above 100%. Algorithmic stablecoins back each token with code: a mint-burn loop, a delta-neutral derivative position, or a rebasing supply rule. Commodity-backed stablecoins back each token with allocated physical bullion custodied off-chain.
​

Fiat-backed assets carry about 93% of the $298 billion total supply per DeFiLlama. The remaining 7% spreads across crypto-collateralized at roughly $10 billion, algorithmic / synthetic-dollar at around $7 billion, and commodity-backed under $2 billion. Liquidity follows supply, which is why any large stablecoin trade tends to route through USDC or USDT regardless of which category the holder might prefer in principle.
​

Fiat-backed stablecoins are the dominant category by supply, by liquidity, and by regulatory attention. The mechanism is the simplest of the four: the issuer takes one US dollar from a customer, mints one token, and holds the dollar (or a near-equivalent like a 1-month Treasury bill) in a custodial account at a regulated bank. Redemption runs the same loop in reverse. Independent accountants publish attestations on a fixed cadence, monthly for Circle and quarterly for Tether, signing off on the asset composition behind the float.
​

USDT from Tether sits at roughly $182 billion in circulating supply as of Q1 2026 per the BDO Italia attestation. Reserves consist primarily of US Treasury bills, with smaller positions in money-market funds, secured loans, gold, and bitcoin. USDT is live on more than 15 chains and carries the deepest liquidity of any single stablecoin, especially on Tron, Ethereum, and centralized exchange order books.
​

USDC from Circle carries about $77 billion in supply per Deloitte attestations in 2026. Reserves are 100% cash and short-duration Treasuries held with BNY Mellon and BlackRock. Circle publishes monthly attestations and a daily reserves dashboard. USDC is the default fiat-backed asset in DeFi protocols, partly because Circle ran a public redemption window during the March 2023 Silicon Valley Bank episode that resolved the depeg within 72 hours.
​

PYUSD from PayPal is issued by Paxos under New York Department of Financial Services trust supervision. Supply was roughly $3.5 billion in April 2026 across Ethereum, Solana, Arbitrum, Stellar, and a LayerZero PYUSD0 variant on nine additional chains. PYUSD is the only major fiat-backed stablecoin issued by a public consumer payments company.
​

RLUSD from Ripple launched in December 2024 under the same NYDFS framework as PYUSD, dual-issued on Ethereum and the XRP Ledger. USDG from Paxos under MAS regulation in Singapore distributes a share of reserve yield back to ecosystem partners (Robinhood, Kraken, Galaxy, Anchorage). FDUSD from First Digital Labs in Hong Kong carries about $415 million in supply and is the most widely used non-USDC, non-USDT pair on Binance.
​

Fiat-backed risk concentrates in three places. Reserve quality: the difference between a basket of cash plus T-bills and one that includes commercial paper or undisclosed loans shows up under stress. Custodial failure: USDC traded as low as $0.87 on March 11, 2023 when $3.3 billion of its reserves sat at the failing Silicon Valley Bank, with par restored only after the FDIC backstop was announced. Regulatory action: BUSD's wind-down in February 2023 followed an NYDFS order against Paxos, taking a fully collateralized stablecoin from $16 billion supply to zero in roughly six months.
​

Crypto-collateralized stablecoins keep their reserves onchain rather than in a bank. A user locks volatile crypto (ETH, wstETH, WBTC, sometimes USDC itself) in a smart contract called a vault, and the protocol mints stablecoin against that collateral at a collateralization ratio above 100%. If the collateral price drops below the liquidation threshold, keepers repay the debt, take the collateral at a discount, and the borrower loses any excess.
​

DAI / USDS from MakerDAO (now Sky) is the longest-running crypto-collateralized stablecoin, launched in 2017 and rebranded to USDS in 2024. The protocol backs USDS with a mix of onchain crypto (ETH, wstETH, WBTC), real-world asset Treasury exposure through vaults like Monetalis and BlockTower, and stablecoin reserves via the Peg Stability Module. Combined DAI plus USDS supply was about $13.4 billion in April 2026 per DeFiLlama. Minimum CR varies by vault type but typically lands between 145% and 175%.
​

LUSD from Liquity takes only ETH as collateral, runs at a 110% minimum CR, charges no recurring borrowing fee, and holds an immutable contract set with no admin keys. Liquity V2, branded BOLD, expands the collateral set to LSTs and adds user-set interest rates that compete in a continuous auction. GHO from Aave mints against deposits inside Aave V3 lending pools; borrowers pay an interest rate set by the AaveDAO (around 4.3%) that flows directly to the DAO treasury. crvUSD from Curve uses a continuous-liquidation mechanism called LLAMMA that gradually converts collateral into stablecoin across a price band, softening the impact of a fast drawdown.
​

The category's recurring risk is collateral cascade. When ETH drops 30% in a few hours, vaults near the liquidation threshold all hit it at once, keepers race to claim discounts, oracles can lag the spot market, and the protocol's surplus buffer absorbs the difference between the debt repaid and the collateral seized. MakerDAO faced exactly this on March 12, 2020 ("Black Thursday"), when network congestion let some keepers buy ETH collateral for zero DAI bids; the protocol covered the resulting bad debt by minting and auctioning MKR governance tokens, a mechanism the Maker docs still describe.
​

Capital efficiency is the trade-off. A 150% collateralization ratio means a holder ties up $1.50 of crypto for every $1 of stablecoin. The case for the category is censorship resistance: redemption does not depend on a bank account or an issuer staying in business.
​

Algorithmic stablecoins are the smallest of the four buckets and the most varied. The original "purely algorithmic" design used a two-token mint-burn loop with no exogenous backing, with TerraUSD plus Luna being the canonical instance. That design has not held a peg through a stress event since 2022 (see the next section). The post-Terra category survives mostly through delta-neutral synthetic-dollar designs: instead of relying on reflexive demand, the protocol holds a long spot position in volatile collateral and an offsetting short perpetual, so spot price moves cancel against perp PnL and funding payments serve as the yield source.
​

USDe from Ethena is the most prominent example. Supply reached approximately $5.4 billion in April 2026 per DeFiLlama. A user mints USDe by depositing ETH or stETH; Ethena simultaneously opens a short perpetual on a venue like Binance, OKX, or Bybit equal in size to the spot leg. The peg holds because the dollar value of (long spot + short perp) stays roughly constant. Yield (paid as sUSDe to stakers) comes from the funding rate paid by long-side perp traders. The Ethena docs describe the mechanism, the venue diversification policy, and the off-exchange settlement custody arrangements.
​

USR from Resolv uses a similar delta-neutral architecture but splits risk into two tokens: USR for stable-value holders and RLP as a liquidity buffer. FRAX from Frax Finance began in 2020 as a fractional-algorithmic design and has since rebranded its core product as frxUSD with full collateral backing.
​

The structural risks cluster in three places. Funding-rate inversion: when sentiment turns deeply bearish, perp funding rates can flip negative for an extended period, and the protocol pays rather than receives. Venue counterparty risk: the short legs sit on centralized exchanges, and an exchange failure (FTX-style) would force unwinding into a stressed market. Basis compression: the rate paid to stakers can compress as more capital chases the trade. Most assets called "algorithmic stablecoins" today are actually delta-neutral synthetics with real collateral.
​

Commodity-backed stablecoins peg to a physical commodity rather than a fiat currency. Gold dominates the category; silver and other metals exist but carry negligible supply. The mechanism mirrors fiat-backed assets: the issuer custodies allocated bullion at a vault provider, mints one token per defined unit (typically one troy ounce of gold), and stands ready to redeem.
​

PAXG from Paxos represents one troy ounce of London Bullion Market Association (LBMA) good-delivery gold held in Brink's London vaults. Each token is allocated to a specific bar, and Paxos publishes a serial-number lookup tool. Supply was approximately $2.2 billion in April 2026, tracking a gold spot price around $4,600 per ounce. Redemption for physical gold is available for whole-bar quantities (roughly 430 ounces); fractional balances trade on secondary markets.
​

XAUT from Tether Gold follows the same one-token-per-troy-ounce model with bullion vaulted in Switzerland. PAXG and XAUT combined are roughly 78% of all commodity-backed stablecoin supply. KAU and KAG from Kinesis tokenize one gram of gold and one ounce of silver respectively, vaulted at Brink's locations, and pay holders a yield share of network transaction fees.
​

The category's risks are vault risk, insurance counterparty risk, and basis between spot gold and the token's secondary-market price. Commodity-backed assets carry wider bid-ask spreads and less depth on most exchanges, which is why the category serves portfolio diversification and inflation-hedge use cases more often than payments or DeFi collateral.
​

Each category trades a different combination of capital efficiency, decentralization, regulatory exposure, and tail-risk profile. The matrix below summarizes the trade-offs at the four-axis level: what backs the token, how the peg is enforced, the dominant failure mode, and a live example with current supply.

Table 1. The four categories on backing, peg mechanism, primary risk, and supply.
​

Fiat-backed dominates because institutional treasuries, payment processors, and exchanges already understand the design. Crypto-collateralized retains a permanent niche in DeFi because no centralized issuer can freeze a wallet that holds it. Synthetic dollars fill a yield-bearing slot fiat-backed assets structurally cannot, since most issuers retain reserve yield rather than pass it through. Commodity-backed sits in the portfolio-diversification corner.
​

TerraUSD is the most expensive stablecoin failure on record and the cleanest illustration of why category design matters. The protocol launched in 2020 with a two-token mint-burn loop: holders could always swap one Luna for $1 of UST, and $1 of UST for one Luna. The arbitrage closed any peg deviation as long as the market valued Luna above zero. Anchor Protocol, a UST-denominated lending application, paid a fixed 19.5% yield to UST depositors and pulled supply to over $18 billion by early 2022.
​

The structural flaw was reflexivity. Luna's market cap had to exceed outstanding UST for the swap mechanism to absorb a peg break. When UST demand fell on May 7, 2022 (large Anchor withdrawals — including a 375 million UST withdrawal — triggered the cascade), arbitrageurs swapped UST for Luna and immediately sold Luna into a thinning bid. Luna supply expanded from 350 million to over 6.5 trillion tokens in 72 hours. UST traded at $0.30 by May 11 and never recovered. Aggregate UST and Luna losses ran to approximately $45 billion per the CFTC complaint against Terraform Labs, with a separate SEC settlement covering fraud charges against the founder.
​

The teaching point is not that algorithmic stablecoins are impossible. A peg held only by reflexive demand has no floor. Every algorithmic-stablecoin design that has held par through a stress event since Terra (USDe, USR, frxUSD) has done so by holding actual collateral, just structured differently from a bank account. The "algorithmic" name now describes the peg mechanism, not the absence of backing.
​

Two regulatory regimes have responded. The GENIUS Act (Generating and Establishing National Innovation for U.S. Stablecoins) passed both chambers and was signed into law on July 18, 2025. It authorizes payment stablecoins from insured depository institution subsidiaries, federally qualified non-bank issuers under OCC supervision, and state-qualified issuers in substantially similar jurisdictions. Reserves must be cash, demand deposits, short-duration Treasuries, or repos collateralized by the same. The act prohibits marketing algorithmic stablecoins as payment stablecoins. Bill text sits at congress.gov. MiCA (the EU Markets in Crypto-Assets regulation) took effect for stablecoins on June 30, 2024, defining e-money tokens (EMTs) and asset-referenced tokens (ARTs); both must hold reserves at credit institutions. MiCA effectively excludes purely algorithmic designs from the EU market and triggered a temporary USDT delisting on regulated European venues during 2024. The regulation text lives at EUR-Lex. Both regimes leave room for crypto-collateralized (DAI / USDS) and commodity-backed (PAXG) stablecoins to operate; synthetic dollars (USDe) sit in a gray zone the OCC will need to resolve in implementing rules.
​

Stablecoin holders typically need more than one category and more than one chain. A treasury might custody USDC on Ethereum, hold operational balance in PYUSD on Solana, run DeFi yield through DAI on Arbitrum, and keep an inflation hedge in PAXG. Moving balance between those positions involves bridges, swaps, and chain-by-chain accounting that gets expensive fast. Eco is a stablecoin orchestration network that takes an intent ("deliver USDC to address X on Base, source from any stablecoin I hold") and routes it through solvers across 15 supported chains in one atomic operation. Eco Routes exposes the routing layer through a CLI and API; Eco Portal handles 1:1 stablecoin swaps and cross-chain transfers in a single click. Either the intent clears end-to-end or it reverts, which removes the partial-failure recovery code that complicates manual cross-category rebalancing.
​

Fiat-backed (USDC, USDT), crypto-collateralized (DAI / USDS, LUSD), algorithmic or synthetic-dollar (USDe), and commodity-backed (PAXG, XAUT). Fiat-backed dominates supply at about 93% of the $298 billion total per DeFiLlama; the other three share the remaining 7%.
​

Fiat-backed assets from regulated issuers (USDC, PYUSD, RLUSD) carry the lowest tail risk under normal conditions but depend on bank custody and regulatory continuity. Crypto-collateralized assets like DAI / USDS reduce custodial risk but add collateral-cascade risk. The right choice depends on what failure mode a holder can tolerate.
​

TerraUSD relied on a mint-burn loop with Luna as the absorbing asset; when UST demand fell, Luna supply expanded into a falling market and both assets imploded. Pure-algorithmic designs without collateral have not held par through a stress event since. Current "algorithmic" stablecoins like USDe hold real collateral structured through delta-neutral derivatives.
​

The issuer holds one US dollar (or a Treasury bill near-equivalent) for every token outstanding, custodied at a regulated bank. Users redeem 1:1 through the issuer or an authorized partner. Arbitrageurs keep the secondary-market price within a few basis points of par by minting or redeeming whenever the price diverges.
​

USDC is fiat-backed: Circle holds cash and Treasuries with BNY Mellon and BlackRock against every token outstanding. DAI / USDS is crypto-collateralized: MakerDAO (Sky) mints against onchain collateral locked at over 100%. USDC is more capital-efficient; DAI is more decentralized.
​

Yes, by swapping one for another on a DEX or aggregator. The swap is a market transaction at the prevailing price, not a category conversion at par. Eco Portal routes 1:1 stablecoin swaps across 15+ chains, which lets a holder rebalance between USDC, USDT, DAI, PYUSD, and others without manually bridging.
​