What are the four types of stablecoins?

Fiat-backed (USDT, USDC, FDUSD, EURC), crypto-collateralised (DAI, USDS, GHO), algorithmic (the Luna UST design — now effectively extinct at scale), and commodity-pegged (PAXG, XAUT for gold; smaller experiments for oil and other commodities). Each family holds peg differently and breaks differently.

Which stablecoin type is the safest?

Fiat-backed stablecoins from regulated issuers — USDC under NY DFS, USDT under quarterly BDO attestation, EURC under MiCA — have the strongest combination of reserve verifiability and stress survival. The honest qualifier: 'safest' is relative, no stablecoin has zero risk, and the failure modes for fiat-backed (counterparty bank failure, regulator action) are different from crypto-backed (collateral liquidation cascade) and from algorithmic (death-spiral runs).

Is DAI safer than USDC because it is decentralised?

Not in 2026. DAI's collateral pool has included a meaningful share of USDC since 2020, which means DAI inherits some USDC counterparty risk. The Sky Protocol (formerly Maker) has diversified the collateral mix and introduced USDS as the rebranded primary stablecoin, but the inheritance of fiat-backed reserve risk is real. Decentralisation reduces some risks and adds others.

Why are commodity-pegged stablecoins less popular?

Gold-backed tokens like PAXG and XAUT exist and have stable markets, but they track gold not US dollars, which means they are not interchangeable with USDT or USDC for most crypto use cases. The market for a tokenised gold position is smaller than the market for a tokenised dollar position; the latter is the dominant use case in crypto.

Are there any working algorithmic stablecoins?

Pure algorithmic designs of the UST type effectively do not exist at scale. Hybrid designs — partially-algorithmic with meaningful collateral — exist (Frax has moved to fully collateralised; USDD operates with substantial reserves on Tron). The pure category is excluded from EU MiCA and the proposed US GENIUS Act.

Four Stablecoin Types · Fiat, Crypto, Algo, Commodity

Family 01 · Fiat-backed

Peg mechanismThe issuer holds real-world reserves — typically cash deposits and short-dated government bonds — equal to or greater than the tokens in circulation. Holders can redeem tokens with the issuer for the underlying asset. Arbitrageurs keep the secondary-market price near peg because any meaningful gap is profitable to close.

Reserve compositionFor USDT (mid-2026): 84% cash equivalents (mostly Treasury bills), the rest in corporate bonds, precious metals and other investments. For USDC: roughly 80% in the BlackRock-managed USDXX money-market fund, 20% in cash at chartered US banks.

Primary failure modeCounterparty bank failure for a custodian holding part of the reserves (USDC and the SVB weekend in 2023), reserve shortfall (the historical NYAG concern with Tether before 2021), or regulator-forced wind-down (BUSD in early 2023 under NY DFS).

Where it worksSpot trading on centralised exchanges, cross-border payment, working capital, the default for any first-time stablecoin holder. The deepest liquidity by an order of magnitude over any other family.

Where it does notHolders who want fully on-chain transparency on every reserve unit; holders in jurisdictions where the regulator of the issuer is politically hostile; long-term holders who want diversification beyond a single issuer.

Market shareRoughly 95% of total stablecoin market cap as of mid-2026.

Family 02 · Crypto-collateralised

Peg mechanismThe stablecoin is minted against a basket of crypto collateral held in a smart contract. The collateralisation ratio is over 100% — typically 130% to 150% — so that the position remains solvent even if the collateral price drops. If the ratio falls below a liquidation threshold, the protocol auctions off the collateral to repay the stablecoin debt.

Reserve compositionFor DAI / USDS in 2026: a mix of ETH-staked positions (stETH and similar), USDC (which makes the design partly inherit USDC risk), various real-world-asset positions added since 2022 (tokenised T-bills, MakerDAO RWA vaults), and smaller positions in other crypto assets.

Primary failure modeCollateral price drop too fast for liquidation auctions to keep up (the "Black Thursday" March 2020 episode where DAI briefly traded above peg because liquidators could not get the keeper bots to fire); USDC contagion in the collateral mix (the March 2023 weekend caused DAI to depeg in sympathy); smart-contract risk in the underlying protocol.

Where it worksDeFi positions where on-chain transparency of collateral matters; holders philosophically aligned with decentralised governance; positions inside specific protocols (Maker / Sky native users, Aave borrow positions). LUSD remains the most "pure" version of the crypto-collateralised design (ETH-only collateral, no fiat-backed positions).

Where it does notFirst-time stablecoin holders looking for the simplest possible exposure; holders who want zero exposure to a smart-contract layer; cross-border payment use cases where receiver familiarity matters.

Market shareRoughly 3% of total stablecoin market cap as of mid-2026.

Family 03 · Algorithmic

Peg mechanismThe token holds peg through code-based mint-and-burn arbitrage with a paired token, without (in the pure form) any exogenous collateral. When the stablecoin trades below peg, arbitrageurs burn it to mint the paired token at peg value and sell the paired token for profit. The mechanism is elegant in calm weather and structurally fragile in a coordinated run.

Reserve compositionFor pure algorithmic designs: none. For hybrid designs that learned the Luna lesson, partial collateral: USDD on Tron holds significant USDC, TRX and BTC reserves; FRAX moved to fully collateralised in 2023; smaller experimental designs vary.

Primary failure modeThe death spiral. As confidence in the stablecoin falls, holders exit; the burn mechanism mints more of the paired token to absorb the exits; the paired token's price falls under the dilution; each subsequent burn produces more of the paired token at a lower price; the system enters a self-reinforcing collapse. Luna and UST went from a combined 60 billion to under 200 million in ten days in May 2022.

Where it worksResearch and academic study. Pure algorithmic designs are not a holdings category in 2026.

Where it does notAnywhere a holder is allocating real capital. EU MiCA explicitly excludes pure algorithmic designs from the stablecoin category. The proposed US GENIUS Act does the same.

Market shareLess than 0.5% of stablecoin market cap as of mid-2026, mostly from hybrid designs that should arguably be reclassified as partially-collateralised.

Family 04 · Commodity-pegged

Peg mechanismThe issuer holds one unit of the underlying commodity (typically one troy ounce of gold) for each token. Holders can redeem tokens for the physical commodity, usually with a fee and a delivery process. PAXG holds gold in LBMA-good-delivery vaults; XAUT similarly.

Reserve compositionOne-to-one with the commodity. PAXG publishes vault locations and quarterly attestations. XAUT publishes a daily report of vault holdings.

Primary failure modeThe commodity itself moves in price — PAXG tracks gold, not a fixed USD value, so the dollar-denominated price moves with the gold market. Counterparty risk on the vault custodian. Liquidity is meaningfully thinner than for fiat-backed stablecoins.

Where it worksHolders who want gold exposure with crypto-rail composability (lend gold-backed tokens, use as DeFi collateral, transfer cross-border without physical shipping). Inflation hedging when the holder's local currency is volatile and they prefer gold over fiat-backed stablecoins.

Where it does notAnywhere the user case requires a stable dollar value. Commodity-pegged tokens are not interchangeable with USDT or USDC for the typical crypto use case.

Market shareRoughly 1-2% of stablecoin market cap as of mid-2026, dominated by PAXG and XAUT.

The cross-family comparison

Dimension	Fiat-backed	Crypto-backed	Algorithmic	Commodity
Peg target	Single fiat (USD, EUR)	Single fiat (USD, EUR)	Single fiat (historically)	Commodity (gold, etc)
Reserve location	Off-chain, bank custody	On-chain smart contract	None (pure form)	Off-chain, vault custody
Verifiability	Attestations, periodic	On-chain in real time	Not applicable	Attestations, vault audits
Liquidity (2026)	Deepest	Moderate, DeFi-focused	Minimal	Thin
Primary risk	Counterparty bank, regulator	Collateral liquidation, smart-contract	Death spiral	Commodity-price, custodian
Survived 2022-2024 stress	Yes (with one painful exception, USDC in 2023)	Mostly, with DAI inheriting USDC contagion	No (Luna)	Yes, no major depegs
Regulator status	Increasingly tight; MiCA, NY DFS, MAS	Lighter; falls under DeFi frameworks	Excluded from MiCA, GENIUS	Specific commodity-token rules
Typical APY (calm market)	0% (token itself); 3-5% in lending	3-7% (Spark, Aave DAI pool)	Historically 15-20% (Anchor)	0% (gold itself does not yield)
Best for	Working balance, cross-border, first stablecoin	DeFi, philosophy	Research	Gold exposure with crypto rails

The history that explains why the categories settled here

2014-2018 · Fiat-backed dominates from the start

USDT launched in 2014 on the Omni Layer (later migrated to Ethereum, Tron and other chains). For four years it was effectively the only fiat-backed stablecoin with any market depth. The 2017 bull market produced the first major USDT controversies — questions about reserves, the relationship with Bitfinex, the role in BTC price formation. The questions did not slow USDT growth; the absence of alternatives kept the market concentrated.

2018-2020 · USDC enters; DAI grows; the framework forms

USDC launched in 2018 with a more regulated posture, immediately becoming the second-largest stablecoin. MakerDAO launched DAI in late 2017 with single-collateral (SAI), then migrated to multi-collateral (DAI) in November 2019. The framework of "fiat-backed vs crypto-collateralised" became clear; the algorithmic category was small and experimental.

2020-2022 · The algorithmic bubble

Iron Finance (TITAN), Empty Set Dollar, Basis Cash, Frax (initially partially-algorithmic) and several other algorithmic designs reached non-trivial size. Most experienced episodic instability. The defining event was Iron Finance's TITAN crash in June 2021, where the algorithmic stablecoin IRON traded below 0.75 and TITAN fell from 64 to near zero. The collapse was a clear pre-Luna warning; the market did not generalise the lesson.

May 2022 · Luna ends the algorithmic case

UST and Luna collapsed together. 60 billion of combined market cap erased. The contagion took down Three Arrows Capital, Celsius, Voyager and contributed to the FTX collapse later that year. The market for pure algorithmic stablecoins effectively closed. Hybrid designs continued but at lower market cap and with substantial collateral additions.

2022-2024 · Regulators codify what the market already learned

MiCA was passed in 2023 and came into full force in June 2024. The US discussed several stablecoin frameworks (GENIUS, STABLE, Clarity Act). The HK Monetary Authority published its stablecoin licensing framework. All of these excluded or constrained pure algorithmic designs. The category was effectively codified out of the regulated stablecoin space.

2024-2026 · Fiat-backed consolidation

Circle's IPO in June 2024 was the major structural event of this period. USDC moved from a private fintech to a publicly listed company with SEC reporting obligations. Tether continued to publish quarterly BDO attestations and accumulated one of the largest non-sovereign holdings of US Treasury bills. The two-token concentration (USDT + USDC at over 75% of stablecoin market cap) became the working norm.

Examples that do not fit cleanly

USDe (Ethena) and the synthetic-dollar category

Ethena's USDe (launched 2024) is sometimes called a "synthetic dollar". The mechanism uses ETH and BTC perpetual futures shorts to neutralise the USD value of the underlying assets, producing a near-stable token. The yield comes from funding rates on the perp shorts. It is not pure fiat-backed (no off-chain reserves), not crypto-collateralised in the DAI sense, not algorithmic in the UST sense. Some commentators classify it as a fifth category; the desk treats it as a hybrid worth watching but not yet a default holdings category.

Tokenised Treasury products (BUIDL, USDY, OUSG)

BlackRock's BUIDL, Ondo's USDY, Ondo's OUSG and several similar products are tokenised holdings of short-dated US Treasuries. They are not stablecoins in the traditional sense — they pay yield, the on-chain price reflects accrued interest, and the holder is buying a tokenised security rather than a payment token. They are increasingly used as a stablecoin alternative for treasury management; the desk does not classify them as stablecoins for the purpose of this article.

RLUSD (Ripple) and the late-entrant fiat-backed category

Ripple's RLUSD launched in December 2024 as a NY DFS-regulated fiat-backed stablecoin. It sits in the same family as USDC and PYUSD but has narrower venue coverage as of mid-2026. Useful inside the Ripple / XRP Ledger ecosystem; less useful as a primary holding because of liquidity depth.

What the desk uses, by family

To put a working stake in the ground:

Fiat-backed working balance. USDT on Binance for global venue trading, USDC on Coinbase / Kraken for US-regulated venue use. Split is roughly 60/40 USDT/USDC depending on the week.
Crypto-collateralised DeFi position. Small DAI position used in specific Aave borrow contexts and in MakerDAO's PSM module. Not a primary holding.
Algorithmic. Zero. The desk has held no algorithmic position since the Luna collapse. The category is not a holdings category.
Commodity-pegged. Small PAXG position as a gold proxy with crypto-rail composability. Not for stablecoin-equivalent use; classified as a commodity position in the portfolio.

If you only remember three things

Fiat-backed dominates because the trade-off (off-chain custody for on-chain liquidity) is the best fit for the dominant use case (working capital for crypto activity).
The "decentralised stablecoin" debate is partly philosophical: DAI inherits USDC risk through its collateral mix, which means the practical difference between a USDC position and a DAI position is smaller than the marketing suggests.
Algorithmic stablecoins are a category that the market has, in 2026, voted against. The regulators have codified the vote. New designs that claim to solve the algorithmic problem deserve careful reading; the historical track record of the category is not encouraging.

If you want to act on this

For a working stablecoin balance, the operational defaults are USDT on a global venue (Binance most common) or USDC on a US-regulated venue. The Binance referral link opens a registration page pre-filled with the StableDesk referral code BN16188; registering through that link does not change your fees. Full disclosure on the disclaimer page.

The four stablecoin types compared