A side-by-side breakdown of how fiat-backed, crypto-backed and algorithmic stablecoins each maintain their peg, and the specific failure mode that has taken each category down.
Nadia OkoroPolicy & Consumer Editor · Regulation, tax, stablecoins and the products retail users hold
Fiat-backed stablecoins hold cash and short-dated government paper and depend on the issuer and its custodians. Crypto-backed stablecoins are over-collateralised on-chain and depend on liquid collateral and working liquidations. Algorithmic designs rely on market incentives rather than reserves, and have failed repeatedly because the mechanism prints more of a falling asset exactly when confidence goes.
Stablecoins now settle more on-chain volume than Bitcoin and Ethereum combined in most months, yet the word "stablecoin" covers three fundamentally different engineering approaches with very different failure modes. Fiat-backed, crypto-backed and algorithmic stablecoins all promise a dollar peg, but they get there through mechanisms that trust different things: a bank, a smart contract, or a market's collective behaviour. Knowing which is which matters far more than most holders realise, usually right up until the peg breaks.
Tether (USDT) and Circle's USDC are the two dominant examples: for every token in circulation, the issuer claims to hold an equivalent dollar of reserves in cash, short-term Treasuries, or similar instruments, sitting in a bank or custodian somewhere off-chain. The mechanism is conceptually simple, which is exactly why it scaled fastest. The trust, however, is entirely off-chain and entirely dependent on the issuer actually holding what it says it holds, disclosing it honestly, and staying solvent.
This is where the two diverge in reputation. Circle publishes monthly attestations from a major accounting firm and holds reserves predominantly in short-duration Treasuries and cash at regulated banks. Tether has a longer history of opacity, including a 2021 settlement with the New York Attorney General over misrepresenting its reserves, though its disclosures have improved since. The risk with fiat-backed stablecoins isn't smart contract risk, it's counterparty and regulatory risk: a frozen bank account, a regulatory crackdown, or a reserve composition that turns out to include more risk than advertised.
MakerDAO's DAI (now largely folded into Sky) takes a different approach: instead of trusting a bank, it trusts overcollateralized crypto locked in transparent, auditable smart contracts. Users deposit ETH or other approved assets worth more than the DAI they mint, typically 150% or higher, and the entire reserve is visible on-chain in real time rather than disclosed monthly by an accountant. Nobody can quietly rehypothecate the collateral or lie about the reserve ratio, because anyone can check it themselves.
The trade-off is that the collateral itself is volatile, so the system needs a much larger buffer than a fiat-backed stablecoin does, and it needs automated liquidations to maintain that buffer as prices move. This is more complex, more gas-intensive, and historically has needed to lean on centralized stablecoins like USDC as part of its own collateral mix to stay liquid during stress, which somewhat undercuts the "fully decentralized" pitch. It's a genuine improvement in transparency over fiat-backed models, but not a clean escape from centralization risk.
Algorithmic stablecoins try to maintain the peg with no full collateral backing at all, instead using a second token, arbitrage incentives, and market confidence to keep supply and demand in balance. Terra's UST was the largest and most consequential example, using a mint-and-burn relationship with its sister token LUNA: when UST traded above a dollar, arbitrageurs could burn LUNA to mint UST for profit, and vice versa when it traded below.
That mechanism worked beautifully in calm markets and catastrophically in a run. In May 2022, large UST withdrawals pushed the peg down, arbitrageurs minted UST by burning LUNA in size, LUNA's price collapsed under the new supply, and the collapsing LUNA price destroyed the very incentive meant to restore the peg. Roughly $40 billion evaporated in a matter of days, in what's now the textbook case study in every discussion of algorithmic stablecoin design. The death spiral wasn't a bug; it was the mechanism working exactly as designed, just in the direction nobody had priced in.
The post-Terra generation has mostly avoided pure algorithmic designs. Ethena's USDe, for instance, maintains its peg through a delta-neutral hedge, holding staked ETH as collateral while shorting an equivalent amount via perpetual futures, funded partly by the yield those perpetuals generate. It's not fiat-backed, not simply overcollateralized in the DAI sense, and not algorithmic in the Terra sense; it's a fourth category that trades reserve simplicity for exposure to funding rates and derivatives market plumbing. Meanwhile, tokenized-Treasury stablecoins are pulling fiat-backed models further into regulated, yield-bearing territory, blurring the line between a stablecoin and a money market fund.
Fiat-backed stablecoins fail through counterparty and regulatory events: a bank freeze, a reserve fraud, a jurisdiction banning redemptions. Crypto-backed stablecoins fail through collateral shocks that outrun the liquidation engine, or through creeping reliance on the very centralized assets they were built to avoid. Algorithmic stablecoins fail through reflexivity, the peg-defence mechanism and the token's price being tied together so tightly that a loss of confidence becomes self-reinforcing rather than self-correcting.
None of this makes one category strictly safer than another in every scenario; it makes them differently exposed. The practical question for anyone holding a stablecoin isn't "is it pegged today" but "what specific event would break this peg, and how would I know before it happened." For fiat-backed tokens, that means reading the attestations. For crypto-backed tokens, it means checking the collateralization ratio and collateral mix. For anything resembling an algorithmic design, it means asking what happens to the peg mechanism itself during a genuine loss of confidence, because that's the one scenario the model is least likely to survive.
An accessible breakdown of how fiat-backed, crypto-backed and algorithmic stablecoins each try to hold their peg — and the real failures that show what happens when they don't.
No order book, no market makers standing by — just a formula and a pool of reserves. Here's the mechanics behind the constant-product curve that prices most of DeFi.
Cross-chain bridges move billions between blockchains every week — and have lost billions more to exploits. Here's how the plumbing actually works and how to size up the risk before you use one.