What is a stablecoin depeg?

A stablecoin depeg is not one event. It is two very different events that look identical on a price chart, and confusing them is how traders lose money on coins that were never actually broken.

The word stablecoin contains a promise, and the promise is arithmetic: one token, one dollar, forever. When the number on the screen reads $0.94, something in that arithmetic has broken. What most people never learn is that two entirely different things break, they produce identical-looking charts, and only one of them is a real emergency. In March 2023, USDC traded as low as 87 cents while Circle’s redemption desk continued honoring dollars at par. In May 2022, TerraUSD traded down through the same levels and never came back, taking roughly $60 billion with it. Same visual, opposite outcomes. Learning to tell them apart in real time is one of the more valuable skills in this market, and it starts with understanding what holds a peg up in the first place.

What a depeg actually is

A stablecoin depeg occurs when the market price of a stablecoin deviates from the value it is designed to track and does not promptly revert. Most stablecoins are soft-pegged, meaning small deviations are expected and normal. A coin trading at $0.998 is not depegged. It is a market with an ordinary bid-ask spread.

The threshold is about degree and duration together. A move to $0.995 for a few minutes during a volatile session is noise, corrected by arbitrage almost immediately. A move to $0.90 that persists for days is a broken mechanism. Between those poles is a spectrum, and where a specific event falls on it depends on how far the price went, how long it stayed, and crucially whether the primary redemption channel kept working.

Depegs run in both directions, which surprises people. A stablecoin can trade above its peg when demand outstrips the supply that can be minted quickly, or when a liquidity pool becomes imbalanced. During the March 2023 turmoil, Tether briefly traded as high as $1.15 as capital fled USDC and Curve’s main stablecoin pool skewed hard. Trading above a dollar is a depeg by definition, and it signals stress just as a discount does.

What holds a peg up

Four mechanisms hold the price near target, and understanding them is what lets you diagnose a break.

Reserves and backing. Each token is supposed to have real value behind it: cash, short-term Treasury bills, repo, crypto collateral, or some mixture. Under the GENIUS Act, signed in July 2025, US payment stablecoin issuers must hold full reserves in liquid assets and disclose their composition monthly. Reserves are the ultimate backstop, but they only defend the peg if holders can actually reach them.

Arbitrage through mint and redeem. This is the real engine. If a token trades at $0.98 and an authorized party can redeem it with the issuer for a full dollar, they will buy every cheap token on the market and redeem for a two-cent profit. That buying pressure lifts the price. The mechanism works precisely as well as the redemption channel does, and no better. Note the constraint: Tether’s minimum redemption is $100,000, which means most holders cannot redeem directly and must sell into the market instead.

Secondary market liquidity. Between arbitrage windows, ordinary exchange liquidity absorbs flow. Deep order books and well-balanced pools soak up sell pressure with minimal price impact. Thin ones do not.

Collateral design. Crypto-backed stablecoins such as DAI overcollateralize, requiring roughly 150% collateral value, with automatic liquidation if it falls below a threshold. Algorithmic stablecoins hold no meaningful collateral and rely instead on minting and burning a paired token to absorb supply and demand.

The four are ranked by fragility. Fiat-backed coins with functioning redemption are the sturdiest. Overcollateralized crypto-backed coins are next, vulnerable when their collateral itself is impaired. Algorithmic designs are the weakest, because their backing is circular: the stablecoin’s value depends on demand for a token whose value depends on the stablecoin.

The distinction that matters

Here is the single most useful idea in this article. When a stablecoin’s exchange price drops, ask one question: is the redemption channel still working?

If it is, you are watching a liquidity depeg. Too many people want out at once through a venue with insufficient depth. The price on that exchange falls because the order book cannot absorb the flow, not because the dollars behind the token vanished. The asset’s redemption value is intact. Arbitrage will close the gap once someone with redemption access shows up.

If it is not, you are watching a reserve depeg. The backing is impaired, inaccessible, or was never sufficient. Nobody arbitrages, because there is no profitable trade in buying a token you cannot redeem for more than you paid. The gap does not close. It widens.

Cain O’Sullivan, co-founder of Hyperdrive, has made this point directly: an exchange price is often not a true representation of the actual redemption value of the asset. The market price and the redemption value are two different numbers, and during stress they diverge violently.

The practical test has three parts. Check whether the issuer’s mint and redeem facility is still operating. Check whether the discount is uniform across venues or concentrated on one. And check whether the reserves are disclosed, liquid, and reachable. A discount on one exchange while the coin trades near par elsewhere is a venue problem. A discount everywhere with redemption frozen is an issuer problem.

The cases worth knowing

TerraUSD, May 2022. The reserve failure. UST was algorithmic, holding no meaningful collateral, and maintained its peg through a mint-and-burn relationship with its sister token LUNA. When confidence broke and holders rushed the exit, the mechanism did what it was designed to do and minted LUNA to absorb the selling. LUNA’s supply exploded from roughly 342 million to about 6.5 trillion tokens, destroying its value, which destroyed the only thing backing UST. Around $60 billion of value was erased and the broader market lost several hundred billion more. The US Treasury Secretary cited it as evidence of rapidly growing risks in the sector. This is the textbook reserve depeg: no collateral, circular backing, no recovery.

USDC, March 2023. The liquidity depeg that looked fatal. Circle disclosed that $3.3 billion of its roughly $40 billion in reserves was trapped at the collapsed Silicon Valley Bank. USDC fell to around $0.87. DAI depegged alongside it, because USDC made up over half of DAI’s collateral, a clean illustration of contagion through collateral chains. But the reserves were real. When the FDIC announced a systemic risk exception covering the bank’s depositors, the peg restored fully. Anyone who panic-sold at $0.88 realized apermanent loss on an asset that was solvent the entire time.

Tether, June 2023. The pool imbalance. USDT slipped to about $0.977 when its share of Curve’s main stablecoin pool ballooned past 70% against a balanced target near a third. Nothing about Tether’s reserves changed. The pool became lopsided, and a lopsided pool prices the overweight asset down. Kaiko noted it looked like a possible deliberate attempt to break the peg.

USDR, October 2023. Collateral that could not be sold. Real USD faced roughly $10 million in redemption requests that drained its liquid DAI reserves, leaving collateral consisting largely of tokenized real estate. The assets existed. They could not be liquidated fast enough to meet redemptions. Illiquid backing is functionally the same as absent backing during a run.

USDe, October 2025. The oracle depeg. During a market-wide liquidation event of roughly $19 billion, Ethena’s USDe briefly printed 65 cents on Binance while trading at near parity on decentralized venues such as Curve. The protocol was around 110% collateralized and its design held. What failed was Binance’s internal oracle and order book depth. The 65-cent print was a venue artifact, not a valuation.

xUSD, November 2025. The disclosure shock. The token fell roughly 77% in a single day after its issuer disclosed a $93 million loss tied to one external fund manager. It had traded at a dollar until the moment the mechanism behind it was revealed to be impaired.

How a depeg spreads

Depegs rarely stay contained, and the transmission runs through three channels.

Collateral chains. When one stablecoin backs another, impairment travels. DAI followed USDC down in 2023 for exactly this reason. Today’s version of the question is which assets sit inside which reserves, and tokenized money market funds have begun appearing in stablecoin reserve baskets, which creates new linkages worth tracking.

Oracle-driven liquidation. This is the most destructive channel and the least understood. DeFi lending protocols price collateral using oracles. If an oracle reports the market price of a stablecoin at $0.90, the protocol sees undercollateralized loans and liquidates them, even when the stablecoin is fully backed and will be at par within hours. Those liquidations dump more supply, pushing the price lower, triggering more liquidations. O’Sullivan describes the result as a death spiral in an asset that never actually failed. The fix is pricing collateral by redemption rate instead of by market quote, anchoring valuation to actual backing.

Reflexive panic. A depeg is a bank run in public. Once the price prints below a dollar, holders who never thought about reserves start selling, which pushes it further below, which convinces more holders. In 2022, Tether processed over $13 billion in redemptions within a week as fear spread from Terra, and Tether was not the coin that failed.

Why the stakes keep rising

The stablecoin market now exceeds $300 billion, with the largest coins holding close to 90% of supply, and the sector functions as the settlement layer for most crypto trading. That combination of scale and concentration is the systemic worry. A depeg of a minor token is a bad day for its holders. A depeg of a dominant one is a repricing of the entire market’s unit of account, because trading pairs, collateral, and DeFi accounting are all denominated in it.

The GENIUS Act addresses one half of the problem well and the other half not at all. Full liquid reserve requirements and monthly disclosure make reserve failures materially less likely for compliant US issuers, which is the half that killed Terra. It does nothing about liquidity depegs, oracle mispricing, or thin pools on individual venues, which is what produced the USDC, USDT, and USDe episodes. The likeliest future depeg is therefore not a collapse. It is a solvent coin printing a scary number on one exchange while a protocol somewhere liquidates people who did nothing wrong.

The three families, ranked by how they break

Because the failure mode depends almost entirely on the design, it is worth walking each family and naming exactly what kills it.

Fiat-backed coins. USDC and Tether are the archetypes. Each token is meant to correspond to a dollar or dollar-equivalent held off-chain in cash and short-term Treasuries. The strength is obvious: the backing is boring, liquid, and reachable. The weakness is that it lives in the traditional banking system, which introduces a counterparty crypto cannot audit. USDC did not depeg in 2023 because of anything on-chain. It depegged because a bank in California failed. That is the structural exposure of the entire category: the safest stablecoin design is only as safe as the least safe bank holding its reserves. The second weakness is redemption access. If direct redemption carries a $100,000 minimum, as Tether’s does, then the arbitrage that defends the peg is available to a small set of large players, and everyone else is a price-taker on the secondary market. The peg is defended on your behalf by people whose interests may not align with yours during a panic.

Crypto-backed coins. DAI is the reference. Collateral sits in smart contracts, overcollateralized at around 150%, with automatic liquidation if the ratio breaks. The strength is transparency: you can verify the collateral yourself, on-chain, right now, which is impossible with a bank account. The weakness is that crypto collateral is volatile, and a market crash triggers mass liquidations exactly when liquidity is thinnest. The subtler weakness is composition. DAI followed USDC down in March 2023 because USDC made up more than half its collateral. A crypto-backed coin backed substantially by a fiat-backed coin inherits every risk of the fiat-backed coin and adds liquidation mechanics on top. Verifying that collateral exists is not the same as verifying it is uncorrelated.

Algorithmic coins. TerraUSD is the tombstone. No meaningful collateral, with stability maintained by minting and burning a paired token. The design’s appeal was that it scaled without needing reserves, which is another way of saying it produced dollars from confidence. The failure is not a bug; it is the mechanism working as specified. When holders sold, the protocol minted LUNA to absorb them, which diluted LUNA, which reduced the value of the only thing backing UST, which prompted more selling. The technical term is a death spiral and the plain term is that the backing was circular. Every purely algorithmic design shares this property: the collateral’s value depends on demand for the stablecoin whose value depends on the collateral. Under stress, both variables go to zero together.

The practical hierarchy that falls out is unromantic. Fiat-backed coins with transparent reserves and functioning redemption fail rarely and recover when they do. Crypto-backed coins fail more often, recover usually, and inherit whatever their collateral is exposed to. Algorithmic coins fail rarely and terminally. The frequency ranking and the severity ranking run in opposite directions, which is exactly why the category confuses people: the coins that wobble most are the ones most likely to come back, and the one that never wobbled until the day it died is the one that took $60 billion with it.

Reading a depeg without panicking

If you hold stablecoins, the practical guidance is unglamorous and short.

Know what backs your coin, and whether that backing is liquid. Cash and short-dated Treasuries are reachable in a crisis. Tokenized real estate is not. Algorithmic backing is not backing.

Watch redemption, not price. The exchange quote is the noisiest signal available during stress. Whether the issuer is still honoring redemptions at par is the informative one.

Compare across venues. A discount on one exchange and parity elsewhere is a liquidity event. A uniform discount is a solvency question.

Do not sell into a liquidity depeg on reflex. Every holder who dumped USDC at $0.88 in 2023 paid twelve cents for the privilege of being wrong about a solvent asset. Equally, do not hold an algorithmic coin through a break on the theory that it recovered last time, because Terra did not.

Understand that spreading holdings across several stablecoins reduces single-issuer exposure and does nothing about correlated collateral. If three coins all hold the same asset in reserve, you own one risk wearing three names.

The uncomfortable truth is that a depeg is a test that reveals what a stablecoin always was. The peg was never a property of the token. It was a claim about a mechanism, and stress is the only thing that ever checks whether the claim was true.

Disclaimer: This article is for information and educational purposes only and does not constitute financial or investment advice. Stablecoins carry issuer, reserve, regulatory, and liquidity risk, and past recoveries do not indicate future ones. Nothing here is a recommendation to buy, hold, or sell any asset. Always do your own research. Information is accurate as of July 16, 2026.

Frequently Asked Questions