Most traders treat a stablecoin as a resting place, the thing you sit in between positions. That is a reasonable use and it quietly involves holding a credit instrument issued by a private company. Understanding what keeps the number at one is the difference between a considered decision and an assumption.
A stablecoin unit is a liability of whoever issued it, recorded on a public ledger. Holding one means holding a promise that it can be exchanged for something worth a defined amount, and the strength of that promise is the strength of the issuer and its collateral. Nothing about the technology changes this: the ledger records who holds the claim, it does not guarantee the claim.
This is not a criticism, it is a classification, and getting it right reorders the risks. The relevant questions become the ones asked of any credit instrument. What backs it, where is that collateral held, who can seize or freeze it, what happens if the issuer fails, and who checks. A holder who has not asked these is exposed to all of them anyway.
The technology settles the transfer. It does not create the value. What creates the value is a promise made by an identifiable party.
Nothing in the code fixes the market price at one. What fixes it is that when the market price falls below one, somebody can buy units cheaply and redeem them with the issuer for the full amount, pocketing the difference. That buying pressure is what pushes the price back.
Every part of that sentence is a condition rather than a guarantee. Redemption has to be available, in size, quickly, and to the participant who noticed the gap. When any of those fails, the mechanism that restores the peg is not weakened, it is absent, and the price is then set by whatever people will pay for a claim they cannot currently cash.
This explains an observation that confuses newcomers: the same discount can be trivial or terminal depending on why it appeared. A small gap while redemptions are flowing normally is arbitrage friction. The same gap while redemptions are paused is the market pricing a claim, and the two look identical on a chart.
The dominant model holds reserves in bank deposits and short-term government securities, and issues units against them. It is the simplest design and the one whose risks are the best understood, because they are the ordinary risks of a money market portfolio plus the ordinary risks of a bank.
The concentration of those risks is where the interesting part sits. Reserves held as deposits are exposed to the specific banks holding them, and a stablecoin has previously traded below its peg because a bank holding part of its reserves failed over a weekend, entirely independently of anything the issuer had done. Short-term government paper is exposed to interest rates and to settlement, which are smaller risks but not zero ones.
Most issuers publish periodic attestations rather than full audits, and the distinction matters. An attestation confirms that stated assets existed at a stated moment, examined under a limited scope agreed with the issuer. It is meaningfully better than nothing and it is not an audit of the entity, its liabilities, or the periods between observations.
A second model issues units against volatile collateral locked in a contract, holding substantially more value than it issues. If the collateral falls in value beyond a threshold, positions are liquidated automatically to restore the ratio. The design accepts that the backing moves and compensates with a cushion.
It removes the bank and replaces it with two different exposures. The first is the collateral itself, which can fall faster than liquidations can execute, which is precisely what happens during the market conditions that make liquidations necessary. The second is the code: the contracts that hold the collateral and run the liquidations are the system, and a defect in them is not recoverable by a support conversation.
The honest summary is that this model trades counterparty risk for market and technical risk. Whether that is an improvement depends entirely on which of the three you are better placed to assess, and most holders are not better placed to assess any of them than the others.
A third family attempted to hold the peg with no meaningful collateral at all, using a mechanism that created and destroyed a companion token to absorb imbalances. When the price fell below one, the system issued more of the companion token to buy back units, betting that demand for the companion would absorb the supply.
The failure mode was visible in advance and it arrived exactly as described. Confidence in the companion token depended on the peg holding, and the peg depended on demand for the companion token. When both fell together the system issued increasing quantities into a collapsing market, and the reflexivity that had supported the design on the way up worked identically in reverse and much faster.
What survives is not a rule about one design but a test applicable to all of them. Ask what backs the unit when confidence is gone, because that is the only condition under which the backing is ever tested. A mechanism whose collateral is its own reputation has nothing behind it at exactly the moment something is needed.
Ask what backs a unit on the worst day, not on an ordinary one. Every design works while nobody is asking.
They are worth separating because they resolve very differently, and the chart looks the same for all four in the first hour.
Something backing the units is worth less than assumed, whether a failed bank, a fallen asset, or a security that cannot be sold at its marked price. This is the most fundamental case and the slowest to resolve, because it requires the shortfall to be made up from somewhere.
The collateral is intact but the path to it is closed: a banking partner is unavailable, a jurisdiction has intervened, or the issuer has paused processing. The arbitrage that restores the peg cannot run, so the discount widens even though nothing is wrong with the backing.
The peg holds everywhere except one market where a large seller has consumed the book. This is a local execution event rather than a depeg, and it is routinely reported as one. The tell is that other venues are unaffected.
Holders sell because other holders are selling. It is self-fulfilling in the short term and self-correcting when the backing turns out to be intact, which is why several historical depegs recovered fully within days and a few never did.
Traders think of stablecoin risk as something that applies to holdings, and it applies to positions too. If your collateral, your quoted prices and your settlement all happen in the same unit, then a movement in that unit is a movement in everything you own simultaneously, including the value of the margin supporting your open positions.
This matters most at the point where it is least visible. A pair quoted against a stablecoin does not show you the stablecoin's price. If it moves against its peg, a position that appears unchanged has changed, and a margin buffer sized in that unit is a different buffer than it was yesterday.
The practical consequence is not to avoid stablecoins, which would be impractical and would substitute worse risks. It is to know which unit you are denominated in, to avoid holding all of the balance in one issuer if the amount matters to you, and to treat the settlement asset as a position rather than as a neutral background.
A stablecoin that pays a return is paying it from something, and identifying the source is the entire analysis. Broadly there are three sources, and they carry very different risks under the same headline number.
The first is the issuer's own reserves: government paper earns interest and some issuers share part of it. This is the most transparent source and its return is bounded by prevailing short-term rates, so a yield well above them is not coming from here. The second is lending: your units are lent to somebody who pays for them, which means you hold their credit risk and the risk of whatever collateral secures the loan. The third is trading strategies executed with the deposits, which is the least transparent and the one where an advertised rate reveals the least about what is happening.
The useful discipline is to refuse to compare rates that come from different sources. A number produced by holding government debt and a number produced by lending to leveraged traders are not the same product with different pricing, and a comparison that puts them side by side has already lost the important information.
Five questions cover almost everything, and none requires specialist knowledge. What is the collateral, precisely, and not in a marketing summary. Who holds it and in which jurisdiction. Who can redeem, at what size, and under what conditions can that be suspended. Who examines the reserves, how often, and is it an audit or an attestation. And what has historically happened to this unit under stress, because most of them have been tested at least once.
None of this makes stablecoins a bad instrument. They are the settlement layer that makes the rest of crypto trading practical, and the alternatives all involve either more counterparty exposure or considerably more friction. The point is narrower: a holding you have not classified is a holding whose risks you have accepted without listing them, and the list here is short enough to work through once.
Stablecoins are not the risk-free part of a portfolio. They are the part where the risk is easiest to overlook because the number does not move.
No. It is a claim on an issuer, recorded on a ledger, backed by whatever that issuer holds. Bank deposits carry deposit insurance in most jurisdictions and a regulatory framework built over a century; a stablecoin unit generally carries neither. The convenience is real and the instrument is a different one.
Arbitrage. When the market price falls below one, somebody buys cheaply and redeems with the issuer for the full amount, and that buying restores the price. The mechanism requires redemption to be available, fast and at size. Where it is not, nothing pushes the price back and it is set by what people will pay for a claim they cannot cash.
Because they start differently. A local liquidity event on one venue or a wave of selling with the backing intact resolves quickly, because the arbitrage still works. A genuine shortfall in collateral or a suspension of redemption does not resolve until the underlying problem does, and sometimes it never does.
No, and the difference is substantive. An attestation confirms that stated assets existed at a stated moment under a scope agreed with the issuer. An audit examines the entity, its liabilities and its controls over a period. Most stablecoin reserve reporting is the first, which is better than nothing and is regularly described as though it were the second.
They are differently exposed rather than safer. They remove the bank and the issuer's balance sheet, and replace them with the volatility of the collateral and the correctness of the contracts. Whether that is an improvement depends on which risks you can actually assess, and for most holders the answer is neither.
It reduces concentration on a single issuer and increases the number of issuers you are exposed to, so it is a real trade rather than a free improvement. For balances large enough to matter it is generally worth doing, on the reasoning that the failure modes are issuer-specific rather than shared.
One of three places: the issuer sharing interest earned on reserves, lending your units to a borrower, or trading strategies run with the deposits. The first is bounded by short-term rates, so anything well above them is coming from one of the other two. A rate quoted without naming its source is missing the only information that matters.
Yes, if your margin and your quoted prices are denominated in it. A pair quoted against a stablecoin does not display the stablecoin's own price, so a position that looks unchanged has changed, and the buffer supporting it is worth something different. The settlement asset is a position, not a neutral background.