Stablecoins: what actually backs the big ones

A stablecoin holds its value by some mechanism, and the phrase “backed one to one” describes at least three completely different arrangements — which differ in what the backing asset is, who holds it, and whether you personally can exchange the token for it.

Three structures

Fiat-reserved: the reserve is the product

The largest stablecoins work this way. A company takes dollars, issues tokens, holds the dollars in some form, and destroys tokens when they come back.

The design is simple. The risk is entirely in the details of the reserve, and those details vary far more than the shared “one to one” phrasing suggests.

What the reserve is actually invested in

“Backed by dollars” almost never means a vault of banknotes. It usually means a mixture of bank deposits and short-term government securities, and it can include other instruments.

Those choices matter. Short-dated Treasury bills are about as liquid and low-risk as financial assets get; they can be sold quickly at close to face value. Bank deposits are a claim on a bank and carry that bank's credit risk, above whatever deposit insurance covers — which, at the scale a stablecoin issuer operates, is a small fraction. Longer-dated or less liquid instruments introduce the possibility of having to sell at a loss if redemptions arrive faster than expected.

The two records in this archive show the range. USDC's reserves sit mainly in an SEC-registered government money market fund managed by a third party, with the remainder as bank deposits. Tether publishes a quarterly composition of a reserve pool it manages itself. Both meet the description “fiat-reserved”. They are structurally different positions.

Who is holding it

A reserve managed by an independent asset manager inside a regulated fund structure has different properties from one held directly by the issuer. Neither is automatically safe. What differs is who can do what with the assets, what reporting obligations attach, and how the assets would be treated if the issuer got into trouble.

Attestation and audit, again

Because it comes up on every stablecoin page ever written: an attestation is an accountant reporting on figures management prepared, as at a date, within a scope set by the engagement. A financial statement audit is an opinion on a complete set of accounts over a period, under auditing standards. Most stablecoin reserve reporting is the former.

That is not an accusation. It is a description of what the document is, and a reader who thinks they are getting an audit when they are getting an attestation has misunderstood their position. The Tether record sets this out in detail.

Overcollateralised: the buffer is the product

The second family puts the collateral on chain. To create a dollar's worth of the token, you lock more than a dollar's worth of crypto — often substantially more — in a contract. If the collateral value falls toward the debt, the position is liquidated automatically and the tokens are bought back.

What you gain is verifiability. You do not need to trust a report; you can read the contract and see exactly what is locked, right now. That is a genuine advantage and it is not available in the fiat model at all.

What you take on instead:

• Collateral volatility. The buffer exists because the collateral moves. In a sharp fall it can move faster than liquidations execute.
• Liquidation mechanics under stress. Liquidations need buyers, chain capacity and working price feeds. All three are least reliable exactly when they are most needed.
• Oracle dependency. The system needs to know what the collateral is worth. That comes from a price feed, which is a dependency — see the record on how those feeds actually work, including the fact that they update on thresholds rather than continuously.
• Capital inefficiency. Locking well over a dollar to issue a dollar limits how large the system can get relative to demand.

A practical note: several designs in this family have gradually added fiat-reserved stablecoins to their own collateral, which means the on-chain verifiability increasingly rests on an off-chain reserve somewhere underneath. Worth checking what a given system's collateral actually consists of rather than assuming it is all crypto.

Algorithmic: the incentive is the product

The third family holds no meaningful backing. It maintains the peg through a mechanism — typically minting and burning a second, floating token to absorb pressure on the first.

In calm conditions this works, and it works with far less capital than the other two designs need. That efficiency is the appeal.

The failure mode is that the mechanism is reflexive. The stablecoin's peg depends on the absorbing token having value; that token's value depends on confidence in the system; confidence depends on the peg holding. When the peg slips, the mechanism issues more of the absorbing token, diluting it, which reduces its value, which weakens the peg further. The loop runs in the wrong direction and it runs fast.

This is not a theoretical concern. Designs of this type have collapsed, in one case erasing a very large amount of value over a few days in 2022, and taking a good deal of the surrounding market with it. Regulatory frameworks written since have generally treated uncollateralised designs as a distinct and higher-risk category.

Some current designs sit between families — partially collateralised, or backed by delta-neutral positions rather than by reserves. Those are their own structures with their own dependencies, and the right question for any of them is the same as for the rest: what specifically is behind it, who holds that, and what happens when the thing it depends on stops behaving?

The question nobody asks: can you redeem?

Here is the part that most stablecoin explainers omit entirely, and it is the part that decides whether backing matters to you personally.

A peg holds in normal markets because of arbitrage. If the token trades at 99 cents, someone buys it cheaply, redeems it with the issuer for a dollar, and pockets the difference. That buying pressure pushes the price back. The reserve is what makes the arbitrage profitable.

But this only works if redemption is actually available. And for most fiat-reserved stablecoins, direct redemption is not available to ordinary holders. It requires an account with the issuer, identity verification, and often a minimum size well beyond what an individual holds.

So the practical position for most people is: you cannot redeem, and you are relying on professional participants who can. Your exit is the secondary market — an exchange — and your price is whatever the order book says.

That has two consequences worth internalising. In normal conditions it makes no difference; the arbitrage works and the peg holds. In stressed conditions, when everyone wants out at once, the people who can redeem have an exit that you do not, and the gap between the reserve's value and your realisable price is the market's liquidity, not the issuer's backing.

None of which means a well-backed stablecoin is no better than a poorly backed one. It means the backing protects the instrument, and market liquidity protects you, and they are not the same thing.

How a depeg actually happens

Depegs follow a small number of patterns.

The reserve is impaired. Something the reserve holds loses value or becomes inaccessible. In March 2023, USDC traded below a dollar after part of its reserve was disclosed as held at a bank that failed. The funds were ultimately recovered and the peg restored, which makes it a clean illustration of the mechanism: the token followed the reserve's perceived quality, then followed it back.

The mechanism unwinds. An algorithmic design enters its reflexive loop. Fast, and historically not recoverable.

Collateral falls faster than liquidations run. The overcollateralised failure mode, usually during a sharp market-wide fall.

Liquidity dries up without anything being wrong. Large sell pressure meets thin order books and the price moves even though the backing is intact. This resolves once arbitrage catches up, and it is the most common kind of small depeg.

Reading which one is happening in real time is difficult, and the honest advice is that if you cannot tell, you probably cannot act on it usefully either.

What the rules now require, and what they do not

Several jurisdictions have brought stablecoins inside financial regulation over the past few years, and the frameworks share a recognisable shape even where the detail differs.

The common requirements: reserves held one-for-one, segregated from the issuer's own money, restricted to short-dated liquid assets, held with regulated custodians, reported on regularly, and redeemable at par by holders. Issuers generally need authorisation. Uncollateralised algorithmic designs have mostly been treated as a separate and higher-risk category, and in some places effectively excluded.

What that changes for a holder is real but narrower than it sounds. A regulated issuer has constraints on what it may do with your money and a supervisor who can ask questions. That is a genuine improvement over an unconstrained one.

What it does not do:

• It does not make the token a bank deposit. Deposit insurance schemes generally do not cover stablecoin holdings, even where the reserve sits at an insured bank — the insurance follows the account holder, which is the issuer, not you.
• It does not guarantee the peg. Reserve rules reduce the chance of a shortfall. They do not create a buyer at par when the market is short of them.
• It does not apply everywhere. Rules are jurisdictional, and a token issued outside a given framework may still be freely tradeable there.
• It does not solve redemption access. A right to redeem at par is only as useful as the operational route to exercise it, which for most retail holders remains an exchange rather than the issuer.

The practical reading: regulation moves the risk from “is the reserve there at all” toward “can I get to it when I want to”. That is a better problem to have and it is still a problem.

How to read an issuer's disclosures

A short checklist to apply to any stablecoin's transparency page.

1. What exactly does the report cover? The reserve, or the entity? As at a date, or across a period?
2. Who prepared it, and under what standard? Attestation standards and auditing standards are different bodies of rules.
3. How often? Monthly, quarterly, weekly holdings. More frequent is more informative.
4. What is the reserve invested in? Not the headline category, the actual instruments and their maturities.
5. Where is it held? A fund, a bank, the issuer's own accounts. Segregated from operating funds or not.
6. Who can redeem, at what minimum? This determines whether the arbitrage that holds the peg has enough participants.
7. Can the issuer freeze balances? Most can. Know whether yours does.

If you have worked through that and are ready to actually acquire some — including the perfectly reasonable route of buying a stablecoin first and deciding what to do next later — the account, the verification and the pair mechanics are covered in the step-by-step on opening an account.

Primary references

• The EU Markets in Crypto-assets Regulation for reserve, custody and redemption requirements.
• The US GENIUS Act for permitted issuers, one-to-one reserves and redemption rules.
• The UK FCA's cryptoasset regime overview for stablecoin backing, safeguarding and redemption rules.

This guide describes structures rather than reporting current reserve figures. The March 2023 USDC episode is public record. Specific issuer arrangements are set out with their sources on the Tether and USDC records, both checked August 2026. For current composition, read the issuer's own report rather than any article.

Related: the Tether record · the USDC record · how supply figures work generally