What a Stablecoin Is, and Why the Peg Holds

A stablecoin is a digital token designed to hold a constant value, almost always one US dollar. The idea exists to resolve a specific conflict: blockchains offer settlement that is fast, global, and running every hour of every day, but their native assets swing wildly in price. Nobody invoices a supplier in something that might drop 15 percent before it arrives.

A stablecoin is the attempt to keep the rails and drop the volatility: a dollar-denominated unit that moves like software.

That framing sets up the whole cursus. This lesson answers the mechanical question, why does it hold a dollar? Lesson 2 asks what businesses actually do with it. Lesson 3 covers the rules and the risks.

Start with the definition that now carries legal weight in the US. The GENIUS Act of 2025 defines a payment stablecoin as a digital asset issued for payment or settlement and redeemable at a predetermined fixed amount, for example one dollar.

Read that last clause carefully, because it is the entire mechanism in three words. Redeemable at par. A stablecoin is not stable because someone declares it stable, or because software says so. It is stable because it is a claim, and the claim can be exchanged for a real dollar. Everything in this lesson follows from that, including how pegs break.

How does a token trade at exactly one dollar on an open market where nobody controls the price? Not by decree. By arbitrage, and the loop is worth walking precisely because every peg failure is a failure of this loop.

The issuer offers two standing services to authorized parties:

• Mint: give the issuer one dollar, receive one new token.
• Redeem: give the issuer one token, receive one dollar.

Now watch what a profit-seeker does when the market price drifts.

If the token trades at 0.998: buy it on the market for 0.998, redeem it with the issuer for a full 1.00, and pocket 0.002 per token. Doing this buys the cheap tokens (pushing the price up) and destroys supply on redemption (pushing it up further). Under normal conditions this closes the gap within minutes.

If the token trades at 1.002: the reverse. Give the issuer 1.00, mint a token, sell it for 1.002. This adds supply and sells into the demand, pushing the price back down.

Notice what actually enforces the peg: greed, not goodwill. Nobody maintains the price out of duty. The issuer simply stands ready to swap at par, and the profit motive does the rest, automatically, in both directions.

Which leads to the sentence that predicts every depeg in history: the peg holds exactly as well as redemption does. Break redemption and the arbitrage that enforces the price has nothing to work with.

If the peg depends on redemption, then it depends on the issuer's ability to redeem, which depends entirely on what the issuer is holding. That is the reserves, and their composition is not a technical footnote, it is the product.

The question to ask of any stablecoin is blunt: if every holder redeemed tomorrow, could the issuer pay? That has nothing to do with blockchains and everything to do with an old, well-understood problem, whether an institution's assets can meet its liabilities on demand.

This is why the GENIUS Act's core requirement is so specific. Issuers must hold at least one dollar of permitted reserves for every one dollar of stablecoins issued, and it enumerates what counts:

• Coins and currency
• Deposits at insured banks and credit unions
• Short-dated Treasury bills
• Repurchase agreements and reverse repos backed by Treasury bills
• Government money market funds
• Central bank reserves
• Similar government-issued assets approved by regulators

Look at what that list is engineered for. Every item is short-dated, highly liquid, and government-backed. No corporate bonds, no long-dated paper, no lending the reserves out to earn a spread. The reason is the redemption promise: a reserve that cannot be turned into cash today, at face value cannot honor a claim today.

So the deep truth about stablecoins is unglamorous. A stablecoin is a narrow, fully-reserved liability that happens to settle on a blockchain. The token is the interesting part; the peg is banking.

Not every stablecoin backs its claim the same way, and the three designs have very different risk.

Fiat-backed is the mainstream design and the one the GENIUS Act regulates: a real dollar sits somewhere for every token.

Crypto-collateralized avoids a bank by locking volatile crypto worth more than the tokens issued, so the buffer absorbs price falls, with automatic liquidation if it thins. It trades issuer risk for market risk, and it is capital-inefficient by design.

Algorithmic tried to hold a dollar with no external backing at all, using a mint-and-burn relationship with a second, floating token. The claim was that arbitrage alone could substitute for reserves.

Re-read the previous step and you can see the flaw before it happens. If the peg holds because you can redeem for something real, what happens when there is nothing real to redeem for? The arbitrage loop depends on the paired token having value, and that token's value depends on confidence in the peg. It is circular. Confidence is the collateral.

The algorithmic design was tested at scale and failed completely. TerraUSD (UST) collapsed in May 2022, wiping roughly 40 billion dollars across the Terra ecosystem in about 72 hours.

The mechanism deserves precision, because its failure was structural rather than accidental. UST held its peg through arbitrage with a paired floating token, LUNA: you could always burn one dollar of LUNA to mint one UST, and burn one UST to mint one dollar of LUNA. In calm conditions that loop works exactly like the mint-and-redeem arbitrage from earlier, and it looks like it needs no reserves at all.

The fatal detail is what backs the loop. Redeeming UST does not produce a dollar, it produces more LUNA, which must then be sold for dollars. So the system's real backing is LUNA's market value, and that value rests on confidence in the system. The collateral and the thing being collateralized are the same belief.

So the unravelling was mechanical. UST slips below the peg. Holders redeem into newly minted LUNA. That new supply is dumped, pushing LUNA's price down. A lower LUNA price means less backing, which triggers more redemption, which mints even more LUNA. The loop that defended the peg upward becomes the engine destroying it, and it accelerates. When LUNA's market cap fell below UST's circulating supply, the bonding mechanism had, arithmetically, nothing left to redeem into.

The lesson generalizes past crypto: a promise to pay is only as good as the asset behind it. Arbitrage distributes confidence; it cannot create it.

Terra proved algorithmic backing fails. A second case proved something subtler and more useful: even a fully-reserved stablecoin depegs if redemption stops working.

In March 2023, USDC held a portion of its reserves in deposits at Silicon Valley Bank, which failed. The exposure was a minority of reserves and was ultimately recovered in full. USDC was, in the end, solvent.

It depegged anyway, trading as low as roughly 88 cents.

The reason is the exact mechanism from earlier. The bank failed on a Friday, and redemption was paused over the weekend. With redemption unavailable, the arbitrage loop could not operate. Nobody could buy discounted USDC and redeem it for a full dollar, because the redemption window was shut. The only price left was whatever panicked sellers would accept in a thin weekend market. When banking hours resumed and redemption reopened, the peg recovered.

Sit with what that demonstrates. The reserves were fine. The issuer was solvent. The design was sound. And the peg still broke for a weekend, purely because the mechanism that enforces the peg was temporarily unavailable.

So a stablecoin's peg depends not only on having reserves but on the reserves being reachable, right now, which means it inherits the operating hours, counterparty risk, and fragility of the banking system it sits on. A token that settles 24/7 is backed by a system that does not. That gap is precisely where the peg lives, and Lesson 3 returns to it as a live risk.

Put the cases side by side, because together they define the entire risk surface.

Two distinct failure modes, and both trace to the same sentence: the peg holds exactly as well as redemption does.

• Terra had no real thing to redeem into. Fatal, and unfixable.
• USDC had the real thing but could not reach it for 48 hours. Painful, and temporary.

Which yields the checklist for evaluating any stablecoin, and it is strikingly free of technology:

• What backs it? (a real asset, or a belief?)
• Can the issuer actually pay if everyone redeems at once?
• Is redemption available, reliably, when it is most needed?
• Who is the issuer, and what happens if their bank fails?

These are questions about credit and liquidity, not about blockchains. A stablecoin is a claim on an institution. The chain determines how the claim moves, which is the subject of Lesson 2, and it is genuinely useful. But it never determines whether the claim is good.

Redemption at par lets arbitrageurs buy a discounted token and redeem it for a full dollar, pushing the price back; the peg breaks when redemption fails, either because reserves are not real, as with Terra, or because they are temporarily unreachable, as with USDC.