Stablecoins and How Pegs Are Held

Crypto markets trade every hour of every day. Banks do not. A trader stepping out of Bitcoin at two on a Sunday morning cannot wire the proceeds anywhere, so exchanges needed something that behaves like a dollar but settles on the same rails as the coins. That is a stablecoin: a token supposed to be worth exactly one dollar, always.

Nothing in the software makes that true. A blockchain is excellent at proving you own 1,000 units of something and useless at proving what a unit is worth. The peg is a promise made off-chain, enforced by traders who are paid to enforce it.

Three families, three ways of promising

Fiat-backed (USDT, USDC). A company takes in $1, parks it in Treasury bills and bank deposits, and issues one token as a claim on that pile. Reserves run roughly 80–90 percent short bills and overnight repo, the rest in deposits for same-day redemptions. The issuer keeps the interest: at a 4 percent bill yield, a $60 billion float earns about $2.4 billion a year.

Crypto-collateralised (DAI and its successors). Nobody holds dollars. A user locks volatile collateral into a smart contract and mints stablecoins against it for less than the collateral is worth, and automated liquidations claw back the shortfall when prices fall.

Algorithmic (UST, and a graveyard of imitators). No meaningful collateral; the protocol mints or burns a second, floating token to absorb supply and demand. It holds while confidence holds, because the backing is worth whatever the stablecoin's survival is worth.

Why the price stays near one dollar

An ETF share tracks net asset value because authorised participants can create and redeem at NAV all day and pounce on any gap. A fiat-backed stablecoin runs the same machine.

above the band: mint wire 1.000, sell token at 1.004

below the band: redeem buy token at 0.996, redeem for 1.000

time

Worked example: closing a two-tick discount

An exchange is quoting USDC at 0.9970 because a fund is dumping size into a thin weekend book. A desk with an issuer account steps in.

1. Buy the discount. 20 million USDC at 0.9970 costs $19,940,000.
2. Redeem at par. Send the tokens to the issuer, receive 20 million dollars by wire the next business day.
3. Net the costs. Gross profit $60,000, less two days of funding on $19.94 million at 5 percent — about $5,540 — plus wire and gas fees of a few hundred dollars.
4. Result. Roughly $54,000 for a two-day hold, and the desk's buying is exactly what lifts the price back toward 1.0000.

That trade sets the width of the band. If redemption were instant and free, the desk would act at 0.99995 and the peg would be watertight. Because it takes a business day, has a minimum ticket around $100,000, and is open only to onboarded institutions, the band is a couple of tenths of a percent wide in calm markets — and far wider the moment anyone doubts step 2.

Worked example: an over-collateralised vault

The crypto-collateralised design replaces the wire with a liquidation engine. Take a minimum collateral ratio of 150 percent and a liquidation penalty of 13 percent.

• Open. Deposit 5 ETH at $3,000, so $15,000 of collateral. The most you can mint is $15{,}000 / 1.50 = 10{,}000$; draw the full amount and you sit at exactly 150 percent, with no headroom.
• Price falls. ETH drops to $2,800. Collateral is $14,000 against 10,000 of debt — a 140 percent ratio, so the vault is liquidatable.
• Liquidation. A keeper repays the 10,000 debt and is paid from your collateral: 10,000 plus the 13 percent penalty of 1,300, so $11,300 of ETH (4.036 coins) leaves the vault. You keep $2,700, or 0.964 ETH.
• Your P&L. You still hold the 10,000 stablecoins you minted, ending with $12,700 where holding 5 ETH would have left you $14,000. The penalty cost $1,300.
• The system's view. Every token outstanding is still over-backed, because the shortfall came out of the borrower rather than the token.

Keep the peg safe by making the borrower absorb volatility. That works until collateral gaps down faster than keepers can sell, which is why these systems hold a surplus buffer and have migrated toward Treasury-backed collateral.

What breaks

Pegs fail at step 2, not step 1. In March 2023 Circle disclosed that $3.3 billion of USDC's reserves sat at Silicon Valley Bank, which had just been closed. Nothing about the token changed; what changed was the belief that a dollar of reserves would become a dollar of cash on Monday. USDC traded down to roughly 0.87 over that weekend and snapped back within hours of regulators guaranteeing the deposits. The reserve was fine — access to it was in doubt, and that is all a peg is.

The algorithmic version fails faster. Terra's UST offered redemption into a floating token, LUNA, and paid roughly 19 percent to hold it. When redemptions began in May 2022, minting LUNA to honour them crushed LUNA's price, so each further redemption needed still more of it. Around $18 billion evaporated in four days.

Where it shows up

Stablecoins are the dollar leg of most crypto trading, the margin collateral behind perpetual futures, the unit of account in lending protocols, and now a buyer of Treasury bills large enough to appear in auction statistics. The questions worth asking are narrow: who can redeem, at what minimum, on what settlement lag, and what sits in the reserve. Those four answers give you the width of the arbitrage band, and the band is the whole trade.

Key terms

• Peg — the target price, almost always one unit of fiat currency.
• Reserve — the assets an issuer holds against tokens outstanding; composition and liquidity matter more than headline size.
• Mint / redeem — the primary-market round trip that enforces the peg, usually limited to institutional clients.
• Over-collateralisation — posting more collateral value than stablecoin minted, so price falls hit the borrower first.
• Depeg — a sustained deviation from par; almost always a doubt about redemption, not about backing.
• Attestation — a point-in-time reserve confirmation, weaker than an audit.