On 7 May 2022, Terra's UST stablecoin de-pegged from one dollar. Six days later, LUNA had fallen from $87 to less than $0.00005, UST sat at twenty cents, and roughly $45 billion in market value had vanished. UST had been called a stablecoin for years. It was not stable, and in hindsight it was barely a coin.
Ten months later, on the weekend of 11 March 2023, Circle's USDC fell to about $0.87 after the company disclosed that $3.3 billion of its cash reserves — around 8% of the float — was trapped at the failed Silicon Valley Bank. The reserves existed. The redemption rail was closed for the weekend. The peg cracked anyway.
These two events are usually told as the same story: "stablecoins can break." They are not the same story. UST died because its design was reflexive and had no exogenous collateral. USDC wobbled because a bank failed on a Friday and US markets were shut until Monday. Different machines. Different failure modes. Same word on the label.
This article is the mental model. We will define what a stablecoin actually is, walk the three archetypes, explain why redemption is the thing that holds the peg, and lay out the issuer risks that the 2022–2025 cycle made concrete. The next four parts of this series build on it.
What "stablecoin" actually means in 2026
The stablecoin sector is roughly $315 billion in April 2026, almost double the ~$170 billion at the end of 2024. USDT alone accounts for about $184 billion. USDC is around $77 billion. Ethena's USDe is near $5.9 billion. Sky Protocol's USDS sits in the multi-billion range. PYUSD is near $1 billion and Ripple's RLUSD around $1.6 billion.
A stablecoin is not defined by being on a blockchain. It is defined by a promise: that one token can be redeemed for one unit of a reference asset — almost always the US dollar — with the issuer or protocol that minted it. The blockchain is just the ledger.
What the word hides is that there are three completely different machines holding that promise:
- Fiat-backed: a centralised issuer holds dollar-equivalent reserves and redeems at par.
- Crypto-collateralised: a smart contract over-collateralises with on-chain assets and uses liquidations and arbitrage to defend the peg.
- Algorithmic: a protocol tries to hold the peg with reflexive token mechanics and no exogenous collateral.
These three share a label and almost nothing else. Treating them as one product is the single biggest mistake retail users make. For a deeper first-principles walkthrough of fiat-backed, crypto-backed, and algorithmic designs, we go further into each one elsewhere; here we are interested in their risk profiles.
Archetype 1: Fiat-backed (USDT, USDC, PYUSD, RLUSD)
A fiat-backed stablecoin is, in the simplest framing, a custodial deposit token. You give the issuer a dollar; the issuer holds something dollar-shaped — cash, T-bills, reverse-repo, sometimes commercial paper or precious metals — and mints you one stablecoin. To exit, an authorised participant gives the token back and the issuer wires a dollar.
The two giants illustrate the spectrum. Tether's Q4 2025 BDO Italy attestation reported $186.5 billion in liabilities backed by $122.3 billion in T-bills, $19.3 billion in reverse repo, $17.4 billion in gold and $8.4 billion in bitcoin, with $6.3 billion in excess reserves. Circle, audited by Deloitte, reported $75.88 billion in reserves against $75.81 billion of USDC outstanding as of 31 October 2025 — almost entirely cash and short-duration Treasuries.
These coins are centralised by design. The issuer can freeze any address. Between 2023 and 2025, Tether blacklisted 7,268 addresses, freezing $3.29 billion in USDT. Circle blacklisted 372 addresses, freezing $109 million. PYUSD reserves are held in dollar deposits, Treasuries and cash equivalents under New York State trust-company supervision and behave more like Circle than Tether.
The trade-off is explicit. You get the tightest peg available because redemption is a wire transfer with a regulated issuer. You also accept that the issuer can lock your tokens on receipt of a court order, and that the coin is only as good as the bank holding the deposits and the credit quality of the T-bill book.
Archetype 2: Crypto-collateralised (USDS, sUSDe, crvUSD, GHO)
A crypto-collateralised stablecoin replaces the bank with a smart contract. To mint $100 of USDS or crvUSD, you lock perhaps $130 to $170 of ETH, staked ETH or other approved collateral. The contract enforces the over-collateralisation; if the price of your collateral falls below a threshold, anyone can liquidate it and burn your debt. The peg is held by arbitrageurs who mint when the price drifts above $1 and burn when it drifts below.
Sky Protocol — the rebrand of MakerDAO — migrated DAI to USDS through 2024 and 2025. The Sky Savings Rate currently pays 3.75–4.5% APY by routing protocol revenue back to depositors. Ethena's USDe is a different beast: a delta-neutral synthetic dollar that holds spot ETH or BTC and shorts the same notional in perpetual futures. When funding rates are positive, the short pays you and the dollar value stays roughly flat. When they invert, the model is stressed.
The risks shift entirely. Fiat-backed coins worry about the issuer's bank, custodian and balance sheet. Crypto-collateralised coins worry about oracle manipulation, smart-contract bugs, governance attacks, perp-market dislocations, and cascading liquidations during a violent move. They are usually more decentralised, but they are not safer in absolute terms — they are differently risky.
Archetype 3: Algorithmic — and why most are dead
A pure algorithmic stablecoin holds its peg with no exogenous collateral. The protocol mints and burns a paired volatility token to absorb supply and demand. Terra's UST and LUNA were the canonical example. It worked while inflows held — Anchor's 19.5% APY drove tens of billions into UST — and it died when they reversed.
The death spiral is structural, not a bug. When UST traded below $1, holders were supposed to burn UST for $1 of newly minted LUNA. As LUNA's price fell, defending the peg required minting exponentially more LUNA, which crushed LUNA's price further, which made the peg defence more dilutive. By 13 May 2022, LUNA had hyperinflated and UST sat at $0.20. About $45 billion was gone in a week.
The contagion was real even for survivors. USDT briefly traded as low as $0.9485 on 12 May 2022 as panicked holders dumped it for USDC. The peg held — Tether honoured roughly $7 billion of redemptions in a few weeks — but the episode underlined that even fiat-backed coins move when liquidity vanishes from the secondary market.
Post-Terra, regulation has effectively closed the door on pure algorithmic designs for retail. The GENIUS Act, MiCA, and the broader 2026 stablecoin rulebook require fiat backing for any payment stablecoin offered to the public in the US and EU. Algorithmic experimentation has not vanished, but it has migrated to the over-collateralised end of the spectrum.
How redemption actually works (and why the peg holds)
Every stablecoin lives in two markets at once. The primary market is the issuer's mint-and-burn rail: an authorised participant — typically a market maker, exchange, or institutional desk — wires $1 to the issuer and receives one freshly-minted token, or returns one token and gets $1 back. The secondary market is everywhere else: Binance, Coinbase, Curve, Uniswap.
Arbitrage links the two. If USDC trades at $0.995 on Curve, an authorised participant buys it on the secondary market, redeems with Circle at $1, and pockets half a cent. That activity buys up the cheap supply and pushes the price back toward par. Federal Reserve research published in February 2024 makes the same point in plainer language: secondary-market liquidity is a shock absorber, not a peg mechanism. The peg is held in the primary market.
This is why the SVB weekend mattered so much. Circle's reserves were intact. The bank holding $3.3 billion of them had failed, the FDIC had not yet announced the backstop, and US wires were closed until Monday. The primary mint-and-burn rail was effectively shut. The secondary market did what it always does without a primary anchor: it priced uncertainty. USDC fell to $0.87 and recovered within 72 hours once the federal backstop reopened the rail.
If you remember one thing from this article, remember this: when the redemption rail closes, the peg is whatever the secondary market guesses. Reserve quality alone does not save you over a weekend.
Issuer risk: what 2022–2025 actually taught us
The last cycle gave us a clean typology of how stablecoins fail.
- Bank-run risk. The USDC depeg in March 2023 was not a reserve shortfall. It was a banking-rail failure. Any fiat-backed coin inherits the credit risk of its custodian banks.
- Asset-quality risk. A fiat-backed coin is only as good as its T-bill and commercial-paper book. Pre-2022 Tether held large amounts of commercial paper; the shift to short-dated Treasuries and reverse repo measurably tightened its risk profile.
- Freeze-list risk. Holding USDT or USDC means accepting that the issuer can freeze your tokens if served with a valid court order. This is policy, not a bug.
- Smart-contract and oracle risk. USDS, USDe, crvUSD and GHO live in code. Oracle manipulation, governance capture and contract bugs have all caused depegs in this category.
- Reflexive design risk. Pure algorithmic coins have a track record of catastrophic failure. Terra is the cleanest case; there are others.
Not every depeg is the same event. USDC March 2023 held and recovered in three days. USDT May 2022 held and bottomed near $0.95. Terra did not hold. Learning to tell those apart is most of the skill of reading a stablecoin headline when a 'dollar' is no longer a dollar.
What the rest of the Stablecoins 2026 series covers
This was Part 1 of five. With the mental model in place, the next parts get specific.
- Part 2 covers the GENIUS Act tier — USDC, USAT, PYUSD, RLUSD, AUSD — and what "federally regulated payment stablecoin" actually means in practice.
- Part 3 unpacks USDT offshore: same ticker, three trust models depending on whether you hold it on Tron, Ethereum or TON.
- Part 4 is the crypto-collateralised lane — the DAI-to-USDS migration, USDe's funding-rate dependency, and why these designs still matter.
- Part 5 is operational: choosing a chain, sizing your exposure, and spotting de-peg red flags before they hit the timeline.
Each later part assumes you understand the three archetypes, the redemption rail, and the difference between a depeg and a death spiral. That is the work of this article.
