TerraUSD: the death spiral that erased $18 billion

UST was a seigniorage-style algorithmic stablecoin: no reserve account stood behind it, only an arbitrage relationship with LUNA. Holders could always burn one UST to mint a dollar's worth of LUNA, and vice versa, so when UST drifted off par the protocol expected traders to close the gap for profit. That worked while LUNA was large and liquid relative to UST.

In early May 2022 a cluster of withdrawals and large stablecoin swaps hit at the same moment liquidity was unusually thin. UST slipped below a dollar, and instead of arbitrage quietly restoring the peg, the redemption mechanism began minting LUNA into a falling market. The faster holders fled, the more LUNA the protocol printed, and the cheaper LUNA became — a self-reinforcing loop with no floor.

By 2022-05-13 the system had unwound completely. LUNA's supply had hyperinflated by orders of magnitude, the chain had been halted to protect what remained, and UST's roughly $18.77B peak was effectively wiped out. It remains the largest stablecoin failure on record and the canonical example of an algorithmic peg breaking.

Pharos does not surface UST on a public coin page, but it preserves the asset's historical price and supply series as a shadow record so the collapse continues to inform the Pharos Stability Index calibration.

The convertibility that kept UST at a dollar in calm markets is the same mechanism that destroyed it under stress. Burning UST minted a fixed dollar value of LUNA regardless of LUNA's price, so as LUNA fell, each redemption required minting more and more tokens. Supply that started in the hundreds of millions ran into the trillions within days.

The fatal threshold was reached when LUNA's total market value dropped below the UST still in circulation. At that point the backing token could no longer absorb the stablecoin it was meant to support, and rational holders all tried to exit before the value drained away. The result was a classic bank run with no reserve and no lender of last resort — only an algorithm that responded to selling by issuing more of an already-collapsing asset.

This is the structural signature an algorithmic peg carries by design: the stabilizer is a volatile sister token, and the same arbitrage that defends the peg upward will dilute it downward without limit. Pharos flags this exposure through PegScore and the DEWS early-warning surface, which weight reflexive, uncollateralized backing as a high-fragility configuration.

Demand for UST was not organic — it was largely manufactured by Anchor's advertised yield of roughly 20%. By spring 2022, on the order of three-quarters of all circulating UST was deposited in Anchor chasing that rate, which concentrated the entire stablecoin's float in a single venue with a single exit.

The yield was a subsidy, not a return. The reserve funding those payouts was draining at a rate measured in millions of dollars a day, and the community had already begun stepping the rate down, signaling that the headline number could not last. When the rate outlook softened and large depositors started leaving, the withdrawal pressure landed on a peg with no collateral and a narrow liquidity base.

The lesson generalizes beyond Terra: a stablecoin whose demand depends on an above-market, subsidized yield is renting its float rather than earning it. When the subsidy thins, the deposits leave together, and a concentrated venue turns an orderly drawdown into a run. Pharos tracks yield-driven supply concentration as a structural-resilience input rather than treating high yield as a strength.

Terra was deeply wired into the rest of crypto, so the collapse did not stay contained. As redemptions spiked on 2022-05-12, even the largest fiat-backed stablecoin wobbled: USDT briefly traded near $0.95 on some venues before market makers arbitraged it back to par. The deviation was shallow and short-lived, but it showed how a single algorithmic failure can stress otherwise unrelated, reserve-backed assets.

Crypto-collateralized stablecoins took direct damage. MIM, the Abracadabra CDP stablecoin, fell to roughly $0.93 in the same window as collateral values and confidence dropped together. The shock then propagated into leveraged credit: several large lenders and funds carrying Terra exposure or correlated positions failed in the weeks and months that followed.

The episode reset how the market and regulators viewed stablecoin risk. It drew a sharp line between assets backed by liquid reserves and assets backed by their own reflexivity, and it became the reference case in subsequent stablecoin rulemaking and risk frameworks.

An uncollateralized peg defended only by a sister token has no floor. The same convertibility that holds the line in calm markets becomes a dilution engine in a run, and once the backing token's market value falls below the stablecoin's float, the design cannot make every holder whole. Reflexive backing should be read as a fragility, not a feature.

Subsidized yield is a demand mirage. When most of a stablecoin's supply sits in one venue chasing an above-market rate that the issuer pays out of reserves, the float is borrowed and will leave on the same signal. Concentration of supply and the source of its yield matter as much as the headline market cap.

Reserve-defense buys time, not solvency. Spending a Bitcoin treasury into a falling market drained the reserve without changing the underlying mechanism. A peg that depends on continuous intervention is already broken; the intervention only delays recognition.