Algorithmic stablecoins and how UST failed in 2022

An algorithmic stablecoin holds its peg by changing supply rather than by holding assets worth the tokens outstanding. The standard construction pairs the dollar token with a second, freely floating token and lets anyone convert between them at the target price, so that supply contracts when the stablecoin trades below the target and expands when it trades above it. The mechanism works while the floating token has market value, and the collapse of Terra's UST in May 2022 is the clearest demonstration of what happens when that condition stops holding.

The mint-and-burn loop, stated precisely

Consider a dollar token and a partner token whose price floats. The protocol offers a standing two-way trade: destroy one dollar token and receive newly created units of the partner token worth exactly one dollar, or destroy one dollar of the partner token and receive one new dollar token. Both legs are executed at the protocol's internal target price, not the market price, which is what creates the arbitrage.

If the dollar token trades at ninety-eight cents, a trader buys it below par, destroys it through the protocol, receives a dollar of the partner token, and sells that for a two-cent profit. Stablecoin supply falls, which supports the price. If it trades above a dollar, the trade runs the other way: buy the partner token, convert it into new dollar tokens at par, and sell them above par. Supply rises and the premium closes. In quiet conditions this is elegant, self-executing, and requires no reserve, which was precisely the appeal.

What holds the system up is not collateral but the market capitalization of the partner token relative to the stablecoin supply. That figure is the implicit backing, and unlike a reserve it is a price, set by the same sentiment that determines whether anyone wants the stablecoin. The algorithmic stablecoin family is defined by this substitution of a market price for an asset.

Why demand had to be manufactured

An uncollateralized dollar token has no natural yield, no issuer distributing interest, and no advantage over an established alternative except that its ecosystem promotes it. Terra's approach was to attach a lending protocol offering a high, fixed, advertised return on UST deposits, funded not by borrower demand but by subsidy from a reserve set aside for the purpose. That subsidy was the reason for most of the demand, which meant the demand was rented rather than earned.

The dependence created a second problem. A subsidized yield that exceeds what borrowers pay draws deposits that leave as soon as the subsidy is questioned. Terra's operators also accumulated a bitcoin reserve intended to defend the peg, which is an implicit concession that the algorithmic mechanism alone was not considered sufficient. Notably, that reserve was itself a volatile asset, correlated with the very market conditions that would trigger its use.

The failure, step by step

In May 2022 the sequence ran roughly as follows, and the ordering is what matters.

1. Large UST holders withdrew from the lending protocol and sold into on-chain pools, at a moment when those pools were being rebalanced and were thinner than usual.
2. The selling pushed UST below a dollar, which is a routine depeg in most systems and normally resolves through arbitrage.
3. Arbitrageurs did what the design intended: they destroyed UST and received newly minted LUNA to sell. Each conversion issued new LUNA, and the more UST was converted, the more LUNA supply grew.
4. LUNA's price fell under the weight of that issuance. Because LUNA's market value was the implicit backing, its fall meant each remaining UST was supported by less, which raised the incentive to exit first.
5. Conversions accelerated. LUNA supply expanded by orders of magnitude within days and its price collapsed toward zero, at which point converting UST returned almost nothing and the arbitrage stopped functioning entirely.
6. The bitcoin reserve was deployed into a falling market and was insufficient. UST never recovered its peg, and the network was eventually halted.

The critical observation is that no step required fraud, an exploit, or a bug. Every actor behaved rationally and the protocol executed exactly as written. The loop that defended the peg in calm markets is arithmetically the same loop that destroyed it, because the direction of the reflexivity flips once the partner token's fall becomes the reason to exit. This is a design property, not an accident, and it is why the pattern is often described as a death spiral.

What the collapse propagated into

Losses did not stay inside one ecosystem. Firms holding UST, LUNA, or leveraged positions built on them absorbed the impairment, and the failures of Three Arrows Capital, Celsius, and eventually FTX later in 2022 unfolded against that backdrop of insolvency and forced deleveraging. The mechanism of contagion was ordinary: collateral pledged in several places at once, positions unwound simultaneously, and counterparty risk that had been assumed away. Assets with no connection to Terra fell because holders of Terra exposure were selling whatever they could.

Reading a design that claims to be different

The label has since been applied loosely, and some tokens called algorithmic hold substantial collateral and merely automate its management. That is a materially different product. Four questions separate them.

• If the token's supply were fully converted through the mechanism today, what assets would holders receive, and are those assets independent of the token's own price.
• Is the demand for the token paid for by a subsidy, and what happens to supply when the subsidy stops.
• Does the stabilizing action increase the supply of an asset whose price the mechanism depends on.
• Who can intervene, on what authority, and does a protocol emergency pause exist along with the admin key implications of having one.

On this site, peg deviation and circulating supply together describe a stablecoin's state, and supply change shows whether the mechanism is expanding or contracting. For a token with a paired floating asset, its market capitalization and circulating supply are the figures that describe the implicit backing, with the caveat that a market capitalization is a price multiplied by a supply and is not an asset that can be sold at that value.

The next article turns to verification: what a reserve attestation tests, and what it leaves untested. The incidents page records the 2022 failures in sequence, and risk groups the structural failure modes described here.