What an algorithmic stablecoin is, before the postmortem

A fiat-backed stablecoin like USDT or USDC holds a dollar of real-world assets for every token in circulation. An algorithmic stablecoin tries to hold a dollar by code instead of by reserves. The mechanism varies by design; in the Terra version that produced UST, the rule was:

  • Burn one dollar's worth of Luna, valued by the protocol, to mint one UST, subject to the mechanism's fees and limits.
  • Burn one UST to mint Luna worth one dollar at the current market price.

The arbitrage worked symmetrically. If UST traded below a dollar, an arbitrageur could buy UST below peg, burn it to mint Luna worth a dollar, sell the Luna and pocket the difference. If UST traded above a dollar, an arbitrageur could burn Luna to mint cheaper UST and sell it. The mechanism was elegant, fully on-chain, and required no off-chain collateral. It also required Luna to have enough market value to absorb the burns.

The design risk. Redemption creates units of a volatile companion token that must find buyers. Selling pressure can destroy that token's value and require still more issuance. A market-cap ratio is not a solvency test or a precise trigger for this death spiral.

How Terra grew · 2020 to early 2022

UST went live in 2020. Anchor Protocol then became a major source of demand inside the Terra ecosystem: it offered a subsidised yield of roughly 20% APY on UST deposits.

The yield was not generated by lending UST out at 20% — Anchor's borrow-side income was substantially smaller than its deposit-side payout. The gap was filled by a "Yield Reserve" subsidised by Terraform Labs and later by the Luna Foundation Guard (LFG), the foundation behind Terra. Anchor was, in effect, a paid-marketing programme: deposit UST, receive a yield that the foundation paid for, which encouraged holders to mint more UST and grow the network.

The SEC's 2023 complaint, paragraph 75, describes nearly 14 billion UST deposited in Anchor in early May 2022, about 74% of the nearly 19 billion UST outstanding. That concentration made Anchor important to ongoing demand as well as growth.

~19B
UST outstanding
(early May 2022)
~14B
UST in Anchor
(same period)
20%
Anchor APY
subsidised

The Bitcoin reserve, and why it did not save them

The Luna Foundation Guard accumulated Bitcoin as a reserve intended to support UST during a depeg. Public accounts described a reserve of roughly 80,000 BTC before the collapse. That stock of assets was different from cash already available to meet redemptions.

LFG deployed reserve resources during the collapse, but its public account of how all the funds were used was later challenged in U.S. proceedings. The 2025 superseding indictment, paragraphs 43 and 50–51, alleged that some LFG funds remained in commingled accounts and were not used to defend the peg. Those allegations are not a complete transaction reconciliation or a finding on every transfer. Transfers to an intermediary, executed sales and remaining funds must be distinguished; a claim that the whole reserve was exhausted cannot substitute for that accounting.

The collapse · day by day

Saturday, May 7, 2022
UST withdrawals and selling put pressure on liquidity. Anchor deposits were denominated in UST, not a USDC balance to withdraw as described in the old account. Curve's base 3pool and a UST pool using it are also distinct; a precise wallet reconstruction requires transaction-level evidence.
Sunday, May 8 · evening
Pressure continues as holders seek exits from UST. Describing this as a proven attack by a particular actor would go beyond the evidence here. Pool imbalance, withdrawals and market selling need to be distinguished.
Monday, May 9
LFG deploys reserve resources in an effort to support UST. Reserve transfers, loans to trading counterparties and completed market sales are different events; their amounts should not be treated as interchangeable. The intervention does not restore a durable peg.
Tuesday, May 10
The feedback loop becomes visible: UST exits create LUNA, selling pressure lowers LUNA's price and later conversions require more units. The mechanism meant to stabilise UST amplifies the pressure on its companion token.
Wednesday, May 11
UST and LUNA remain under severe stress. A quoted discount is not evidence that an exit can be executed at that price: order-book depth, withdrawal access and network processing all matter during a run.
Thursday, May 12
LUNA issuance surges as confidence collapses. Validators halt the chain amid governance-security concerns. A chain halt adds an operational barrier to exits, separately from the loss in market value.
Friday, May 13 onward
Binance, Coinbase, Bybit, OKX and most major exchanges delist or suspend Luna and UST. Terra announces a fork plan: a new Luna ("Luna 2.0") will be issued via airdrop to holders of the original Luna and UST. The fork ships in late May. The original chain is renamed Luna Classic (LUNC) with UST renamed USTC. Neither recovers in any meaningful sense.

The contagion chain

The collapse did not stay contained. Three Arrows Capital, the largest crypto hedge fund at the time, had concentrated exposure to Luna and to UST-related positions. 3AC could not meet margin calls and entered insolvency in June 2022.

3AC was a major counterparty to Voyager Digital and to Celsius. Both lenders had been borrowing from 3AC, lending to 3AC, or both. Voyager filed for bankruptcy in July 2022. Celsius froze withdrawals in mid-June and filed for bankruptcy soon after. The chain extended further: BlockFi was forced into a credit facility with FTX (later collapsing with FTX), Genesis Capital experienced severe losses (and would eventually file for bankruptcy in January 2023 after FTX), and the broader crypto-lending sector contracted by more than half within six months.

The cleanest way to think about Luna's role is as the detonator. The collapses that followed were already fragile counterparties; Luna was the event that exposed the leverage and the cross-lending that had built up in the bull market.

Who lost what

The market-cap erasure is the headline number — roughly 60 billion combined for Luna and UST at the peak. The retail loss numbers are harder to estimate and more painful. A few categories:

  • Anchor depositors. Anyone holding UST in Anchor for the 20% yield. Many were retail savers in jurisdictions where local currency inflation made the Anchor yield genuinely attractive. The total deposit base was 14 billion at the peak; recovery was effectively zero.
  • Luna stakers and validators. Anyone holding Luna for staking yield or governance participation. Includes a substantial validator infrastructure community that had built tooling on Terra.
  • Exchanges' margin and lending books. Lenders who had Luna as collateral or as borrow exposure faced rapid liquidations during the death spiral. Some had to write off positions.
  • Indirect contagion. Holders who had no Luna or UST exposure directly but who lost funds at Celsius, Voyager, FTX, BlockFi or other lenders that failed in the contagion chain. The total here is harder to pin down; legitimate claims still being processed in some bankruptcy cases in 2026.

The Do Kwon extradition

Do Kwon, the co-founder and public face of Terraform Labs, left South Korea in early 2022 before the collapse. He was charged in South Korea after the collapse with violations of capital markets law. The US Department of Justice and the SEC also filed charges, the SEC alleging that Terra's marketing and structure of UST and Luna had constituted unregistered securities sales.

Kwon was arrested in Montenegro in March 2023 at Podgorica airport on a separate charge of using a forged passport. He served a Montenegrin sentence for the passport offence through 2023 and into 2024. Both South Korea and the United States filed extradition requests; the Montenegrin courts and government went through a prolonged decision process before granting extradition to the United States in December 2024.

Kwon pleaded guilty in August 2025 and was sentenced to 15 years in prison on December 11, 2025. The criminal sentence is distinct from the SEC civil case and creditors' recovery procedures; it does not establish how much any particular holder will recover.

What the regulators wrote into law

Luna's collapse directly shaped the stablecoin frameworks that followed.

MiCA: stablecoin rules from June 30, 2024

MiCA distinguishes e-money tokens referencing a single official currency from asset-referenced tokens referencing other values or rights, including combinations. Classification depends on the token's characteristics. An algorithmic mechanism does not, by itself, place a token outside MiCA or satisfy its issuer obligations.

United States · the enacted GENIUS Act

The GENIUS Act was signed on July 18, 2025, as recorded by the White House. It establishes a payment-stablecoin framework with reserve and issuer requirements. Enactment is distinct from implementation dates and rules; old draft language should not be presented as the current law or as proof of a blanket exemption.

Hong Kong, Singapore, UAE

These jurisdictions have different regulatory perimeters and implementation stages. Hong Kong's Stablecoins Ordinance took effect on August 1, 2025. A shared concern about backing does not establish identical rules or a universal legal category for every algorithmic design.

Algorithmic stablecoins in 2026 · what survives, what does not

The pure algorithmic category is effectively dead at scale. Several hybrid designs exist; most have learned the Luna lesson and added meaningful collateral:

  • Frax (FRAX). Originally a partially-algorithmic design backed mainly by USDC and other reserves. In 2023, the protocol governance voted to move FRAX to fully collateralised. The token continues to operate as a reserve-backed stablecoin in 2026.
  • USDD. USDD's launch announcement is dated May 5, 2022, before the Terra collapse. Its current documentation describes crypto-collateralised debt positions and liquidation, so the legacy mint-and-burn model and reserve basket should not be copied forward. This article does not establish a current market discount.
  • Experimental designs. Assess the actual collateral, exit mechanism and incentives. No dataset cited here supports a supply ceiling for every experimental project.

The market does not seem to want pure algorithmic stablecoins in 2026. The combination of regulatory exclusion, retail trauma and the lack of demonstrated stress survival has kept the category small. Whether a future design will prove the category survivable remains an open research question; the desk's view is that without a credible exogenous reserve, the answer is likely to remain no.

Five lessons every stablecoin holder should keep

Lesson 1 · Subsidised yield is a marketing programme, not a return

Anchor combined protocol income with a subsidised yield reserve; its payout should not be treated as a sustainable return from borrowers alone. Ask where any advertised yield comes from and what happens when incentives end. No fixed percentage separates safe from unsafe products, and a deposit can lose its entire value.

Lesson 2 · The peg holds in calm weather; pricing risk requires bad weather

UST traded near a dollar for two years. The mechanism worked. Inferring from two years of calm that the mechanism would work under stress was a category error. The same point applies to any new stablecoin: the question is not whether the peg holds today but how it behaves on the worst day of the year.

Lesson 3 · Reserve composition matters more than reserve existence

A Bitcoin reserve can lose market value while a dollar peg is under pressure. Its usefulness also depends on who controls it and whether funds can reach the promised purpose. UST's failure does not establish that every announced reserve asset was sold in its defence. Reserve analysis needs both the asset composition and an account of where the funds went.

Lesson 4 · Cross-pool concentration is contagion risk

UST liquidity was concentrated in a small number of Curve pools. A determined large seller could distort the entire on-chain price by emptying one side of one pool. The same risk applies to any stablecoin with thin off-exchange liquidity: a single venue's order-book depth is not the same as ecosystem-wide depth.

Lesson 5 · "It cannot happen here" is the most expensive sentence in crypto

The claim that a peg cannot fail deserves scrutiny even when a project has supporters and reserve assets. Terra's mechanism failed under a run. For another design, examine what buyers, collateral and operational access would remain when holders all want to exit.

Reading list for the Luna postmortem

  • Terraform Labs governance forum and on-chain Anchor data, archived versions on Wayback Machine for 2021-2022.
  • Curve.fi 3pool subgraph data, May 7-13 2022.
  • SEC v. Terraform Labs PTE Ltd and Do Kwon, complaint filed Feb 16 2023; trial verdict April 2024.
  • US Department of Justice indictment of Do Kwon, March 2023, Southern District of New York.
  • Korean Financial Services Commission notice, May 2022 and subsequent.
  • EU Markets in Crypto-Assets Regulation (Regulation (EU) 2023/1114), Articles 36-58 covering stablecoin requirements.
  • Three Arrows Capital liquidation filings, BVI Eastern Caribbean Supreme Court, June 2022 onward.
  • Celsius Network bankruptcy filings, Southern District of New York, July 2022 onward.

Anything you can verify yourself, you should. Corrections to this article live on the corrections page; email [email protected] if you spot an error.