The Terra/Luna Collapse, Day by Day: May 7 to May 13, 2022

Terra before the collapse

Between mid-2021 and early 2022, Terra was one of the busiest stages in DeFi. Its dollar-pegged token UST was the third-largest stablecoin behind USDC and USDT. Its governance token LUNA sat at roughly position seven by market cap among all crypto assets, ahead of Solana and Polkadot. Its flagship lending protocol Anchor offered 19.5% annualized yield on UST deposits — a rate that, when you put it next to anything else in traditional or crypto finance, made no obvious sense, and that almost nobody could resist.

That run of good fortune did not last. Since May 2022, when people refer to Terra, they say "the collapse." This article reconstructs that collapse in order: the mechanism, the day-by-day sequence, and what the principals did at each step.

Terra carried weight in South Korea too: founder Do Kwon made the finance and venture capital category of Forbes Asia's 30 Under 30 list in 2019, and the number of Korean investors who suffered losses in the collapse is estimated at about 200,000 (Wikipedia · Do Kwon).

I. Mechanism: How algorithmic stablecoins were supposed to work

You cannot understand the collapse without first understanding the mechanism it broke.

The UST/LUNA twin-token design

Terra ran on two core assets:

• UST — the algorithmic stablecoin, target price $1.00.
• LUNA — the governance token, price floats freely, absorbs UST's volatility.

The two were bound together by a burn/mint arbitrage mechanism:

1. If UST traded above $1.00 (say $1.02), anyone could burn $1.00 worth of LUNA at the protocol and receive one freshly minted UST, then sell that UST on the open market for $1.02 — pocketing $0.02. This arbitrage pushed UST price back down toward $1.00.
2. If UST traded below $1.00 (say $0.98), anyone could burn one UST at the protocol and receive $1.00 worth of newly minted LUNA. Sell the LUNA, pocket the difference. This arbitrage pushed UST back up toward $1.00 — but it also expanded the supply of LUNA.

This works elegantly when everyone believes it works. Arbitrageurs show up, UST always returns to $1.00, LUNA acts as a sponge that absorbs the volatility. But the mechanism has one hidden precondition: LUNA itself must retain a non-trivial market cap. The moment LUNA's market cap collapses faster than the system can mint, arbitrageurs stop wanting to receive LUNA in exchange for UST. UST then has no path back to $1.00. This is the death spiral that became a vocabulary item in May 2022 — UST falls, arbitrageurs mint enormous quantities of LUNA, LUNA supply explodes, LUNA price crashes, nobody wants to accept LUNA anymore, UST falls further, repeat.

In the days after May 9 this spiral really did run its full course: UST holders kept converting UST into newly minted LUNA, and LUNA fell to practically zero; that week UST fell to around 10 cents (Wikipedia · Terra). Supply minted ever faster while the unit price fell ever faster, showing the positive-feedback risk in a pure-algorithm design when large redemptions and falling confidence occur together.

A capacity check helps explain the risk: compare LUNA's market capitalization at a given moment with the potential volume of UST redemptions, and you can see whether the absorption capacity is thinning quickly. That ratio does not tell you how fast the price will move or where a stable point lies. The feedback actually at work was that UST redemptions burned UST while minting LUNA; expanding LUNA supply and weakening demand pushed the unit price down, so each later redemption required more newly minted LUNA. External collateral or liquidity buffers change that feedback, and Frax's later move to full collateralization was a mechanism change of exactly that kind.

That 19.5% from Anchor

Anchor Protocol was Terra's flagship lending market. Its design was simple on paper:

• Depositors put UST into Anchor and received roughly 19.5% annualized.
• Borrowers posted bLUNA or bETH as collateral and borrowed UST at roughly 11% annualized.
• The staking yield on the collateral was meant to subsidize the deposit rate.

The arithmetic only works if borrow interest plus collateral staking yield exceeds deposit interest. In practice, operating income did not sustainably cover the roughly 19.5% deposit rate. Anchor's Yield Reserve covered the gap and received repeated injections from Terraform Labs and, later, LFG.

The Yield Reserve is the key to understanding Anchor. It was a pool of UST that topped up depositors whenever borrow income fell short of deposit interest. By design, the deposit rate could run higher when the reserve was healthy and was supposed to come down when it was stressed; in practice, for most of the time Anchor was live, the deposit rate sat at or near 19.5%.

The Yield Reserve's balance is public on chain, and its path is easy to summarize: an injection, a stretch of drawdown, another injection. That shows the 19.5% rate depended on outside support; without changes to income, rates or new funding, the reserve would stay under pressure. On its own, though, it does not tell you when or how the system would have run into trouble without the May 7 selling.

II. Peak: How a market cap in the tens of billions was assembled

In early April 2022, LUNA set its all-time high of about $119.50 (Wikipedia · Terra). Around then Terra's Total Value Locked, UST's market cap and LUNA's circulating market cap were each near their highs, and UST was already the third-largest stablecoin; a month later came the collapse, which wiped out close to $45 billion of market value within a week (Wikipedia · Terra).

The peak rested on three legs:

1. Anchor's deposit base reached its high in April. The UST deposited in Anchor made up the bulk of the circulating supply. Put differently, a large share of all UST ever minted had never left Anchor up to that moment.
2. LFG publicly committed to accumulating $10 billion of BTC as a UST reserve. Do Kwon announced this "UST BTC reserve plan" in late March. By May 7, LFG held 80,394 BTC (Wikipedia · Terra).
3. Curve's upcoming "4pool" liquidity pool was about to launch. Composed of UST, USDT, USDC, and FRAX, it was designed to be the deepest stablecoin pool on Curve. Once live, UST would have its strongest on-chain exit path.

Taken together, these three items also produced a very fragile structure. UST's stability now depended on (a) Anchor not running, (b) LFG's BTC not being depleted, and (c) 4pool launching on schedule. Any one of those three failing meant the other two could not catch UST alone.

Put another way, the April peak was a reservoir of capital held in place by Anchor's 19.5%: as long as 19.5% kept being paid, the money stayed; once there was a credible sign of the rate being lowered, or UST's own price started to give, the water would run out.

III. Trigger: That Saturday, May 7

The trigger was pulled on May 7, 2022. Around that time, in preparation for the 4pool launch, the Terra team withdrew a large amount of UST liquidity from the existing UST pool on Curve. The operation itself was meant to move liquidity to the new pool, and there was nothing wrong with it procedurally. But it temporarily cut away a sizeable piece of UST's on-chain exit depth.

On May 7 (a Saturday), large UST-for-USDC sell orders appeared on chain and hit that already-thinned UST pool on Curve, and UST slipped slightly off its peg. Who was selling is not settled in the public record. What is certain is the timing: it landed exactly in the window when the pool's liquidity was at its thinnest.

On May 8, UST crept back to around $1 after the slip. But the UST pool on Curve was already carrying a hidden problem. In normal operation the pool's coins sat roughly in balance; after the Terra team moved part of the UST liquidity out for 4pool, the pool grew shallower, and once large UST sell orders hit it, it tilted visibly towards the UST side. The more imbalanced a Curve pool becomes, the more the slippage on the next trade in the same direction grows non-linearly — so although UST looked to have bounced on May 8, the pool's imbalance had not gone away.

One more detail about 4pool: when the collapse began, 4pool had not yet launched. Had it gone live a few days earlier, UST would have had one more source of on-chain depth, which might have absorbed the selling of May 7. But with liquidity already moving out of the old pool and the new pool not yet connected, that gap was precisely when UST's on-chain depth was thinnest — and precisely when the death spiral started.

IV. The days of the collapse, day by day

The genuine death spiral started on May 9. What follows is arranged by date.

V. Deposits, the warrant, the extradition fight and other protocols

Could Anchor deposits be withdrawn at the time?

Anchor's UST deposits had no lock-up period. Anchor's official documentation says deposits can be withdrawn at any time, unless all the stablecoins in that market have been lent out (Anchor documentation); the 21 days is the wait for redeeming bLUNA — that is, for unbonding staked LUNA — not for UST deposits. So during the run, depositors could take their UST out. The real problem was that the UST they took out was itself falling — it still had to be sold on the secondary market to turn into dollars, the later you sold the lower the price, and at the time nobody knew which moment was the last exit.

Do Kwon: from the collapse to his arrest

Do Kwon was in Singapore when the collapse happened. On September 14, 2022, a Korean court issued a warrant for his arrest, and Seoul prosecutors then started the process for an Interpol Red Notice; in December that year, the Korean side believed he was hiding in Serbia. On March 23, 2023, he was arrested at Podgorica airport in Montenegro, carrying forged Costa Rican documents and about to fly to Dubai, with forged Belgian travel documents on him as well (Wikipedia · Do Kwon).

The Korean investigation and the extradition fight

On May 18, 2022, Han Dong-hoon, the newly appointed Korean justice minister, expanded the economic-crime investigation capacity of the Seoul Southern District Prosecutors' Office (서울남부지방검찰청); on its first day, the newly formed joint financial and securities crime investigation team made Luna / Terra its first case (Wikipedia · Terraform Labs).

After Do Kwon's arrest in Montenegro on March 23, 2023, the United States and Korea both filed extradition requests with Montenegro. Montenegro's courts then went back and forth several times on where to send him, with rulings at one stage leaning towards Korea; in the end, Montenegro's justice minister ordered extradition to the United States in December 2024. Do Kwon was extradited to the US on December 31, 2024; in August 2025 he pleaded guilty to two fraud charges, and on December 11 that year he was sentenced to 15 years in prison (Wikipedia · Do Kwon).

What it did to Curve, Frax and DAI

The large amount of UST sitting in Anchor was exposure inside the Terra ecosystem. But UST was also a core component of Curve's UST pool and of the planned 4pool, so external DeFi protocols had heavy exposure too. When that UST pool became badly imbalanced on May 9, the pool's LPs (liquidity providers) were forced into the extreme, maximum-impermanent-loss case — the other stablecoins they had supplied were swapped out for UST, UST fell to almost nothing, and most of the LPs' principal went with it. As this archive sees it, the DeFi world has been much more cautious since then about stablecoin pools that have no real collateral and rely on an algorithm to hold the peg — a caution that Luna taught the hard way.

Frax was the most-watched "semi-algorithmic stablecoin" at the time. Frax v1 used a hybrid of partly algorithmic and partly collateral-backed design; it came under pressure in the days of the Luna collapse too, but it held, and afterwards Frax moved step by step towards full collateralization. MakerDAO's DAI, primarily over-collateralized, largely held its peg through the collapse.

VI. LFG's BTC reserve

On May 7, before the collapse, LFG held 80,394 BTC (Wikipedia · Terra). The same article records that Terraform Labs drew on LFG's bitcoin and other reserves to steady UST and LUNA and could not restore the 1:1 peg, and that as of May 16 blockchain analysts said how the reserve had been spent was still unclear. LFG gave its own update on May 16: according to Fortune's report that day, only 313 BTC were left, the rest having been sold between May 8 and May 10.

VII. Aftermath: rescue, fork, arrests

After the chain halted, Do Kwon proposed the "Terra 2.0" plan: keep the LUNA name for a freshly launched chain and rename the original collapsed chain "Terra Classic" with its token relabeled LUNC. The community argued about it in two camps for about two weeks, and the new chain finally launched after a governance vote. The two chains have different tokens, supplies and market structures; current price, market cap and on-chain activity require live data rather than treating new LUNA as a continuation of the old market cap.

The legal proceedings since then:

LUNC's burn mechanism and the not-dead embers

Terra Classic (LUNC) — the old chain — was never officially shut down. It has continued operating in a "community-takeover" mode. In September 2022, the Terra Classic community passed a proposal introducing an on-chain transaction tax burn — a share of each LUNC transfer is taken and burned, with the aim of slowly working down the supply minted during the collapse. How the rate was set and how exchanges implemented it has not been consistent over time, so go by Terra Classic's on-chain governance records and exchange announcements; for current supply, price and market cap, look at live data on a block explorer and a market-data site. The thing to remember: it carries no remaining stablecoin functionality.

Jump's Tai Mo Shan settles with the SEC

In December 2024, Tai Mo Shan Limited, part of Jump Trading's crypto business, agreed to pay $123 million to settle with the SEC (Wikipedia · Terraform Labs). The conduct the SEC pointed to dates to May 2021. According to the SEC, UST lost its peg that month and Tai Mo Shan bought large amounts of UST on the secondary market, helping pull the price back to around $1; Terra's side presented that recovery as the algorithm repairing itself, the market did not know about the buying behind it, and investors were misled. Tai Mo Shan neither admitted nor denied the allegations.

Where Korea's crypto user-protection law came from

After the Luna collapse, Korea moved faster on crypto user-protection legislation. In 2023 the National Assembly passed the Virtual Asset User Protection Act (가상자산 이용자 보호법), which took effect on July 19, 2024. The law focuses on two things: exchanges must keep customer assets separate from their own; and unfair trading such as price manipulation and spreading false information carries criminal liability. For how the penalties are set, and what separate rules apply to stablecoin issuance, go by the text of the law and the announcements of Korea's Financial Services Commission.

VIII. Is the algorithmic-stablecoin path dead?

After Terra, "algorithmic stablecoins were the experiment that failed" became the common shorthand. That phrasing is not exactly accurate: Frax v1, with its hybrid algorithmic-plus-collateral design, made it through Luna and afterwards moved to full collateralization.

What does not work is the specific path: pure algorithm, zero collateral, governance-token-absorbs-all-volatility. It has a problem it cannot get around — it requires the entire market to keep its confidence in the governance token's future price. Once that confidence breaks at some critical threshold, the mechanism has no external buffer.

UST was not the first stablecoin to attempt algorithmic stabilization either; NuBits (2014) and Basis (shut down in 2018 before launch) came earlier. Where UST differed was Anchor: the 19.5% rate scaled up UST and its potential redemption pressure much faster before the failure.

IX. After Luna: what to check about a stablecoin

1. Ask first what keeps it stable. USDT and USDC are "centralized reserve stablecoins" — their stability depends on the issuer's balance sheet. UST was an "algorithmic stablecoin" — its stability depended on continued market faith in a governance token. Both kinds can fail, through different mechanisms, and the word "stable" in the name does not tell you which kind you are holding.
2. High yields need a verifiable source. Anchor's 19.5%, and the high annual rates later posted by vaults and "real yield" projects, should be split into borrower income, collateral yield, reserves and outside subsidies. If finite reserves or repeated injections provide most of the yield, its duration depends on inflows, spending and the adjustment mechanism.
3. The scale of a rescue is one signal of crisis depth. LFG deployed roughly 80,000 BTC and still did not hold the peg, showing that selling and redemption demand exceeded that buffer. A rescue announcement should be read alongside liabilities, liquidity and redemption volume; the announcement alone does not prove recovery is impossible.
4. Do not read the stated reason for a "withdrawal pause" at face value. In the days of the collapse, many exchanges paused withdrawals on the Terra network, and the announcements mostly gave technical reasons. The technical reasons may well have been true, but for users the result was the same: the money could not get out. The next time you see withdrawals paused because of "network issues," treat it first as a sign that the risk has escalated, and do not wait for it to fix itself.
5. "Governance tokens" are not lifeboats. LUNA's contract with UST was: "I absorb your volatility." That promise had real value while the market believed in it. The promise itself created no value. Once confidence collapsed within a few days, LUNA absorbed nothing. If a stablecoin design relies on a governance token to "catch" volatility during stress, ask what happens to that token's price when the absorption begins.