Case Study 1: The Terra/Luna Death Spiral — $40 Billion Gone in a Week

Background

In the spring of 2022, Terra was one of the largest blockchain ecosystems in cryptocurrency. Its stablecoin, UST, had a market capitalization of approximately $18.5 billion, making it the third-largest stablecoin behind USDT and USDC. Its native token, LUNA, traded at approximately $80 with a market cap of roughly $28 billion. The Terra ecosystem included over 100 decentralized applications, and its flagship lending protocol, Anchor, held deposits exceeding $14 billion. Terraform Labs, the company behind Terra, was led by Do Kwon, a charismatic and combative Stanford-educated developer who had dismissed critics of UST's design with derision on social media.

By May 14, 2022 — just one week later — UST was worth $0.04. LUNA was worth less than a hundredth of a cent. Over $40 billion in combined market capitalization had been obliterated. The collapse triggered a contagion chain that would topple hedge funds, lending platforms, and exchanges across the crypto industry over the following months.

This case study reconstructs the collapse hour by hour, models the mathematics of the death spiral, and identifies the design decisions that made catastrophe not just possible but, under stress, inevitable.

The Architecture: Elegant and Fatal

Terra's stablecoin mechanism was based on a burn/mint relationship between two tokens:

UST was the stablecoin, targeting $1.00. It had no reserves backing it — no dollars, no Treasuries, no crypto collateral in the traditional sense. Its "backing" was the ability to exchange it for $1 worth of LUNA at any time.

LUNA was the native blockchain token. It accrued value from transaction fees, staking, and — most importantly — the perceived growth potential of the Terra ecosystem.

The mechanism: anyone could burn 1 UST to receive $1 worth of LUNA (at LUNA's current market price), or burn $1 worth of LUNA to mint 1 UST. This created a two-way arbitrage:

  • **UST > $1.00:** Burn LUNA, mint UST, sell UST above $1.00. Profit = (UST price - $1.00).
  • **UST < $1.00:** Buy UST below $1.00, burn UST, receive $1 of LUNA, sell LUNA. Profit = ($1.00 - UST price).

In equilibrium, this mechanism was supposed to keep UST at exactly $1.00 through constant arbitrage pressure. And for over two years, it did.

The Anchor Flaw: Building on Quicksand

Anchor Protocol was the mechanism's Achilles' heel. Launched in March 2021, Anchor was a savings-and-lending protocol on Terra that offered approximately 19.5-20% APY on UST deposits. This yield was not generated organically through lending demand. Borrower interest payments covered only a fraction of the depositor yield. The remainder was subsidized by:

  1. Staking rewards from borrower collateral (bLUNA, bETH)
  2. Direct subsidy from a yield reserve funded by Terraform Labs and the Luna Foundation Guard

At Anchor's peak, approximately $14 billion of the total $18.5 billion UST in circulation was deposited in Anchor — over 75%. The yield reserve was depleting at a rate of approximately $300 million per month in early 2022. At this burn rate, the reserve would reach zero by mid-2022.

The fragility this created: UST demand was not driven by organic use as money — for payments, trading, or savings at market rates. It was driven almost entirely by the unsustainable 20% yield. Remove the yield, and you remove the primary reason most people held UST. The stablecoin's demand was a function of a subsidy, not of utility.

In February 2022, Anchor governance approved a proposal to make the yield rate dynamic, allowing it to decrease. By May, the rate had declined to approximately 18%. Deposits began flowing out. The slow leak had started.

The Luna Foundation Guard: The Insurance Fund

Recognizing the fragility, the Luna Foundation Guard (LFG) — a Singapore-based nonprofit affiliated with Terraform Labs — began accumulating Bitcoin as a reserve to defend the UST peg in a crisis. By early May 2022, the LFG held:

  • ~80,394 BTC (~$3.27 billion at the time)
  • ~39,914 BNB (~$15.8 million)
  • USDT and USDC holdings
  • AVAX and LUNA holdings

Total reserves: approximately $3.5 billion. The strategy was to deploy this Bitcoin to buy UST on the open market if the peg came under pressure, absorbing selling until the arbitrage mechanism could restore equilibrium.

The critical question that would soon be answered: was $3.5 billion enough to defend an $18.5 billion stablecoin against a crisis of confidence?

The Timeline: Seven Days of Destruction

Saturday, May 7

3:00 PM UTC: Large withdrawals begin from Anchor Protocol. Approximately $2 billion in UST is withdrawn over the weekend. On-chain data shows several large wallets — including one wallet holding approximately $500 million in UST — beginning to move funds.

5:00 PM UTC: Approximately $150 million of UST is sold through the Curve Finance UST-3pool on Ethereum. The 3pool (UST/USDC/DAI/USDT) becomes heavily imbalanced — the UST proportion rises from the target ~25% to over 60%.

Curve pool imbalance significance: In an automated market maker (AMM), pool imbalance signals selling pressure. When one asset's share rises dramatically, it means sellers are dumping that asset. Other liquidity providers, seeing the imbalance, begin withdrawing their non-UST assets, further worsening the imbalance.

8:00 PM UTC: UST dips to $0.985 on decentralized exchanges. On centralized exchanges (Binance, FTX), the price briefly touches $0.98 before recovering.

Sunday, May 8

Throughout the day: Additional selling pressure continues. An estimated $350 million in UST is sold through Curve and other venues. The Curve pool remains severely imbalanced.

5:00 PM UTC: The LFG announces it will deploy $1.5 billion in capital to defend the peg — $750 million in BTC loaned to market makers (Jump Trading, among others) with instructions to buy UST, and $750 million in UST to be deployed after the peg stabilizes to restore reserves.

11:00 PM UTC: UST closes the day at approximately $0.975. The first depeg whispers circulate on Crypto Twitter. Do Kwon tweets "Deploying more capital — steady lads."

Monday, May 9 — The First Crash

Early hours UTC: UST drops below $0.95 as Asian markets open. The arbitrage mechanism begins operating at scale: holders burn UST to receive LUNA, then sell LUNA. LUNA's price begins declining.

10:00 AM UTC: UST hits $0.90. The burn/mint mechanism is processing billions in UST redemptions. LUNA's price drops from ~$65 to ~$50 as newly minted LUNA floods the market.

2:00 PM UTC: UST briefly touches $0.60 before recovering to $0.80. The LFG's Bitcoin reserves are being actively sold — on-chain data shows tens of thousands of BTC leaving LFG wallets and flowing to exchanges.

6:00 PM UTC: UST recovers to $0.90 on the back of LFG's intervention. Temporary hope. Do Kwon continues to project confidence on Twitter.

The mathematics at this point:

At $0.90, a UST holder who burns 1 UST receives $1.00 of LUNA. If they sell immediately, they make $0.10 profit. This is the arbitrage working as designed. But consider the aggregate: if $5 billion of UST is burned in a day, that creates $5 billion of new LUNA supply. LUNA's pre-crisis market cap was $28 billion. A $5 billion supply injection represents an 18% dilution — in a single day. This is why LUNA's price is collapsing: the arbitrage mechanism is flooding the market with new LUNA faster than the market can absorb it.

Tuesday, May 10 — Point of No Return

4:00 AM UTC: UST drops to $0.70. LUNA's price has crashed to $30 — a 63% decline from $80 in three days.

The reflexive spiral is now fully engaged. Consider: at $80/LUNA, burning 1 UST mints 0.0125 LUNA. At $30/LUNA, burning 1 UST mints 0.0333 LUNA — 2.67x more tokens. At $10/LUNA, it would mint 0.1 LUNA — 8x more tokens. Each price decline in LUNA accelerates the rate of dilution, which accelerates the price decline.

10:00 AM UTC: UST falls to $0.50. LUNA is at $15. The circulating supply of LUNA, which was 345 million tokens on May 7, has already expanded to over 500 million.

2:00 PM UTC: Terraform Labs introduces emergency measures, voting to increase the UST-to-LUNA burn capacity from $293 million per day to $1.2 billion per day. The intention is to allow the arbitrage mechanism to operate faster. The effect is to accelerate the hyperinflationary spiral in LUNA.

8:00 PM UTC: The LFG's Bitcoin reserves are nearly exhausted. On-chain tracking shows the LFG wallet has been reduced from 80,394 BTC to fewer than 300. Over $3 billion in Bitcoin has been sold in the failed defense.

The LFG reserve depletion demonstrates a critical lesson: a finite reserve cannot defend against a reflexive selling spiral. The reserves buy time, but if confidence is not restored during that time, the reserves simply subsidize the exit of early sellers at the expense of late sellers.

Wednesday, May 11 — Hyperinflation

Early hours: LUNA's price is below $1.00. Its supply has hyperinflated past 1 billion tokens. The burn/mint mechanism is now creating hundreds of millions of new LUNA per hour.

6:00 AM UTC: The Terra blockchain is halted for the first time. The stated reason: to prevent governance attacks. With LUNA worth fractions of a cent, an attacker could acquire a majority stake for almost nothing and take control of the network's governance.

The halting creates additional panic. Users who wanted to burn UST for LUNA can no longer do so. UST falls to $0.30.

12:00 PM UTC: The blockchain resumes. Burning continues at a furious pace. LUNA's supply crosses 10 billion tokens.

6:00 PM UTC: The blockchain is halted again.

Thursday, May 12 — The Aftermath Begins

LUNA's supply has reached into the hundreds of billions. Its price is effectively zero — below $0.001. UST is at $0.15. Major exchanges (Binance, FTX, Coinbase) halt or delist LUNA trading.

Friday, May 13 — Death Certificate

UST: $0.04. LUNA supply: 6.53 trillion tokens. LUNA price: $0.00001. The combined loss: over $40 billion.

The Mathematics of the Death Spiral

We can model the Terra death spiral as a system of coupled differential equations:

Let: - P_UST = price of UST - P_LUNA = price of LUNA - S_LUNA = supply of LUNA - D = rate of UST being burned (sell pressure)

When UST < $1.00, the burn/mint creates:

New LUNA minted per unit time = D / P_LUNA

The selling of this new LUNA creates downward price pressure:

dP_LUNA/dt proportional to -(D / P_LUNA) * selling_fraction

This is a positive feedback loop — as P_LUNA decreases, more LUNA is minted per dollar of burns, which increases selling pressure, which further decreases P_LUNA.

The critical insight: this system has no stable equilibrium below a certain confidence threshold. Above the threshold, rational arbitrageurs buy UST expecting peg restoration, and the mechanism works. Below the threshold, rational actors sell LUNA as fast as it is minted, because holding a rapidly inflating token is irrational. The transition from the stable equilibrium to the unstable equilibrium can happen suddenly — a phase transition, not a gradual decline.

Run code/terra_death_spiral.py to simulate this dynamic with adjustable parameters.

The Contagion Chain

The destruction did not stop at Terra's borders.

Three Arrows Capital (3AC): The hedge fund, managing approximately $10 billion in crypto assets, had significant positions in LUNA. Losses from the collapse, combined with declining values across their portfolio, meant 3AC could not meet margin calls from its lenders. 3AC defaulted on over $3.5 billion in loans from Genesis, Voyager, BlockFi, and others. Its co-founders, Su Zhu and Kyle Davies, fled to avoid creditors and were later arrested.

Celsius Network: The crypto lender had exposure to 3AC's default, held illiquid stETH positions, and had made leveraged bets that went wrong. On June 12, 2022, Celsius froze all withdrawals, trapping $4.7 billion in customer deposits. It filed for bankruptcy in July.

Voyager Digital: The crypto brokerage had lent $650 million to 3AC with minimal collateral. When 3AC defaulted, Voyager was insolvent. Withdrawals were frozen on July 1, and bankruptcy was filed on July 5.

BlockFi: The crypto lender received a $400 million emergency credit line from FTX. BlockFi survived the summer but filed for bankruptcy in November 2022 after FTX itself collapsed.

The thread connecting all these failures was leverage and interconnection. Each entity had lent to, borrowed from, or invested in the others. Terra's collapse was the first domino in a contagion chain that would define the crypto bear market of 2022.

What Warning Signs Existed

The failure was not unpredictable. Multiple analysts and researchers had publicly warned about the specific vulnerabilities that destroyed Terra:

The circular value argument had been articulated by numerous DeFi researchers, including a detailed thread by Kevin Zhou (Galois Capital) in November 2021 explaining why the reflexive mechanism would fail under stress.

The Anchor sustainability problem was well-documented. The yield reserve depletion was publicly trackable on-chain. Anyone could calculate when the reserves would run out.

Historical precedent was clear. Every previous algorithmic stablecoin — Basis, ESD, Iron Finance, Fei — had failed under similar dynamics.

Do Kwon's response to critics was dismissive and, in retrospect, damning. When a researcher warned about UST's vulnerability, Kwon tweeted "I don't debate the poor." When users asked about Anchor's sustainability, he projected confidence without substance. This cult-of-personality dynamic discouraged critical evaluation within the Terra community.

In February 2023, the SEC charged Do Kwon and Terraform Labs with fraud, alleging they misled investors about UST's stability, the use of reserves, and the involvement of a market maker (Jump Trading) in a previous UST depeg in May 2021 that was presented as the mechanism working autonomously.

Do Kwon was arrested in Montenegro in March 2023 while traveling on a forged passport. He was extradited and, in January 2025, a US jury found Terraform Labs liable for securities fraud. Do Kwon faces additional criminal charges.

The Terra case established legal precedent: algorithmic stablecoins can be classified as securities, and their designers can be held personally liable for losses when they misrepresent the mechanism's safety.

Lessons

  1. Circular value dependencies are lethal under stress. Any system where Asset A's value depends on Asset B, and Asset B's value depends on Asset A, has no external anchor. When confidence breaks, the circle reverses from virtuous to vicious with no floor.

  2. Artificial demand creates artificial stability. The 20% Anchor yield manufactured demand for UST. This demand vanished the moment the subsidy became unsustainable, leaving the mechanism to handle exactly the kind of mass selling it could not survive.

  3. Finite reserves cannot defeat infinite spirals. The LFG's $3.5 billion was impressive but irrelevant once the reflexive loop engaged. A finite resource fighting an exponential feedback loop loses every time.

  4. Speed kills. The unlimited burn/mint rate meant the hyperinflationary spiral had no speed limit. By the time anyone could intervene, trillions of LUNA had been minted.

  5. Contagion amplifies losses beyond the initial failure. The $40 billion Terra loss became a $60-80 billion ecosystem loss through interconnected leverage and counterparty exposure.

  6. Critics were right, and they were identifiable. The failure modes were publicly described, mathematically modeled, and historically precedented. The warnings existed. They were ignored by those who profited from the status quo.

Discussion Questions

  1. If you had been an advisor to the Luna Foundation Guard in April 2022, what actions would you have recommended? Could the collapse have been prevented?

  2. The arbitrage mechanism worked exactly as designed — it just made things worse. What does this tell us about the difference between a mechanism working correctly and a mechanism being well-designed?

  3. Do Kwon dismissed critics publicly and aggressively. How does founder behavior affect systemic risk? Should there be disclosure requirements for algorithmic stablecoin risks?

  4. Many retail investors understood the 20% yield as equivalent to a bank savings rate. Whose responsibility was it to ensure they understood the risks — the protocol designers, the media, regulators, or the investors themselves?

  5. If you were designing an algorithmic stablecoin after Terra, what specific mechanisms would you implement to prevent a reflexive death spiral? Is it possible to preserve the design's capital efficiency while eliminating the spiral risk?