May 2022 Terraquake
The loss of value associated with the crash of Luna, the main currency in the Terra blockchain ecosystem, is unprecedented in the crypto space both in terms of scale and speed. Over $40Bn of circulating supply and over $80Bn of fully diluted market value have evaporated. We studied historic examples of pegged crypto assets which seemingly indefinitely lost their peg, a notable example of this being Nubits, which was arguably the first algorithmic stablecoin. Though, this had never happened before to an asset with as large a market cap as UST. As such, the impacts of this de-peg event were felt across the entire crypto ecosystem.
The UST/LUNA system was an extremely impressive example of reflexivity at play, and was able to achieve a scale that no single decentralised stablecoin was able to. However, as is often the case in complex systems, the bigger something gets, the more unstable it becomes. One of the main selling points behind UST was its actual real-world use through Chai, a dedicated payment network that allows convertibility of Terra stablecoins to fiat at the point of sale, and such that in the event of a market crash speculators would rotate out of risky assets into UST and outflows of any size would be manageable. However, greed and yield chasing injected more than $10Bn of assets from speculators into the system. That capital was mercenary and would de-risk at the first sight of worsening conditions, rotating into the next hot thing with better yield.
There could be many hypothetical scenarios where this would be sustainable at a smaller scale, but what is clear with the benefit of hindsight is that reaching such a scale in such a short time made the system fragile. As the old adage goes — slow and steady wins the race.
The events that transpired this week were driven primarily by coins native to the Terra blockchain. Specifically, two tokens: Terra Luna (LUNA) and its sister token Terra USD (UST). UST is a stablecoin which is designed to be pegged to the US Dollar and should trade at an exchange rate of 1:1. UST resides in the subset of stablecoins known as algorithmic stablecoins. The exact definition of an algorithmic stablecoin is subjective but broadly speaking these coins look to maintain their peg without the use of external collateral and through the increasing or decreasing of supply. In the case of Terra, the system is dual token and based on risk layering, where liabilities (UST — the stablecoin) are issued against the value of the equity (LUNA — the risky asset that accrues seigniorage). More specifically, the system allows to “burn” (remove from circulation) $1 worth of LUNA to receive 1 unit of UST or, vice-versa, to “burn” 1 unit of UST for $1 worth of LUNA. For example, if on a centralized exchange UST is trading for $0.8, an arbitrager would be able to buy 1 UST for $0.8, transfer this to the Terra blockchain and exchange this 1 UST for $1 of LUNA. This $1 of LUNA could then be sold, resulting in a $0.2 profit. This process would put upward pressure on the UST price and so result in UST returning to its peg (1:1 exchange rate with the US Dollar).
In recent months, both LUNA and UST had experienced unprecedented growth in market share and adoption. This had been primarily driven by the increase in demand for UST. There were several use cases for UST within the Terra blockchain. The biggest of these is the Anchor protocol which we discuss in more detail further on in the report. Starting from June 2021 the Terra ecosystem experienced an unprecedented increase in total value locked (TVL) that jumped from $0.5 billion to $18 Bn at its peak in mid-May, as displayed in Figure 3. This was accompanied by the continued adoption of the Terra blockchain. This is best exemplified by the total number of wallets (shown in Figure 3) registered with Terra Station, the most popular wallet provider for the Terra blockchain. Furthermore, the terra blockchain had a number of high-profile backers including Jump Crypto and 3 Arrows Capital. A week ago, the Terra blockchain, Luna and UST were the darlings of the crypto markets. With Luna being one of the few assets which had appreciated in value this calendar year.
1. May 7th 19:00 UTC — May 9th 13:00 UTC:
UST is trading at a discount of approximately 2% before eventually recovering to hover around a price of $0.995. This is accompanied by a sell-off of LUNA from $80 to $60 and BTC from $36,000 to $33,000. The catalyst for this was an $85 million swap of UST to USDT on Curve. The Luna Foundation Guard (LFG) lends out their BTC reserves to market makers.
2. May 9th 13:00 UTC — May 10th 19:00 UTC:
UST crashes, trading at a discount as low as 40% to the US Dollar before recovering and trading at a price of approximately $0.9 for 8 hours. Over this period LUNA crashes a further 50% down to $30 and BTC falls as low as $29500 before recovering to $31500. There are voices about a potential bailout from some of Terra’s high-profile backers.
3. May 10th 19:00 UTC — May 12th 01:00 UTC:
UST tumbles as low as $0.2 before recovering to $0.8. LUNA loses over 99% of its value trading slightly below $0.02. BTC falls further to $28000. The founder of the Terra blockchain announces there will be no bailout.
4. May 12th 01:00 UTC — May 13th 15:00 UTC:
UST continues to tumble settling at $0.1. LUNA at this point trading at $0.00005. BTC recovers and moves above the $30000 mark.
Events Play by Play:
The information we provide here is simply our best understanding of the events that played out, given the information available to us at the current time, the details of the event are still being uncovered. Having said that, this is how we believe things played out.
The initial event which shook confidence was a sizable trade of UST for USDC on a Curve pool, the main on-chain reference market and the biggest pool of liquidity for UST. Approximately $85 million UST was swapped for USDC, a modest sum, but a trigger for a snowball of swaps. This likely caused the initial 2% discount. Though UST did rebound, the peg was never fully recovered. At this point, there was a considerable amount of fear spreading in the market regarding a potential permanent de-pegging. This is best exemplified by the Terra Anchor protocol, essentially a savings account where individuals can deposit UST and receive a yield — upwards of 19% per annum. The total UST locked in this protocol peaked at around 14 billion UST. In less than a day, after this $85 million transaction, this number had decreased by 2 billion UST. An unprecedented run to the bank.
It is interesting to note that the timing of this crash occurred at a period when UST was particularly vulnerable. This is due to the migration of liquidity on the Curve decentralized exchange from one pool of liquidity to another. Specifically, the Luna Foundation Guard (LFG) was launching a new liquidity pool, known as the 4-pool (USDC, USDT, UST, FRAX), to replace the competitor 3-pool that includes USDC, USDT and DAI. The LFG is a non-profit organisation run by the creators of the Terra blockchain, and its native assets LUNA and UST. They have several mandates including open market monetary operations to help support the UST peg and providing funding to developers looking to build within the Terra blockchain. The idea is that this would become the dominant source of liquidity for stablecoins which would improve UST adoption outside of the Terra blockchain. To provide initial liquidity for the 4-pool, the LFG removed $150 million of liquidity from the current dominant UST pool on Curve (a meta-pool of UST and shares of the 3-pool). As such, in the interim period between the 4-pool being created and the liquidity being withdrawn, there was $150 million less liquidity for individuals looking to exit UST and it provided a sign of weakness for the on-chain peg, often used as a reference point for centralised trading.
Notably, also, LFG had been buying BTC to create a war chest which could be deployed to help protect the peg, with the idea of having a basket of different assets that are backing the UST, in a manner somewhat akin to the action of central banks defending forex rates with their reserves. Basically, they would look to sell the BTC they acquired and buy UST should UST drop below $0.98. They had acquired over 80 thousand Bitcoins ($3Bn at acquisition cost, $1.5Bn at the time of deployment).
With confidence in the LUNA/UST system failing the LFG lent out $750 million of its BTC reserves to various market makers, which was then used to selectively buy UST. Essentially outsourcing the market operations that the BTC was purchased for initially, this occurred during 2) of the Timeline section. However, the action taken was not enough to fully restore confidence in the system and UST traded at a 10% discount. Anchors deposits were being rapidly drained, with over 4 billion UST having been already removed from the protocol. As it became known just recently, LFG will eventually deploy almost the entirety of their reserves to support UST price over the initial phases of the bank-run and bought $1.5bn UST via various market makers.
At that point confidence in the system was evaporating. LFG plans were public knowledge and due to this, the impact on BTC price caused by the bank-run on UST was amplified, as market participants looked to short BTC in anticipation of the forced selling of collateral. As BTC declined, LUNA started to fall at an increasing pace, due to the inflationary minting-burning mechanism, that converts UST looking to exit the system into newly created LUNA tokens, promptly sold on the market exacerbating the instability of the system. As the BTC price started to near $30k, the fear became overwhelming. As reference support and key price level, on a breakdown, BTC could plunge much further still, testing the $25k level and dragging the rest of the market with it. These vicious dynamics eventually created a large mismatch between the market cap of LUNA and UST, reducing the probabilities that the holders of UST could ever find exit liquidity via LUNA and triggering a massive bank-run (the Terra ecosystem had liabilities much higher than its assets).
Throughout this period there were whispers of a potential bailout agreement which would have involved the OTC purchase of LUNA held by LFG at a heavy discount by a consortium of investors. However, this never materialised; likely in large part to the high uncertainty regarding LUNA’s price, which throughout this period kept declining aggressively.
The founder of LUNA and UST, Do Kwon, finally announced that there would be no external capital injection and that they will simply have to mint enough LUNA such that all the UST that wishes to leave the system at discounted prices could. To help facilitate this, the maximum number of UST that could be burnt and LUNA that could be minted each day was increased massively from 293M units of LUNA to approximately 1.2Bn. This occurs at point 3) of the Timeline section. At this point the amount of LUNA being created was essentially hyperinflationary in nature; see Figure 7 for a graphical representation.
This extreme minting placed unbelievable downward price pressure on LUNA. As such the LUNA price crashed to fractions of pennies. Given the constant downward pressure on LUNA it is highly unlikely that there will ever be enough demand for LUNA to fully absorb all the UST which would want to exit at par.
Lessons learned and a framework for risk-reward in DeFi:
Every event of such scale is an opportunity to reflect, study the possible weaknesses in the design of the system and recognise what further analysis could have been done to take advantage of this event. Here are some of the questions that we ask ourselves to develop a risk management framework for pegged assets.
1. What are the mechanisms for holding the peg and how do they behave in stressed conditions?
2. What are the historical examples where the peg was under-pressure and what happened after?
3. How has the system grown/contracted since then and are there any strong reasons to believe that some of the mechanics which allowed the peg to be maintained are no longer applicable?
4. In what state of the world the faith in the peg is threatened?
5. If there is a bank run (people looking to withdraw assets from the system), how will that play out? What is the likely way in which it would unfold?
6. And the billion-dollar question — how severe the de-peg event can be and how quickly can it play out?
Considering new kinds of derivative products in DeFi, be it liquid staking, liquid ve- tokens to influence liquidity pool rewards, bridged assets or more traditional stablecoins, it is vital to examine the assumption behind the mechanisms that should keep the peg in check. The rewards are usually easy to understand and find, they are the marketing tool. Get an APR of 20%, 30%, 100%, 10000%. We’ve seen everything. The risks, however, often are hidden and are tough to quantify. Thus, the word Ponzi is often used to describe these systems. Labelling that probably grossly undervalues the innovation that is undergoing in the industry, and the fact that creating large incentives to users, ultimately serves as a marketing budget to put the flywheel in motion and allow the system to become self-sustaining, all the while distributing the protocol ownership directly to users.
Given the fragility of these systems, however, we thoroughly study their design to understand possible vulnerabilities and create a detailed framework for liquidating the position based on the indicators that we monitor. Since most of DeFi involves the interplay of a large number of participants in real-time, fear spreads fast and the system is exposed to cascade effects often triggered by the activity of a small number of participants. In most de-pegging events, the trigger usually is a liquidity mismatch (between how much is available and how much is needed to facilitate the exit of everyone who desires to) as a result of a self-fulfilling growing mistrust in the system’s stability. Thus, we are careful to analyse how “rushing to the exit” can play out and try to model scenarios where the system reverts to balance and under what circumstances it finds a new steady state.