The Great [stablecoin] Filter — Is Terra Almost Past It?
“If such advanced life had substantially colonized our planet, we would know it by now.” -Robin Hanson
Like blockchain protocols, stablecoins face their own trilemma, commonly defined as a constraint to a maximum of two of the following three: decentralization, stability, and capital efficiency. Custodial stablecoins such as USDC, USDT, and Paxos are highly stable and capital efficient with a 1:1 collateral ratio, but achieve this at the cost of centralization. Non-custodial, on-chain collateralized stablecoins such as MakerDAO’s DAI achieve stability and decentralization at the cost of capital efficiency. The volatile assets composing the collateralized debt positions (CDP) referred to as vaults require over-collateralization in excess of 150%. MakerDAO additionally exchanges a large degree of decentralization for volatility robustness. 34% of all DAI are collateralized by USDC currently. Should USDC collateralization continue to increase, DAI risks evolution into a synthetic USDC. (The USDC backing in MakerDAO is technically a swap rather than a CDP, but this technical differentiation is beyond the scope of this article).
Seignorage shares stablecoins are the primary category of algorithmic stablecoins attempting to solve the stablecoin trilemma by employing monetary contraction and expansion mechanisms, backed by partial or completely endogenous collateral, to achieve decentralization and capital efficiency. To the extent they lack exogenous collateral, all seignorage shares-based stablecoins are fundamentally belief-based systems. When all economic incentives for stability are tied to rewards endogenous to the system itself, a protocol is at risk of losing stability should protracted or severe volatility shake users’ belief in the stablecoin’s peg. This inherent attribute of algorithmic stablecoins has caused deep disagreement amongst developers, founders, and investors regarding long-term viability.
Due to this belief-based system, seignorage stablecoins demonstrate reflexivity, amplifying directional movements and placing them at risk of entering into a “death spiral” as loss of user confidence amidst economic contraction feeds into a destructive, positive-feedback loop.
The seignorage model is the evolution of monetary policy via code, employing various mechanisms to contract and expand the money supply in accordance with market conditions while continually absorbing volatility in order to maintain the peg. Just as civilization’s history is strewn with the corpses of nation states that employed reckless or untenable monetary policy, so too will crypto history be strewn with the corpses of seignorage stablecoins that failed to provide adequately robust protections against volatility amidst periods of severe and protracted contraction.
Terra
The Terra protocol is an open-source, public blockchain built on the Cosmos SDK and designed around its algorithmic stablecoin, Terra. It was incepted by Do Kwon in 2018. There are two main token pools within the Terra protocol:
- Terra: These are fiat-tracking stablecoins, pegged 1:1 to their respective currency and minted by burning Luna. Terra is designed to be stabilized at a 1:1 peg via an elastic coin supply. Multiple denominations exist and all occupy a single pool. At the time of writing, total supply of Terra by denomination is broken down as follows. TerraUSD’s (UST) supply is 15,387,670,000. TerraKRW’s (KRT) supply is 36,867,150,000 ( $ 30,231,063). Terra’s Mongolian tugrik’s (MNT) supply is 216,300,000 ( $ 73,542). The base Terra stablecoin, TerraSDR (SDT), which tracks the International Monetary Fund’s Special Drawing Rights (SDR), has a current supply of 700,100,000 ( $ 407,822,252). All other existing Terra stablecoins are negligible in quantity. 97.2% of all value in the current Terra supply is captured by UST. With that in mind, I will use Terra and UST interchangeably going forward.
- Luna: Luna is the native Terra blockchain token, staked by network validators in order to secure the network. Transaction fees on the Terra blockchain are paid to validators in Luna (Validation/Mining are largely irrelevant to this article’s analysis. More information on Tendermint consensus on the Cosmos SDK can be found here and specific application of the consensus protocol on Terra can be found here). Luna is the share component within the Terra blockchain’s seignorage shares model, utilized to absorb volatility and stabilize the coin component (UST). At the time of writing, there is a total supply of 720,100,000 LUNA, down from the genesis creation of 1 Billion LUNA.
$1 value of LUNA can ALWAYS be exchanged for 1 UST, and vice-versa, via the protocol’s mint and burn mechanism. Stabilization of UST is proposed to work via the following:
Expansion: When demand for UST increases, UST will begin to trade above the $1 peg. Arbitragers are then able to burn $1 of LUNA in order to mint 1 UST and trade it. This process will continue until the peg is regained. This expands the money (UST) supply and decreases the LUNA supply during periods of economic growth within the Terra ecosystem. As the Terra ecosystem grows and demand for UST causes a decrease in the LUNA supply, LUNA should be expected to grow in value. Of note, following the Columbus-5 upgrade in September 2021, 100% of LUNA is now burnt by the protocol when minting UST, which makes Terra technically no longer a seignorage stablecoin.
Contraction: When demand for UST decreases, the price will trade below the $1 peg, indicating that the money supply needs to contract. Arbitragers are then able to burn 1 UST and mint $1 of LUNA, in a reversal of the expansion process. This process absorbs UST’s volatility by diluting LUNA holders. Monetary contraction is never cost-free. 41.03% of all LUNA is currently staked, implying that Terra validators burden a substantial portion of short-term dilution during periods of economic contraction, in the hope that medium-to-long term ecosystem growth will result in greater incentives. Furthermore, LUNA can theoretically be infinitely diluted, though this is limited by on-chain liquidity parameters that exponentially increase the swap spread beyond a certain quantity (explained further in May 2021 peg deviation analysis further below).
The Great Filter
The Fermi Paradox is a scientific observation that originated amongst a number of prominent physicists in the 1950s. It contends with the lack of alien life evidence given the high probability estimate for it, based upon the universe’s age and size. Economist Robin Hanson attempted to address the paradox by introducing The Great Filter in 1995. He proposed that amongst the evolutionary steps required for a species to advance adequately far enough to colonize the universe, passage of at least one of those steps must be highly improbable, otherwise the universe would be replete with highly-advanced, colonizing species. At the improbable step(s), the universe ruthlessly filters and eviscerates species from the cosmic timeline. If a species successfully passes through the great filter, they will achieve long-term, sustained existence.
The great filter is an analogously appropriate lens through which to assess seignorage stablecoins seeking to solve the stablecoin trilemma. However, unlike humanity’s future, the reflexive design of seignorage stablecoins reveals the improbable evolutionary step — the great filter — that determines long-term viability. The great filter of seignorage stablecoins is highly severe and protracted economic contraction, such as that seen in historical financial crises. The market, much like the universe, is indifferent and unforgiving. Just as in the case of humanity’s future, there are no second chances for stablecoins to pass through the great filter. While decentralization may exist on a spectrum, stablecoin survivability is binary. Algorithmic stablecoins either possess the requisite stability mechanisms to recover any peg deviation arising during market dislocations or they are likely to die.
The whole of crypto remains a speculative, risk-on asset class in 2022. It has never experienced protracted, global, risk-off market conditions such as those in the aftermath of the 2008 financial crisis. Should similar conditions arise before crypto deeply integrates into global financial infrastructure, crypto can be expected to contract faster and more severely than any other asset class. Illiquidity and despondency will drown out all other narratives as both speculators and investors flee to higher financial ground. Is Terra capable of surviving such conditions for a multi-year period?
Ecosystem Development and Token Demand
Terraform Labs raised $150M in July 2021 in order to further develop a burgeoning ecosystem intended to create legitimate, perpetual demand for the Terra stablecoin. The team has clearly communicated their strategic approach to Terra’s ecosystem in the past, betting on legitimate use cases and demand for UST resulting in price stability.
Anchor Protocol:
Anchor is the cornerstone of the Terra ecosystem. It is advertised as a ‘decentralized savings’ protocol that pays a consistent 19.4% yield on UST. The protocol differs from a typical money market such as Aave in that all collateral must be yield-bearing assets, referred to as ‘bonded’ assets within the protocol’s documentation. These bonded assets represent staked assets in proof-of-stake protocols, providing consistent yield. As of now, the bonded versions of staked LUNA, ETH, and AVAX are the only assets available to be utilized as collateral within the protocol. Terraform Labs claims that the use of interest bearing collateral enables the high yield on deposited UST.
Anchor recently expanded its reach to Avalanche, enabling seamless bridging of assets from Avalanche to Terra through Wormhole via an extension the team calls xAnchor. I almost didn’t recognize my UST was bridged from Avalanche to Terra through the app during my first interaction. With no requirement to interact with Wormhole or Terra aside from an approval transaction, most users won’t even know their assets are leaving the Avalanche chain. Terra Station wallet is not required. No understanding of bridging is required. This is the seamless and simple user experience Terraform Labs consistently speaks about. As Terraform Labs continue to expand to other chains, demand for UST can be expected to grow substantially so long as the 19.4% yield remains stable.
19.4% yield on a stablecoin is deeply unaligned with the yields available in other money markets such as Aave, Yearn’s Iron Bank, Benqi, Scream Finance, Blizz Finance, and Trader Joe. The protocol pays UST lenders from collateral staking yields and UST borrower interest rates. During risk-on market conditions when borrowers pay higher rates, it is theorized that lender yield < collateral staking yield + borrower interest rates. This excess contributes to a yield reserve, which currently contains 401,225,046 UST. During lowered borrowing demand, such as much of the past year, this reserve can be drawn upon in order to ensure the 19.4% yield. The yield reserve ran a $100M deficit during the last 30 days.
Terraform Labs has previously stepped in to replenish the yield reserve during extended periods of lowered borrowing demand, with the second injection of $450M in February bringing the total to $520M.
While conversations about lowered, dynamic yield rates to ensure protocol solvency have occurred, there is no evidence that Anchor will actually abandon its 19.4% yield in the near future. The past year has demonstrated that on a high-enough timeframe, the occurrence of yield reserve depletion due to a sufficiently protracted market draw-down is a probabilistic certainty when yields remain fixed. Terraform Labs is making a convicted bet — $520M thus far — that some critical mass of users necessary to create Terra ecosystem sustainability can be generated through subsidized, stable yield.
Yields on Anchor will eventually re-align with the wider market when the yield reserve is permanently depleted and Anchor (ANC) emissions stop incentivizing borrowing. Assuming a generous 9% yield on all staked collateral, Anchor currently generates $517M annual yield on collateral and an additional $320M via UST borrowing interest (at current 11.57% APR), for annual income of $837M. With over 11 Billion UST deposited, Anchor is currently paying out $2.23B in annual yield. While protocol liabilities and assets will continue to fluctuate over time, they can be expected to only further increase Anchor’s deficit. The historical trend is exceptionally clear. Total UST deposits are increasing at a rate far greater than that of total borrows (borrow rates are directly linked to the rate of change in collateral deposits) as speculators continue to swoop in to take Terraform Labs’ freely offered money.
With the above assumptions in play, the protocol would have to pay under 8% on UST deposits to maintain solvency with its current assets and liabilities. The yield rate could settle in a range slightly above 8% as UST deposits decrease in response to decreasing yield, but the 15–16% yield rate claimed by Do Kwon is not realistically sustainable. There is only one reason to pay 11.57% to borrow UST rather than the current 3–5% market rate for other stablecoins; you can earn 19.47% yield on the UST you’ve just borrowed. This recursive demand for UST is unhealthy and precisely what Do Kwon has spoken out against in the past. When yield begins to drop, the borrow demand can also be expected to plummet.
As Anchor yields begin to align with the wider market, investors and speculators can be expected to rapidly move back into exogenously collateralized stablecoins assessed to be lower risk. So long as custodial stablecoins are not facing imminent regulatory risk and Circle is solvent, USDC and other collateralized stablecoins such as DAI will remain the obvious choice for users that prioritize returns and risk minimization over philosophical dedication to decentralization. As both yield and borrowing demand begin to drop , it is highly probable that a bank run on Terra could occur given that 73.4% of the total UST supply is currently concentrated in Anchor.
CHAI:
In June 2019, Terraform Labs partnered with CHAI, a Korean mobile payments company founded by Terra co-founder Daniel Shin. The integration of Terra into the CHAI ecosystem enables purchases with reduced transaction fees for merchants. Prior to the Columbus-5 upgrade, a portion of seignorage was dedicated to incentivizing CHAI users with discounts at participating merchants. Within the CHAI app, users pay with South Korean Won (KRW) which is converted to KRT, transacted to the merchant’s Terra wallet, and then converted back to KRW. The process occurs in the background, not exposing users and merchants to the underlying payment rails. Fees are reduced from 2–3% to <1% by utilizing Terra’s blockchain. Juxtaposed against Anchor, real-world payment integrations such as CHAI generate legitimate and sustainable demand for UST.
Users must elect to use Terra within the CHAI app. The millions of Korean consumers that have the CHAI app and card are not transacting with Terra by default. In an attempt to look into details regarding KRT demand via CHAI, I went to chaiscan.com, which was established as a block explorer-style analytics suite for Terra-specific CHAI payments. It has been defunct for at least the last 96 hours.
I then discovered this article, in which a CHAI card holder was reportedly informed by customer service that the partnership between Terra and CHAI was discontinued, effective as of September 2021. When I reached out directly to CHAI via e-mail to confirm this information and request more detailed information regarding the underlying transaction flow that Terra uses in the CHAI app, they responded:
Despite the discontinuation of Terra’s partnership with CHAI, there has been no significant decline in KRT transaction volumes. As shown below, the cumulative KRT transaction volume has maintained a nearly constant rate of increase since 2020 Q3. This linear relationship reveals stagnation over the last 1.5 years, without any significant transaction volume growth since the initial CHAI integration period ended.
If CHAI was on-boarding a significant number of active Terra wallet users, a sudden, pronounced decline in daily volumes should be present in the below chart at the time the integration ended. The absence of a sharp volume decline casts doubt on CHAI’s impact towards on-boarding users. Terraform Labs has not yet made any public acknowledgment or confirmation of the terminated CHAI partnership.
May 2021 Peg Deviation, Recovery, and Subsequent Protocol Improvements
Following the sell-off in early May 2021 during which BTC declined 45% over two weeks, UST temporarily deviated from its peg for several days. The UST total supply contracted 9.5% from 2.1 Billion to 1.9 Billion over a week period.
The market-wide sell-off caused excessive liquidations of bLUNA in the Anchor protocol, amplifying the directional movement of LUNA. When a slight peg deviation developed, traditional arbitrage methods were unable to close the gap due to on-chain liquidity parameters.
Terra’s oracle design for pricing the on-chain UST/LUNA swap mechanism requires consensus on the exogenous, off-chain LUNA price. If off-chain liquidity < on-chain liqudity, then the off-chain LUNA price can be intentionally bid up, allowing on-chain swaps to UST of a greater (profitable) quantity using the bid up price. The reverse then happens as exogenous prices are bid down and manipulators swap back to LUNA. In order to prevent this type of oracle attack, off-chain liquidity must always be greater than on-chain liquidity. Thus, general on-chain parameters prior to the May 2021 peg deviation allowed for up to $20M in swaps at a 2% spread each day. Beyond $20M, the spread begins to increase exponentially. This effectively limits on-chain liquidity, preventing the oracle attack described above. 1 UST is still swapped for $1 LUNA under these parameters, but the spread is paid to Terra miners as a “tax”.
The trade-off in implementing these parameters is that arbitrage becomes unprofitable if the peg is not quickly regained, due to the spread’s exponential increase beyond $20M. The on-chain liquidity parameters exacerbated the duration and severity of the peg deviation, with the swap spread at 500–700bps for most of the episode. This likely caused further loss of confidence and price decrease as users looked to exit UST and LUNA positions via off-chain markets when arbitrage was no longer profitable. Furthermore, circulating LUNA value dropped below UST value on 23 May, rendering the protocol insolvent. While LUNA can be infinitely diluted so long as it trades above $0, user confidence is further shaken when there are no longer enough assets to cover liabilities.
Despite all of the above issues, dislocated arbitrage incentives slowly re-aligned over several days and enabled the protocol to re-establish its peg, which it has maintained since. The 21-day lock-up period for the 33.6% of the total LUNA supply that was staked at the onset of this episode provided much needed stability. Additionally, Anchor protocol’s uninterrupted yield on UST deposits maintained a baseline demand for UST.
Terra’s capacity to re-establish its peg after a relatively severe market-wide contraction is a hopeful nod to Terra’s long-term viability. In a post-mortem analysis, Jump Trading introduced Prop 90 to adjust on-chain liquidity parameters in a bid to improve Terra’s volatility absorption during future market contractions. It effectively increased on-chain redemptions to $100M with lower spreads.
UST FOREX Reserve
In February, the Luna Foundation Guard (LFG) announced a $1B private LUNA token sale in order to build a reserve of decentralized, non-correlated assets as an additional mechanism to reduce UST’s reflexivity during periods of contraction. During exceptionally volatile periods, such as May 2021, users will be able to swap a 1:1 ratio of UST for $1 of the non-correlated, decentralized assets held by LFG. This reserve will likely act as a safety valve during periods when on-chain liquidity parameters inhibit arbitragers from maintaining UST’s peg. The initial reserve will consist of BTC, but the LFG intends to expand the reserve to other non-correlated, decentralized assets in the future. Do Kwon has communicated plans to grow the reserve to 10B going forward.
In addition to BTC’s legitimate capacity to provide volatility absorption as an asset of last resort, introduction of exogenous collateral reserves profoundly alters user perceptions. User confidence in the protocol increases disproportionally to the relatively small percentage of outstanding UST that can actually be redeemed for BTC.
The Way Forward
While Terra’s BTC reserve and May 2021 recovery are positive indicators, I am skeptical of the protocol’s long-term viability. The May recovery is partially attributable to demand sustained throughout the episode via Anchor’s 19.4% yield. This yield creates artificial, recursive demand that I find highly concerning. At a current burn rate of $100M/month and with no evidence that Terraform Labs is successfully transitioning its subsidized user base to legitimate use cases, the clock is running out. When Anchor’s yield can no longer be subsidized and dissipates away, so too will the market’s ephemeral demand for the 73% of the UST supply currently deposited in Anchor. When Terra finally passes through the great stablecoin filter — when protracted, severe, global risk-off conditions crush all of crypto — I do not have confidence that Terra will survive without legitimate token demand. As Iron Finance demonstrated, partially-exogenous collateral is no guarantee of stability. A scenario in which LUNA trends to zero and never recovers while all outstanding UST drop to the BTC reserve backing is a realistic outcome for the market conditions described above. Protracted risk-off conditions such as those described above will either kill the Terra protocol or provide an extremely risk-minimized opportunity to invest in a steeply discounted algorithmic stablecoin with proven survivability. Reconsideration of this assessment would require legitimate use cases such as CHAI payments and Mirror Protocol to account for a majority of UST demand and/or an appropriate change in the regulatory treatment of custodial stablecoins.
