[ARFC] Risk Parameters for DAI Update

This seems like a pretty natural response given the pace at which Maker are onboarding sUSDe, but we should take a step back to wait on further analysis from Chaos Labs (or other risk partys willing to provide some additional analysis).

Firstly, we are against fully offboarding DAI as collateral and think such a decision is a bit too harsh and unjustified. Aave’s DAI & sDAI market across all deployments totals ~$368.2M with an avg. utilization rate of ~88% (excl. sDAI supplies), which is small relative to Aave’s total TVL but still significant nonetheless. A full offboarding cuts off a decent portion of Aave’s user base and revenue which will likely increase with the launch of MakerDAO’s Endgame subDAO farms.

Secondly, we do agree that DAI’s risk profile has increased and the DAO should act in response by adjusting risk params that we can all agree on are sufficient given the heightened risk profile of DAI. Pending Chaos Labs’ analysis, we should have sufficient time for ample discussion as the 600M sUSDe increase will not happen overnight.

Monet Supply has provided a good thread on the current and future risk situation for MakerDAO under their sUSDe expansion https://twitter.com/MonetSupply/status/1775210473605980407.

Tl;dr:

  • Current 100M exposure is ~2% of DAI backing
  • 600M will represent ~12% of DAI backing, achieved gradually overtime
  • Morpho vault overcollaterlization sitting at avg. of 12.3% excess capital
  • Maker surplus buffer sitting at 61M DAI
  • $140M in MKR/DAI POL acting as a potential backstop and venue for on-chain liquidations
  • Ethena’s insurance fund sits at $27.5M

Futhermore, USDe will arguably become more robust with the nearing addition of BTC-based perp yield farming.

12 Likes

Summary:

Chaos Labs offers analysis concerning the community’s potential strategy to mitigate risk associated with DAI and DAI-centric assets.

Analysis

Dai Direct Deposit Module (D3M) Exposure

We recognize the community’s concerns regarding MakerDAO’s potential rapid increase of exposure to the novel protocol Ethena. As per our Ethena Perpetual Futures Liquidity Assessment Report, we approximated that the insurance fund (IF) should aim for a bootstrapped size of at least $33M in liquidity for the first 1bn of USDe in circulation, under a 300% linear growth rate assumption and an initial supply of 100M. Currently, the insurance fund sits at just $32M, while the total supply of USDe has grown at an unprecedented rate, reaching 2bn in just three months.

However, from the lens of DAI, it’s important to clarify that the proposed 600M D3M cap does not suggest immediate or swift deployment. Instead, it functions similarly to the GHO facilitator capacity, enabling Maker to deploy conditional debt incrementally, up to 100M at a time, based on the overall health of the protocol and cash flow rates from the D3M. From GHO’s standpoint, with appropriate listing parameterization for sUSDe on Aave and anticipated ongoing increases in the GHO borrow cap, a similar outcome can be achieved, with a portion of GHO debt backed by sUSDe.

Much like GHO cap increases, the D3M sUSDe naturally relies on the assumption that the Maker Risk team will act prudently and rationally under the proposed parameterization, gradually allocating debt while maintaining sound risk management practices. The total welfare and rate-limiting of the Ethena protocol, however, are effectively out of Maker’s control and should be assessed accordingly.

D3M Vault Allocations

image

Screenshot 2024-04-04 at 13.41.08

As a result of Ethena prioritizing rewards for USDe utilization over sUSDe, there is a notable difference in the absolute increases in USDe total supply compared to the USDe staked in sUSDe. Currently, only 24% of all USDe is staked, leading to an asymmetric insurance fund accrual mechanism as the protocol scales up. As per the Ethena docs, for “the duration of the shard campaign, the reserve fund will be funded by a portion of the yield generated by the Protocol asset backing, which relates to the proportion of USDe that is not staked.” The remaining surplus, currently 76%, is being directed to the underfunded insurance fund to maintain its growth alongside the rapid growth in USDe supply.

Screenshot 2024-04-04 at 12.53.37 (1)

Gradual Debt Decreases: On redemptions and hardcoding USDe to $1

The price oracle utilized on Morpho hardcodes USDe to $1, while sUSDe bakes in the sUSDe/USDe exchange rate accordingly.

In the event of a mass redemption event, likely triggered by the draining of the insurance fund due to prolonged periods of negative funding rates, it is expected that sUSDe will trade at a discount to its D3M-enshrined value, considering the 7-day redemption cooldown period involved. However, there currently exists minimal rate-limiting redemption mechanisms for USDe, allowing for 2M redemptions per block and a 25% moat of USDT to ensure a stable liquidity buffer. With 83% of the D3M size allocated towards USDe-specific pools, this implies that the collateral ratio (CR) would have to degrade significantly to not only become net unprofitable for liquidators, based on the liquidator incentive function employed on Morpho, but also for Maker specifically, as demonstrated below.

LLTV (Break Even) Morpho Liquidation Incentive USDe CR Threshold (to deter liquidators) USDe maximum CR decrease from CR Threshold before DAI insolvency Time Buffer Under Compounded Max Daily Negative Returns (Days)
0.77 7.41% 0.9259 20.24% 35
0.86 4.38% 0.9562 11.18% 21
0.915 2.62% 0.9738 6.43% 13
0.94 1.83% 0.9817 4.43% 9

As mentioned previously, the aggregate USDe collateral ratio break-even point for Maker sits at approximately .875, and .7875 through the deployment of the surplus buffer. While we do not endorse this type of oracle logic for the imminent listing of sUSDe on Aave, we recommend implementing an oracle that utilizes $\min(CR,1)$ to adequately price USDe during tail events. However, hardcoding such a value may have acceptable implications from the perspective of any shortfall realized by Maker. This is because a substantial amount of negative funding would need to accumulate rapidly to pose a highly dangerous situation and drop the collateral ratio below the break-even point, thereby allowing a time buffer for all parties to react accordingly. Assuming the largest historic daily negative funding rate, on the day of the Merge (-0.077%), is continuously compounded without any insurance fund, it would take the following number of days for Maker to react to such an extreme scenario occurring:

Screenshot 2024-04-05 at 14.53.20

While we naturally anticipate that market dynamics will evolve with the introduction of Ethena’s substantial OI size, we can utilize historical rates as a proxy to estimate the gradual rate of accumulation and, thus, the implied time buffer.

The Tradeoff: Impact of setting DAI LTV to 0

To clarify the discussions in this thread, it’s important to note that modifying DAI’s Loan-to-Value ratio or Liquidation Threshold does not directly impact its supplying or borrowing capabilities but rather solely affects its usability as collateral within the protocol. Considering this perspective, we present the aggregate DAI-collateralized debt LTV below to assess the potential impact of making such changes within the protocol. It’s observed that the current value follows a downward trend, currently standing at a modest 7.02% of the total DAI supplied, with approximately 60% of this value being used to collateralize stablecoins. This indicates a relatively low demand for DAI as collateral, suggesting that most DAI suppliers are primarily interested in earning interest rather than utilizing it as collateral for borrowing purposes.

Screenshot 2024-04-04 at 16.28.51

In dollar terms, the cumulative debt has remained relatively stable despite the growth in DAI supply on Aave V3, currently amounting to a meager $13.8 million. Taking the average total debt value of $10.2M since inception, and assuming an average weighted interest rate of 7% APY and an aggregate 15% reserve factor, this would translate to an annualized revenue of just ~$107K.

Screenshot 2024-04-04 at 16.29.38

Therefore, in the broader context, the proposed adjustments related to DAI LTV would have minimal implications on protocol revenues and overall utilization.

On altering sDAI LTV to maintain consistency

From a risk perspective, there is virtually no difference between adjusting collateral-related risk parameters in DAI markets and sDAI markets. However, unlike DAI, sDAI maintains an aggregate LTV ranging between 30% and 60% due to its inherent yield-bearing features, although its utilization and total value supplied within Aave have fluctuated.

Screenshot 2024-04-04 at 16.46.00

Screenshot 2024-04-04 at 16.44.15

Conclusion:

The introduction of a pioneering product like Ethena introduces unprecedented quantitative considerations in the DeFi space, particularly regarding the impact of accumulating negative funding on the insurance fund and the overall collateralization of the system. Currently, the protocol utilizes a speculative and artificial reward system, which, while contributing to the insurance fund’s convexity with respect to supply growth, makes it challenging to accurately predict the protocol and the insurance fund’s absolute growth in the long term. The Maker Risk team has adopted a facilitator-capacity-esque framework to model potential future increases, and while the theoretical implied D3M risks are relatively modest at its current scale, such a system inherently relies on implicit trust assumptions from Aave’s perspective. Coupled with the minimal revenue DAI collateral brings to Aave, the community could decide to reduce exposure to DAI as collateral. In this case, we suggest the following two options below.

Conservative

Decrease DAI and sDAI LTV by 12 Percentage Points, LT 1 Percentage Point Per 100M Deployed:

Based on the details outlined in this post, there is a planned gradual reduction in the LTV ratios for DAI and sDAI on Aave V3, transitioning from 77% to 75%, accompanied by a decrease in the LT from 80% to 78% across most chains. Considering these anticipated adjustments and the maximum implied percentage of total DAI backing by (s)USDe at 12%, we conservatively recommend a linear reduction of an additional 12 percentage points in LTV. Regarding the LT, we suggest a conditional decrease of 1 percentage point for every 100M in D3M exposure, with a maximum reduction of 6 percentage points, assuming the full 600M is deployed. LT reductions naturally come with potentially induced liquidations, thus if the amount liquidated due to LT changes is deemed too large, we will modify accordingly. This process will begin with a 1% reduction from the current value to account for the initial 100M, once the initial expected decrease has been completed.

Given current inconsistencies in DAI LTs across the different chains, we will aim to align the LTs and LTVs in the first stage of this proposal. This means that in instances where the LT is currently set higher, the initial reduction would be more than the 1% suggested above. This standardization will be carefully planned to minimize the impact of LT reductions on the likelihood of liquidations.

Specification

Chain Asset Current Expected LTV Recommended LTV Current Expected LT Recommended LT (1st stage) Recommended LT (Last Stage)
V3 Ethereum DAI 75% 63% 78% 77% 72%
V3 Arbitrum DAI 75% 63% 78% 77% 72%
V3 Optimism DAI 75% 63% 80% 77% 72%
V3 Polygon DAI 75% 63% 78% 77% 72%
V3 Gnosis WXDAI 75% 63% 78% 77% 72%
V3 Avalanche DAI.e 75% 63% 80% 77% 72%
V3 Metis m.DAI 75% 63% 80% 77% 72%
V3 Ethereum sDAI 75% 63% 78% 77% 72%
V3 Gnosis sDAI 75% 63% 78% 77% 72%
V2 Ethereum DAI 75% 63% 87% 77% 72%
V2 Polygon DAI 75% 63% 80% 77% 72%
V2 Avalanche DAI 75% 63% 80% 77% 72%

Aggressive

Decrease DAI and sDAI LTV to 0%, LT 1 Percentage Point Per 50M Deployed:

Given Maker’s allocation of a significant portion of potential DAI backing to a new, riskier asset, should the community wish to take more aggressive action, we propose a temporary reduction in the LTV to 0%. Regarding the LT, our recommendation is to conditionally decrease by 1 percentage point for every 50M in D3M exposure, with a maximum reduction of 12 percentage points, assuming the full 600M is deployed. LT reductions naturally come with potentially induced liquidations, thus if the amount liquidated due to LT changes is deemed too large, we will modify accordingly. This proposal prioritizes a risk-averse strategy to hedge against potential future actions by MakerDAO, considering the minimal revenues generated from collateral usage. This process will begin with a 2% reduction from the current value to account for the initial 100M.

Given current inconsistencies in DAI LTs across the different chains, we will aim to align the LTs and LTVs in the first stage of this proposal. This means that in instances where the LT is currently set higher, the initial reduction would be more than the 2% suggested above. This standardization will be carefully planned to minimize the impact of LT reductions on the likelihood of liquidations.

Specification

Chain Asset Current Expected LTV Recommended LTV Current Expected LT Recommended LT (1st stage) Recommended LT (Last Stage)
V3 Ethereum DAI 75% 0% 78% 76% 66%
V3 Arbitrum DAI 75% 0% 78% 76% 66%
V3 Optimism DAI 75% 0% 80% 76% 66%
V3 Polygon DAI 75% 0% 78% 76% 66%
V3 Gnosis WXDAI 75% 0% 78% 76% 66%
V3 Avalanche DAI.e 75% 0% 80% 76% 66%
V3 Metis m.DAI 75% 0% 80% 76% 66%
V3 Ethereum sDAI 75% 0% 78% 76% 66%
V3 Gnosis sDAI 75% 0% 78% 76% 66%
V2 Ethereum DAI 75% 0% 87% 76% 66%
V2 Polygon DAI 75% 0% 80% 76% 66%
V2 Avalanche DAI 75% 0% 80% 76% 66%
15 Likes

Thank you for the prompt analysis @ChaosLabs

From our perspective we would lend our voices to adopting a conservative approach while we take more time to determine how Aave should interact with DAI in light of these possible risk vectors.

5 Likes

I would like to thanks @ChaosLabs for this brillant analysis, in the light of their feedback, the ACI is now supportive of the “Conservative” option allowing a balanced & consensus solution.

sDAI is an attractive form of collateral as a intrinsic yield-bearing collateral, considering its current reasonable footprint in Aave and its business potential, we would suggest to leave it out this proposal.

12 Likes

I’m supportive of a conservative approach here.

As a user (and holder) of Aave, it would be disconcerting seeing such a rash change as 0% applied from the cuurent ~75%. An aggressive change doesn’t look necessary and it may cause users to think twice about borrowing with assets when their health factor drops.

Dear Aave Community,

As LlamaRisk continues to assess the collateral risk associated with Ethena’s USDe stablecoin, we are providing initial feedback ahead of the upcoming snapshot vote on updating DAI risk parameters. While a comprehensive risk assessment is still in progress, we share a number of concerns compiled below. Based on the analysis of these concerns, Ethena’s integration with Maker DAI does not pose an immediate systemic risk for Aave.

  1. High Collateral Concentration and Transparency Concerns: A significant portion of Ethena’s collateral is concentrated on centralized exchanges, particularly Binance, exposing the protocol to increased counterparty risks. Moreover, Ethena’s collateral is held by a few custodians, including Copper, Cobo and Ceffu. Greater transparency regarding verifiable collateral balances vs. circulating USDe would help assess overall risk.
    image
    Source: Ethena, April 7th, 2024
  2. Effectiveness of Delta-Neutral Hedging: Limited liquidity on centralized exchanges could pose significant challenges, particularly during market stress or volatility. If Ethena needs to execute large trades to maintain its delta-neutral positions or manage collateral, the lack of liquidity could result in substantial slippage and unfavorable execution prices. As Schmeling, Schrimpf, and Todorov (2023) highlight, the crypto futures basis (or “crypto carry”) can become very large and vary strongly over time due to the interplay between trend-chasing retail investors and the relative scarcity of arbitrage capital. During market downturns, this dynamic can lead to a breakdown in the effectiveness of delta-neutral hedging strategies. The effectiveness of Ethena’s approach relies on several factors:
    • Accurate pricing data: The hedging positions must be based on reliable and up-to-date price information to ensure proper alignment with the collateral’s value.
    • Efficient execution: Ethena needs to execute the necessary trades efficiently to maintain the delta-neutral position, minimizing slippage and transaction costs.
    • Sufficient market liquidity: The futures markets used for hedging must have adequate liquidity to absorb Ethena’s trades without significant price impact.
    • Management of funding rates: Perpetual futures contracts are subject to funding rates, which can impact the cost of maintaining the hedging positions. Negative funding rates could lead to a liquidity crunch and mass sell-off, threatening the stability of the peg.
  3. Underfunded Insurance Fund: Ethena’s insurance fund appears underfunded relative to the protocol’s scale. Based on Chaos Lab’s research of Nov. 2023 (Perpetual Future, Liquidity and Funding Rate Considerations for Ethena), during the most significant funding rate inversion event in the past three years (The Merge, September 2022), Ethena could have faced losses of up to 4.3% due to a prolonged negative funding period, potentially exceeding the insurance fund’s capacity. As of writing, the insurance fund stands at $32.6m, or about 1.6% of the total USDe supply. It is also worth noting that a significant amount of the reserve fund (about 11.3m at the time of writing) is comprised of Uniswap V3 USDe-USDT LP positions, which could trade below par-value in a drawdown scenario.
    image
    Source: Ethena (April 7th, 2024)
  4. Morpho Oracle Hardcoded to 1:1: Maker’s integration with Morpho Blue, through the DIRECT-SPARK-MORPHO-DAI loan facility supporting DAI-USDe/sUSDe isolated lending pairs, raises concerns. The Morpho oracle being hardcoded to a 1:1 ratio may not accurately reflect market conditions and risks. Moreover, the aggressive Liquidation LTV (LLTV) on some of Morpho Blue’s isolated markets (up to 94.5%, or leverage up to ≈18x) makes them more susceptible to bad debt and faulty liquidations, potentially leading to losses for Maker and impacting DAI’s stability.
  5. Leverage and Liquidity Risks: As Ethena grows, the ratio of available liquidity to protocol size becomes crucial. The main venues for secondary liquidity, such as Curve and Uniswap, are relatively small compared to Ethena’s market cap. The 7-day withdrawal queue on staked ETH collateral could amplify risks during market stress, potentially causing further instability in the secondary market.
    image
    Source: Curve and Uniswap V3 pools, April 7th, 2024
  6. Ethena’s Risk Management and Governance: Limited information regarding Ethena’s risk management practices, governance structure, and decision-making processes hinders stakeholders’ ability to assess the protocol’s capability to navigate challenges and maintain USDe stability.
  7. Legal and Regulatory Concerns:
    a. Regulatory stance on USDe: Ethena Italia S.r.l. holds a VASP registration in Italy, but this does not exempt it from additional regulatory requirements in other European countries where it may operate. The upcoming MiCAR regulation in the EU will introduce a unified scheme for crypto-asset service providers that Ethena needs to strategize for.
    b. Compliance with laws and regulations: Ethena must conduct thorough AML checks on funds and clients as a registered crypto service provider. However, current obligations still need to meet the full scope of consumer protection under MiCAR. Ethena must provide more comprehensive risk disclosures and documentation to align with evolving standards.
    c. Impact of regulations on USDe as collateral: If Ethena faces difficulties obtaining necessary MiCAR licensing, its operations with the USDe token could be deemed unauthorized, severely impacting protocols using USDe as collateral. Restrictions on offering yield on e-money tokens is another key aspect Ethena must navigate. Close monitoring of ESMA consultations and proactive engagement with regulators is advised to mitigate risks.

Despite the concerns raised, Ethena’s integration with Maker DAI does not pose an immediate systemic risk for Aave. Maker’s max exposure of 1B USDe is expected to remain below 20% of DAI’s total backing, and Maker has demonstrated resilience in handling risky collateral, including over 3B of off-chain real-world assets (see our DAI Exposure to Real World Assets report), without experiencing significant adverse consequences.

Based on these considerations, we believe the conservative option presented by Chaos Labs to gradually reduce DAI LTV parameters on Aave is prudent at this stage. That said, the risk environment is dynamic and warrants ongoing monitoring. As a general principle, we believe in taking proactive steps that sufficiently address and respond to changing risk factors in a way that is least disruptive to Aave users.

Moreover, market participants using leverage on staked USDe should exercise extreme caution due to the limited liquidity relative to USDe’s market cap, which could lead to depegging events during large redemptions.

8 Likes

We support an additional option for the community to vote on, with sDAI excluded from the conservative option outlined in the post.

1 Like

After gathering feedback from community and Risk Service Providers, the current proposal has been escalated to ARFC Snapshot.

It has been taken into account latest feedback, therefore there will be 3 options:

  1. Conservative
  2. Aggressive
  3. Abstain
    Voting will start tomorrow, we encourage everyone to participate.
3 Likes

And how about DAI emode parameters?

As Allez Labs shares its risk analysis on this proposal to set DAI LTV to 0, here is a recap of timelines and the reasons that have led to this proposal to mitigate DAI’s risks.

Timeline/Context

On March 21st BA Labs posted a risk assessment of Ethena and the integration with Morpho, the initial plan was to allocate 100M DAI across different (s)USDe collateral markets in Morpho Blue through the Spark DAI vault. On the same day Chainsecurity’s audit of the integration was published. The initial allocation was made on March 29th.

Collateral/LLTV 77% 86% 91.5% 94.5% Total
sUSDe 5M 10M 30M 5M 50M
USDe 5M 10M 30M 5M 50M

On March 21st, Block Tower made another proposal to directly purchase Ethena (s)USDe by reducing exposure to USDC and a legal assessment by Steakhouse Financial followed suit on March 28th.

On April 1st, BA Labs posted a second analysis regarding the Spark DAI vault and its allocation to (s)USDe backed markets, proposing the following parameters.
DDM-Params-Forum

Source: Maker forum

With some pushback on the forum, after only a day of discussion the proposal was approved for voting, less than a week later, on April 7th 2024, the proposals passed:

  • increasing the credit line up to 1B DAI
  • approving a direct swap of 50M USDC to (s)USDe through Maker’s affiliated RWA entities

In practice, a 1/2 multisig of anons now has power to increase the credit line by 100M DAI every 24 hours up to 1B DAI. However as defined in Maker governance they act upon recommendations from the Stability Scope Advisor which is BA Labs who at this stage recommends a maximum allocation of 600M DAI.

Following the approval, on April 8th, a direct deposit into Morpho was triggered bringing the total assets of the Spark DAI vault on Morpho to 200M DAI.

Collateral/LLTV 77% 86% 91.5% 94.5% Total
sUSDe 10M 20M 30M 10M 70M
USDe 10M 80M 30M 10M 130M

Increasing the allocation into riskier markets (94.5%) and into the non yield bearing USDe, this is in line with the adopted end allocations (below).

Collateral/LLTV 77% 86% 91.5% 94.5% Total
sUSDe 10M 50M 30M 10M 100M
USDe 10M 330M 150M 10M 500M

The timelines and decision making processes around the gradual allocations ((100M DAI - accumulated interest) per) have not been laid out clearly.

Did Maker follow its collateral onboarding frameworks?

There are two collateral onboarding frameworks in Maker, [MIP6] and the Spark collateral onboarding framework. The allocations in (s)USDe collateral markets on Morpho falls under the second framework.

Maker’s collateral onboarding frameworks are centered around a profitability analysis, while they have no direct control over the rate at which DAI is borrowed in Morpho Markets, being the dominant and often only player in these markets they have lots of room to keep it in line with expectations in normal market conditions.

While the framework was not intended for the Ethena type of derivative it explicitly states that amendments can be made for projects that prove scalable demand, undoubtedly Ethena falls under that category.

  • (s)USDe has not been listed on blue chip lending markets for more than 2 months (Compound/Aave as per the collateral onboarding framework)
  • the average market cap of the circulating supply of sUSDe/USDe has not been over 1B for over 3 months
  • no secondary liquidity analysis has been conducted, the conclusions around the primary market are not commensurate with the investment made, indeed it is recognized that there are a lot of unknowns and unquantifiables.

Similarly, while it is unclear under which framework the direct purchase of (s)USDe falls there is a swath of concerning accommodations made.

Maker will deploy 50M USDC into Ethena through its RWA entities, with preferential conditions:

The analyses provided and allocation decisions

As we understand it, the rationale for making the allocation into (s)USDe collateral markets is that the risk is worth the reward. In our opinion the understanding of the risks and reward are not commensurate with the investment made.

On risk

The analysis posted on Maker’s forum is a thorough deep dive into Ethena’s risks and was as comprehensive as can be in listing them. However the conclusions around the primary market redemptions, custodian risk, LST and hedging risks are admitted to be unclear at best.

There has been a shift from getting exposure the funding rate towards gaining exposure to ENA:

The allocation decision in the first analysis is centered around where user profitability/lending efficiency lies within the different Morpho collateral markets by taking into account leverage and funding rates without considerations of risks. This analysis de facto does not include USDe markets as they are non-yielding. The second allocation recommendation (April 1st) was solely based on past pool performance without consideration of the underlying market structure and liquidation curves.

As of April 3rd USDe pools were the riskier pools across the whitelisted markets, their collateral is non yielding and users were paying the highest rates, yielding a concerning time to liquidation because of accumulated interest.

Spark DAI allocation (on April 3rd), assuming 30% funding rate

Market DAI Allocation Collateral at risk (-10% drop) Current borrow rate 6 day avg borrow rate Time to liquidation Time to liquidation (6d avg rate)
USDe 77% 5000000 1170000 46% 27% 11w 19w
USDe 86% 10000000 6800000 62% 32% 8w 16w
USDe 91.5% 30000000 29200000 108% 74% 5w 7w
USDe 94.5% 5000000 5160000 63% 44% 8w 11w
sUSDe 77% 5000000 1550000 50% 31% 26w >52w
sUSDe 86% 10000000 3510000 44% 34% 37w >52w
sUSDe 94.5% 5000000 3410000 52% 39% 23w >52w
sUSDe 91.5% 30000000 21180000 43% 51% 40w 24w

The borrowing demand in these no-yield collateral markets is a consequence of Ethena points program which is expected to last until September. Additionally there was an expectation that funding rates would go up and that Maker will allocate more in these markets and drive the borrow rates on these leveraged positions further down.

Despite USDe collateral markets being the riskiest to deploy into, the second allocation is heavily weighted towards them.

Collateral/LLTV 77% 86% 91.5% 94.5% Aggregate allocation
sUSDe 10M (+5M) 20M (+10M) 30M (+0M) 10M (+5M) 70M (+20M)
USDe 10M (+5M) 80M (+70M) 30M (+0M) 10M (+5M) 130M (+80M)

This shows a clear shift of Maker’s exposure from the funding rates towards the ENA points program.

The speed of this new allocation is unprecedented

At a higher level more allocation will be made into a new protocol - Morpho Blue - and new collateral - (s)USDe - at a faster pace than Spark’s original credit line using battle tested software. There is no doubt that this is concerning.

On rewards

Unlike in Maker core vaults or in Spark, Maker has no direct control over the rate at which DAI is borrowed in Morpho Markets, however being the dominant player in these markets they can keep rates in line with expectations in normal market conditions by dynamically supplying and withdrawing.

The recent reallocation to USDe though underscores that operationally there is not yet a commitment to at all times maintain the aggregate return of the Spark DAI vault in line with reasonable risk premia and well above the DSR as stated in the second risk analysis.

The integration with Morpho and allocation to (s)USDe collateral markets are publicly depicted to be about gaining exposure to the funding rate. In practice the allocation decisions and concentration into USDe markets instead of the yielding sUSDe markets shows a pivot from gaining exposure to the funding rate towards gaining exposure to the ENA token and its points program.

Analysis from Maker should be provided on the expected premiums and reallocation decisions in light of the recent shift of focus towards indirect ENA farming value capture.

Maker’s Balance Sheet Analysis

An aggregate view of Maker’s balance sheet (April 8th)

Collateral DAI (M) DAI % Backing Stability Fee Risk scoring (Allez Labs Risk Score)
RWA 2,075 42.64% 100% CR 3.70% Low
(st)ETH core vaults 890 18.29% 609% CR 13.40% Low
PSMs 590 12.12% 100% CR 0% Low
Spark (Aave instance) 920 18.91% NA 13.30% Medium
WBTC core vaults 191 3.93% 358% 14.70% Medium
Spark (Morpho) 200 4.11% 113+% CR Variable High
Total 4,866 100%

Source: Makerburn.com

Excluding the Morpho instance, Maker charges on average 7.42% for DAI borrows and yields 13% on 1.53B DAI that is locked in the DSR leaving ~3.16% ( 150M annualized ) for expenses, buybacks etc. While the allocation in Morpho is risky, given it’s less than 5% of Maker’s current balance sheet it does not meaningfully change DAI’s risk profile, however as time progresses and the allocation grows to $600M it will tend to become more and more risky reaching 11.3+% of DAI fixing other BS items.

Assuming most of the flow of DAI to USDe is through the PSM, users taking leverage on (s)USDe using DAI will progressively decrease Maker’s RWA exposure from current ~55+% to ~45-% exposure in favour of (s)USDe CDPs.

Users can use DEX liquidity available on Curve to take leverage on (s)USDe using DAI and liquidity in the DAI/USDe and sDAI/sUSDe pools has been growing and recently surpassed $75M.

(s)DAI/(s)USDe DEX liquidity
Source: Allez Labs on Dune

However the capital efficiency of the PSM is unmatched and already a few hours after the allocation there’s been significant outflows from PSM-USDC-A, - 41%USDC in the last 48h.

PSM USDC Balance
Source: Allez Labs on Dune

There are of course other user, cyclicality and market dynamics at play explaining this move, but users taking leverage on Ethena is a huge part of this and will only be more pronounced as the allocation continues to grow.

InstantPSMUSDCOutflowRelatedToEthena
CumulativePSMUSDCOutflowRelatedToEthena
Source: Allez Labs on Dune (1), (2)

It’s reasonable to estimate that at full allocation (600M DAI), most of the liquidity used to take leverage will flow out of Maker’s RWA exposure. First flowing out of the PSM and later Maker will decrease its T-Bills exposure to replenish the PSM. Alternatively Maker and Ethena could work to ensure that there are enough incentives in size for DEX LPs to support the leveraging demand.

Similarly users deleveraging from (s)USDe are likely to go through the PSM and thus replenishing it over time.

While it is still early, borrowers in (s)USDe collateral markets on Morpho have not shown sensitivity to the cost of capital, this is likely to continue for the coming months. As a result the expected changes to Maker’s balance sheet are likely to sustain over that period with high risk DAI increasing from 4.1% to 17.6%

Conclusions

Whilst the introduction of Ethena as a collateral asset in the Maker ecosystem enhances Maker’s profitability, there is a mismatch between what is currently understood and the speed and size of this new allocation. As such, reasonable preventative measures should be taken on Aave while we continue to monitor the risks and Maker’s balance sheet.

DAI and sDAI in Aave

As pointed out there is north of 340M DAI/sDAI across Aave deployments, we do not think that completely offboarding either are commensurate measures to the recent changes in Maker’s risk profile, additionally for Aave to remain competitive we suggest treating DAI and sDAI differently.

DAI

As pointed out by @ChaosLabs DAI is mostly a borrow asset in Aave, maintaining it as a collateral asset only exposes Aave to tail risk events that are not compensated by the revenue it currently generates as collateral.

We are in support of any plan to transition DAI to become only a borrowable asset, i.e. decreasing slowly the LT. We suggest the LTV be set to 0 which does not affect existing users.

sDAI

sDAI is a yield bearing asset used only as collateral on Aave, it is not a significant revenue generator for Aave. However updating its LT and LTV to be non competitive or as offboarding sDAI is not in the favor of a large portion of Aave’s user base as suggested by @WintermuteGovernance. As a measure of its significance: Aave V3 Ethereum is the second largest “holder” of sDAI next to the Gnosis bridge.

In our opinion, updating the sDAI parameters to become completely uncompetitive is not a productive route. Within a reasonable cap sDAI as collateral, the risks are diversified, and it does not significantly increase insolvency risk for Aave.

We suggest lowering the supply cap of sDAI to 65M on V3 Ethereum as we continue to monitor Maker’s balance sheet without changing its 80% LT and 77% LTV.

Moreover as precautionary measure and for consistency the sDAI oracle - which has been left out of the CAPO updates in light of the changes of its compounding rate - should be updated to use the capped price of DAI instead of the chainlink price as a source for DAI_TO_USD, cf sDAI oracle and DAI oracle.

On the proposed options

As a practical matter there is a live ARFC snapshot from Chaos Labs, sDAI being excluded from the snapshot we are in favor of Option 1 which intends to decrease DAI’s LT and LTV progressively.

Overall both will decrease the risk in Aave, however our opinion is a measured approach in reducing the LT is a better solution without a thorough analysis of the affected user positions.

Ultimately we believe the above stated measures, lowering the LTV and LT for DAI while decreasing the supply cap for sDAI provide the best balance between reducing risk for the Aave protocol while maintaining competitiveness.

In this analysis Allez Labs highlights recent changes to DAI’s risk profile, changes which are set to accelerate hence the proposed precautionary approach. This is not a condemnation of Ethena which does present significant risks but also equally significant opportunities of which we will share analyses soon.

10 Likes

After snapshot monitoring, the current ARFC Snapshot has just ended, reaching both Quorum and CONSERVATIVE as winning option, with 533K votes.

Therefore, the ARFC Snapshot has passed with CONSERVATIVE option as elected choice.

Next step will be the publication of an AIP for final confirmation.

2 Likes