[ARFC] Stablecoin Interest Rate Curve Update - 03.04.2025

[ChaosLabs]

Overview

In light of recent market changes, Chaos Labs recommends setting the target borrow rate for most stablecoins to 6.50%. These changes would be implemented via the direct-to-AIP process.

Motivation

The utilization for stablecoins across Aave has resumed falling after a brief improvement following the last IR curve adjustment, calling for a further reduction in target rates to better align with the market-priced interest rate. The chart below shows the median stablecoin utilization on each Aave instance, with the lines weighted by the total borrows in each deployment.

A closer look at the largest stablecoin markets indicates that borrow rates have stabilized between 6% and 8%.

As a result, we recommend setting the target rate for all native stablecoins to 6.5%.

Please note that GHO, USDS, pyUSD, and bridged stablecoin versions are to be managed separately and are not included in these recommendations. Additionally, after syncing with Growth SPs, we will refrain from performing changes to the Polygon Instance.

Specification

Protocol	Instance	Asset	Current Slope1	Rec. Slope1
Aave V3	Ethereum Core	USDC	9.50%	6.50%
Aave V3	Ethereum Core	DAI	9.50%	6.50%
Aave V3	Ethereum Core	USDT	9.50%	6.50%
Aave V3	Ethereum Core	LUSD	9.50%	6.50%
Aave V3	Ethereum Core	FRAX	9.50%	6.50%
Aave V3	Ethereum Core	crvUSD	9.50%	6.50%
Aave V3	Ethereum Core	USDe	9.50%	6.50%
Aave V3	Ethereum Prime	USDC	9.50%	6.50%
Aave V3	Ethereum EtherFi	USDC	9.50%	6.50%
Aave V3	Ethereum EtherFi	FRAX	9.50%	6.50%
Aave V3	Arbitrum	DAI	9.50%	6.50%
Aave V3	Arbitrum	USDT	9.50%	6.50%
Aave V3	Arbitrum	LUSD	9.50%	6.50%
Aave V3	Arbitrum	USDC	9.50%	6.50%
Aave V3	Arbitrum	FRAX	9.50%	6.50%
Aave V3	Optimism	DAI	9.50%	6.50%
Aave V3	Optimism	USDT	9.50%	6.50%
Aave V3	Optimism	sUSD	9.50%	6.50%
Aave V3	Optimism	LUSD	9.50%	6.50%
Aave V3	Optimism	USDC	9.50%	6.50%
Aave V3	Base	USDC	9.50%	6.50%
Aave V3	Metis	m.DAI	9.50%	6.50%
Aave V3	Metis	m.USDC	9.50%	6.50%
Aave V3	Metis	m.USDT	9.50%	6.50%
Aave V3	Avalanche	DAI.e	9.50%	6.50%
Aave V3	Avalanche	USDC	9.50%	6.50%
Aave V3	Avalanche	USDt	9.50%	6.50%
Aave V3	Avalanche	FRAX	9.50%	6.50%
Aave V3	Avalanche	AUSD	9.50%	6.50%
Aave V3	Gnosis	WXDAI	9.50%	6.50%
Aave V3	Gnosis	EURe	9.50%	6.50%
Aave V3	Gnosis	USDC.e	9.50%	6.50%
Aave V3	BNB	USDC	9.50%	6.50%
Aave V3	BNB	USDT	9.50%	6.50%
Aave V3	BNB	FDUSD	9.50%	6.50%
Aave V3	Scroll	USDC	9.50%	6.50%
Aave V3	ZkSync	USDC	9.50%	6.50%
Aave V3	ZkSync	USDT	9.50%	6.50%

Next Steps

We propose making these updates via the direct-to-AIP process.

Disclosure

Chaos Labs has not been compensated by any third party for publishing this ARFC.

Copyright

Copyright and related rights waived via CC0.

[LlamaRisk]

Summary

LlamaRisk supports these adjustments. Given the persistently tight market conditions, aligning the protocol more closely with broader rates is essential. Lowering slope1 would help achieve this while mitigating further borrow volume contractions. However, due to persistent volatility, rates are expected to vary more. Therefore, it is important to continuously monitor the evolving utilization rates, as @ChaosLabs has been proactively doing.

At the time of writing, the utilization rate for stablecoins remains below the optimal threshold. Below, we provide a breakdown of how borrow rates would be impacted across major stablecoin markets:

Asset, Market	Current Borrow Rate, APY	New Borrow Rate, APY
USDC, Core	6.38%	5.01%
USDT, Core	6.4%	5.08%
DAI, Core	7.98%	6.23%
USDC, Arbitrum	6.84%	5.36%
USDC, Base	7.1%	5.55%
USDC, Avalanche	6.89%	5.4%
USDT0, Arbitrum	7.79%	6.09%

Regarding GHO, with the current borrow rate ranging from 4.51% to 6.45% on Core, the proposed changes would keep it aligned with — or slightly discounted compared to — the adjusted stablecoin borrow rates. As a result, no immediate action is recommended for Aave’s stablecoin.

Disclaimer

This review was independently prepared by LlamaRisk, a community-led decentralized organization funded in part by the Aave DAO. LlamaRisk is not directly affiliated with the protocol(s) reviewed in this assessment and did not receive any compensation from the protocol(s) or their affiliated entities for this work.

The information provided should not be construed as legal, financial, tax, or professional advice.

[stani]

Thanks @ChaosLabs for putting forward the proposal and @LlamaRisk for the comments. Given the market dynamics of interest rates in DeFi, a faster checkpoint would favour a better outcome for the users (especially given the governance delays). The demand has decreased since early February, I personally prefer to see more faster turnaround on rate optimizations as they mostly affect business side of the protocol. I understand that i might not always be easy when the market fluctuates and more data is needed but the objective should be to implement optimizations as quickly as possible.

[ChaosLabs]

At Chaos Labs, we are committed to ensuring Aave’s markets remain efficient by keeping utilization close to UOptimal. At the same time, we highly prioritize the user experience on Aave.

Since the demand started decreasing in December, we have proactively proposed and executed multiple interest rate reductions to align rates with market averages. While it may be tempting to undercut the market or predict future rate trends, doing so comes with significant risks. Reducing rates too aggressively can lead to overutilization, causing highly volatile borrow rates that ultimately degrade the borrower experience and drive users to alternative platforms.

To resolve this issue and drastically improve Aave’s responsiveness to market conditions, we are pioneering the use of Interest Rate Risk Oracles. This solution will enable real-time interest rate updates and allow Aave to adapt dynamically to market demand. By continuously monitoring market conditions and adjusting rates accordingly, Risk Oracles will maximize pool utilization, hence driving up suppliers’ rates while also mitigating the risks of overutilization and sudden interest rate spikes.

In the meantime, the Risk Steward Parameter Change ARFC provides a stepping stone toward greater flexibility by reducing governance time for interest rate updates. With the adjustment of range for Slope 1 to 2% from the previous 0.5%, this will allow for faster reaction while maintaining minimal risk.

We appreciate the ongoing discussions and are committed to refining these mechanisms to benefit both users and the protocol’s stability.

[ChaosLabs]

In order to further expedite the IR update process, while we still recommend the execution of this AIP and the reduction of Slope 1 to 7.5% across the board, we will proceed to implement preliminary 0.5% changes to the assets to the extent permitted by the Risk Stewards.

Acknowledging the dynamic nature of rates, we continue to monitor and provide recommendations as needed.

[stani]

Appreciate the rate cuts from December however the reactivity hasn’t been fast enough to capture the existing market environment. Hence the ask isn’t to change rates aggressively instead to proceed faster on the actions needed to take the rates to correspond the existing market conditions into prod.

Borrowers can be less sensitive for short term price movements of interest as long as the borrow costs are adjusted promptly as well. For more predictability for borrower costs there is of course GHO as well.

Regarding the interest rate oracles, I do think the right trend is towards automatization. However I do have one specific concern related how the business decisions would be involved into any automatization that the protocol will adopt. It’s true that interest rate decisions do relate to risk pricing, however, interest rates should be widely driven by business decision of the DAO. What kind of pricing the DAO wants to apply to facilitate business. We have very good people now at the DAO with great business sense and it would be important to ensure that any interest rate oracle automatization is based both on risk and business.

[TokenLogic]

To date, the sole application of Interest Rate Risk Oracle has been performing sub-par and furthermore, the Risk Oracle overwrites the Risk Stewards if a manual update is pushed. The spike in wETH Borrow Rates due to a temporary improvement in stETH staking yields triggered a refinancing event that led to AUM, that had taken months to secure, moving away from Prime. This event discourages the builders from moving the capital back to Prime due to the uncertainty the automated methodology.

The Prime wETH reserve, where Interest Rate Risk Oracle is being trialed, is facing lower utilisation and Aave DAO is generating less revenue as a result. Whilst a lot of effort went into generating a native wstETH yield on Prime, the wETH utilisation on Prime is less than Core where the Interest Rate Risk Oracle is not implemented and the wstETH native yield rate is lower than Prime.

Base upon our observations, a three day cadence on the Risk Steward is not the blocker. The amount of time between updates can be improved, adjusting the Slope1 from 9.50% to 7.50%, takes x4 0.50% updates and 9 days to fully implement. This thread is 9 days old and only 1 update has been progressed.

[ChaosLabs]

From our perspective, the Interest Rate Risk Oracle is not solely about increasing AUM, but rather about ensuring Aave’s revenue remains strong, maintaining a balanced risk framework, and reducing governance overhead. There are a few key areas where we have a different view:

Operational Overhead of Manual Updates

We’d like to highlight that the claim that this thread is 9 days old with only one update progressed is inaccurate. The thread was actually 6 days old at the time of the response, and two updates have been successfully implemented, aligning with the expected timeline given the 3-day timelock for Risk Stewards’ changes.

Furthermore, it is unreasonable to expect these 0.5% changes to be performed manually over more than 50 assets across 14 different Instances every 3 days. This creates significant operational overhead for all parties involved with drafting, creating, and signing the transactions, consequently slowing down other necessary governance processes.

This emphasizes the need for Interest Rate Risk Oracles for the Aave DAO to handle this process more efficiently and minimize the operational overhead.

As the model’s inner workings and logic are explained on the forum, as they were here, and hence are readable and understandable by users, it creates a further layer of predictability for borrowers. The algorithm can also be adjusted to consider multiple factors, making it highly flexible and allowing for an optimal balance between Risk and Business decisions.

wstETH Utilization vs. Revenue

It is suggested that lowering wstETH supply yields led to an increase in utilization. While that may be true, it ignores the direct impact on revenue generation. As revenue is both a function of utilization and interest rate, maximizing utilization by continuously decreasing the interest rate, such as what was done for wstETH, did not mitigate the drop in revenue.

As it is visible from the chart below, the weekly revenue derived from LRT/wstETH looping within the Prime instance has reduced from $9K a week to $2K while maintaining the risk from the LRT assets unchanged and limited liquidity within the wstETH pool.

If there is no demand for lower rates, reducing them further is not a viable long-term strategy. The primary goal of the Prime instance should not just be attracting capital but ensuring sustainable and safe revenue generation for Aave DAO.

LRT Market-Wide Effects

While the algorithm for the Interest Rate Risk Oracle dynamically adapts to market conditions, it is still affected by the broader environment, and the recent outflow of LRTs has affected the market driving a constant downward pressure on wstETH supply rate and hence on the market utilization.

The claim that the AUM that left following the spike in wETH borrow rate had taken months to secure is inaccurate, as over 83% of the AUM that left following the spike, was driven by automated systems, of which Instadapp represents the majority share with over 77% of the funds being withdrawn. A significant portion of the funds that were withdrawn later returned following the adjustment of the curve.

These algorithms used by Instadapp to dynamically allocate the wstETH supply to the optimal markets also emphasize how the majority of the Prime Instance liquidity is both highly concentrated and highly mercenary, hence extremely sensitive to minimal changes.

wETH Collateralization is a New Factor

A major factor that led to the halting of the Prime instance growth was the inability for WETH to be used as a collateral within the market, and while this factor has now being resolved, the time since the change has been limited and hence the full impact of the change is still unfolding. We expect the future growth of the Prime Instance, now that stablecoin liquidity is present and WETH is enabled as collateral, to support the growth in demand from wstETH users going forward.

[ChaosLabs]

Over the last months the stablecoin supply across the instances has remained stable and has even continued to grow on selected chains. However, the borrow demand for stablecoins decreased following a reduction in price of the collateral assets such as Ethereum.

This drop in borrow demand caused an additional reduction in average Interest Rate since the original recommendation, hence we reduced the target Interest Rate for all stablecoins by an additional 1%, bringing the target IR for native stablecoins to 6.5%.
We updated the original recommendation to reflect these changes.

[sid_areta]

Reading through this discussion (and please correct me if I’m wrong), it seems like there is a trade-off here between “awkward” rate changes that occur frequently and may be uncomfortable for borrowers, causing adverse impacts, vs. speed and reactivity, which allows Aave to always be competitive in the market and move quickly.

We’re more inclined towards interest rate risk oracles with limits on the size of the movement. However, we think TokenLogic’s point regarding the refinancing effect and impact on revenue is valuable. Historical data such as this may be useful inputs into the risk oracle and its recommendations, and over time, it could become “smarter” and evolve from solely being based on the Lido staking rewards minus the buffer. It would be good to understand if this is possible.