Overview

Lending protocols are the most important component within DeFi systems. However, different lending protocols have introduced various architectures due to considerations of governance mechanisms. This article will primarily discuss the architecture of lending protocols from the perspective of governance systems. From the starting point of this article, the focus will mainly be on the relationship between Collaterals and Liabilities and the factor configurations of lending protocols.

Collaterals and Liabilities

Monolithic

Monolithic remains the most mainstream lending protocol architecture. In this architecture, users can use multiple assets as Collaterals to borrow other assets. At the same time, the Collaterals deposited by users can also be borrowed by other users, meaning that the users’ Collaterals can also earn interest.

In the Monolithic architecture, to mitigate risks, some tokens can only be borrowed but cannot act as Collaterals. For example, in the diagram below, users can deposit wstETH and USDC as Collaterals to borrow LINK tokens, but they cannot use LINK tokens as Collaterals to borrow WBTC.

The lending protocol that best fits the Monolithic architecture is AAVE V3, currently the largest lending protocol on Ethereum. In AAVE V3, most assets can serve both as Collaterals and Liabilities (i.e., the “Normal” situation in the diagram below). AAVE V3 also features an Isolation Mode, where certain assets can only be used as Collaterals to borrow stablecoins but cannot be used as Collaterals to borrow other tokens.

The Monolithic architecture is relatively capital efficient. For ordinary users, they can directly deposit multiple tokens as Collaterals to borrow the assets they need, while the Collaterals also earn interest. However, the biggest issue with this architecture is that most tokens cannot be included as Collaterals. Once unstable value tokens are introduced as Collaterals, it could lead to significant losses for the lending protocol.

AAVE V3 has extremely strict token approval processes, so currently, AAVE V3 only supports around 30 tokens as Collaterals. Many of these 30 tokens are under Isolation Mode and cannot be used as general Collaterals.

Isolated Pairs

Isolated Pairs is another commonly used lending protocol structure. In this architecture, the Collaterals deposited by users can only be used to borrow a specific token. Additionally, the Collaterals deposited by users cannot be borrowed by others, meaning the Collaterals do not earn interest.

Currently, the third-largest lending protocol on Ethereum, Morpho Blue, uses this model. Within this protocol, there are a series of discrete markets, and users can choose to deposit Collaterals into a specific market to borrow the corresponding Liabilities tokens. However, since the Collaterals are not borrowed by others, users cannot earn interest on their Collaterals. The diagram below shows the market for WBTC and USDC, where the WBTC deposited by users can only be used to borrow USDC tokens, and the WBTC deposited will not earn interest.

Therefore, within Morpho Blue, there is another group of users who provide USDC to the lending market to earn interest. For example, users may inject USDC into the market shown in the diagram above, where WBTC is used as Collateral, to earn interest. Thus, in the Isolated Pairs system, users are divided into the following two categories:

1. Depositors, who provide Liabilities tokens to the market for other users to borrow, earning lending interest.
2. Borrowers, who provide Collaterals to borrow tokens but forgo the interest income on their Collaterals.

For depositors, in the Morpho Blue protocol, the same token may be borrowed in multiple markets. For example, there may be markets where WBTC is used as Collateral to borrow USDC, as well as markets where wstETH is used to borrow USDC. The interest rates paid for USDC in these markets are different. The diagram below shows the interest rates for markets where different Collaterals can be used to borrow USDC.

Therefore, if users want to deposit USDC and earn a higher interest rate, they need to allocate their USDC to different markets. This is not a simple task, as some markets may choose highly volatile Collateral tokens. By providing USDC to such markets, users may incur losses due to severe token fluctuations that could lead to liquidation.

As a result, a third type of market participant often emerges in the Isolated Pairs markets—Curators. Curators are typically professional financial institutions that analyze the risk and return of each market and design their own fund allocation strategies. Users can then delegate their USDC to Curators to manage based on the Curators’ fund allocation strategies and returns.

Another advantage of Isolated Pairs is that it completely avoids the governance issues found in the Monolithic architecture. Any user can create a lending market on their own without going through governance. For example, USD0++, as a new type of yield-bearing token, would find it difficult to enter lending protocols like AAVE, but it can directly establish a market within Morpho Blue.

From a developer’s perspective, Isolated Pairs also offers a significant advantage. Since this architecture does not involve the secondary lending of Collaterals, there is no need for code related to recording interest on Collaterals, making the overall implementation simpler. Morpho is currently one of the most popular lending protocols among developers.

Isolated Groups

This architecture is currently only used by Compound V3 within the lending protocol. Under this architecture, different Liabilities have different Collaterals.

A concrete example is in the mainnet version of Compound V3. If we want to borrow ETH, we can provide the following Collaterals:

Compared to the Isolated Pairs system, Isolated Groups allow users to provide multiple types of Collaterals for a single asset. This model theoretically enables developers to freely deploy lending markets and configure the relationship between Collaterals and Liabilities. However, the only protocol currently using this model, Compound V3, has still introduced a governance model based on COMP tokens. The creation of markets and modifications to Collaterals all require votes from COMP token holders.

A bit awkwardly, Compound has encountered two instances of code vulnerabilities. These two errors can be specifically referred to in the “Compound Errors“ and “Overcompensated“ reports written by Rekt. However, neither of these errors had an impact on the assets within the lending protocol, so Compound remains the fifth-largest lending system on the Ethereum mainnet.

Isolated Groups with Mutual Collateral

This lending protocol architecture is also rarely used. Currently, the 18th-ranked lending protocol on the DeFiLlama TVL leaderboard, Silo Finance, utilizes this architecture.

The diagram above shows the lending relationships in the Silo Arbitrum version, specifically the borrowing and lending interactions of Silo tokens. Under the Isolated Groups with Mutual Collateral architecture, tokens are interlinked and can act as Collaterals for each other. In the Silo Arbitrum version, all lending markets contain ETH and USDC tokens, along with a third token specified by the market creator, which in this case is the Silo token.

In the Silo documentation, ETH and USDC are referred to as “bridge assets,” while the third token designated by the market creator is called the “base asset.” The naming of bridge assets makes sense because users can create chain-like lending paths based on ETH and USDC to meet their needs. For example, users can use Silo tokens as Collateral to borrow ETH and then use ETH as Collateral to borrow ARB. Through this path, we essentially achieve the goal of borrowing ARB using Silo tokens.

The advantage of Isolated Groups with Mutual Collateral is that it combines the risk isolation of Isolated Groups, allowing users to choose lending markets based on their individual risk preferences and requirements. It also helps to some extent mitigate liquidity fragmentation, as users can design chain-like lending paths using bridge assets to indirectly achieve certain borrowings. Finally, due to the Mutual Collateral relationship, the Collaterals deposited by users can earn interest.

However, Silo also allows users to configure Collateral Status. This option enables users to prevent their Collaterals from being borrowed out, ensuring the safety of the Collateral, but at the cost of forgoing the interest earned on the Collateral.

Parameters

In the second part of this article, we will discuss the parameter issues within lending protocols, focusing mainly on the Collateral Factor, which is most relevant to the capital efficiency of lending protocols. This parameter measures the discount applied to Collaterals within the lending protocol. The higher the Collateral Factor, the lower the discount applied to the asset within the protocol. Generally, stable assets like ETH have higher Collateral Factors, while low-market-cap assets tend to have lower Collateral Factors.

Global Paternalism

Global Paternalism is the most common Collateral Factor parameter configuration model within lending protocols. Simply put, it means that the protocol itself sets and modifies the Collateral Factor parameters. For example, within AAVE, the DAO (Decentralized Autonomous Organization) of AAVE proposes changes to the parameters of certain Collaterals. The diagram below shows the parameters configured by AAVE governance when ezETH is used as Collateral.

For both the protocol and users, Global Paternalism is the simplest solution.

The Invisibal Hand

Currently, lending protocols like Morpho, which do not introduce a governance system, cannot implement Global Paternalism. Therefore, Morpho uses a different approach. In this approach, any user can create a market with any Collateral Factor parameters, and depositors will place their assets into the markets they deem most reliable.

For example, in Morpho, we can see two markets with different parameters for the USD0++ / USDC pair, and users will deposit their USDC assets into different markets based on their risk preferences.

Conclusion

This article mainly introduces the common architectures of DeFi lending protocols. In simple terms, different lending protocols have emerged due to varying dependencies on governance. We can broadly divide lending protocols into two categories:

1. Strongly Dependent on Governance: Lending protocols where parameters and collateral choices require strict governance processes.
2. Not Dependent on Governance: Lending protocols that allow for the free creation of markets and parameter setting, with users “voting with their feet.”

For protocols that are strongly dependent on governance, the governance mechanism is the core element. Compound and its forked projects have encountered governance issues multiple times, which led to protocol problems. Currently, AAVE has the most excellent governance mechanism among all protocols. The major advantage of protocols strongly dependent on governance is that, under the premise of trusting the protocol’s governance mechanism, depositors do not need to perform any other operations beyond depositing assets. This is particularly attractive to organizations with a foundation structure, like the Ethereum Foundation.

For protocols that are not dependent on governance, everything is decided by the market. This raises higher requirements for participants in lending, as both depositors and borrowers need to pay attention to market dynamics and frequently reallocate funds in different lending markets to obtain the best rates. This is extremely unfriendly for foundation users, who are unable to frequently operate funds due to the limitations of the mechanism. Of course, users can choose to delegate their funds to a third party, but currently, the third parties within Morpho are not more trustworthy than AAVE.

Disclaimer：

1. This article is reprinted from [Techflow]. The copyright belongs to the original author [Techflow]. If you have any objection to the reprint, please contact Gate Learn Team, the team will handle it as soon as possible according to relevant procedures.

2. Disclaimer: The views and opinions expressed in this article represent only the author’s personal views and do not constitute any investment advice.

3. Other language versions of the article are translated by the Gate Learn team, not mentioned in Gate.io, the translated article may not be reproduced, distributed or plagiarized.