Compound III is the state-of-the-art lending protocol deployed by Compound governance in August 2022. It is most notably unique from competing collateralized lending protocols because each market instance has a single borrowable asset.
What does this actually mean?
→ On Compound v2, a user who lends WBTC can borrow UNI, DAI, and other assets against it.
→ On Compound III, a user who lends WBTC can borrow USDC only.
So at first glance, Compound III looks like a slimmed-down version of Compound v2. Why did Compound use this design?
The answer is: Compound III is about simplicity. It strips away unnecessary features and use cases in order to optimize.
> “By removing every unnecessary feature and use-case, upgrading the risk engine (and capital efficiency), and focusing on a single borrowable asset, the protocol has the potential to be the safest & most appealing tool for borrowers ever designed.”
- rleshner
Compound v2 enabled borrowing 20+ assets, but at the cost of:
Security. Long tail assets provide vectors for attacks – in 2022 Mango Markets was hacked and Aave incurred bad CRV debt through oracle manipulation of less liquid assets.
UX. The Compound v2 dashboard has 27 APYs. The Compound III dashboard has 2.
And for what? While v2 users could borrow long-tail assets, nearly 95% of borrow volume came from borrowing stablecoins, followed by ETH and WBTC. So, Compound III’s design optimizes for the main use case demonstrated by the actual behavior of users.
Compound has deployed two markets on mainnet today – one market where users can borrow USDC, and another instance where users can borrow ETH. And so, while users on Compound III can only borrow USDC or ETH today, these two assets should cover the majority of the borrowing volume observed on Compound v2.
In the rest of this post, we will provide a deeper look at the technical concepts that make Compound III special, along with an analysis of Compound III’s growth and traction to date. Finally, it will provide several ideas and opportunities for where the protocol might expand in the future.
Key concepts
Compound III aims to shake up the status quo in DeFi lending. There are three key concepts that differentiate it from its competitors.
Concept #1: Users only borrow one Base Asset per each market.
The first thing that stands out when looking at Compound III is that each market has only one asset that can be borrowed – the base asset. This is new and different from Compound v2 and Aave v2/v3.
To illustrate this point, take the Compound III cUSDC market on mainnet as an example. In the cUSDC market, users supply one or more of the 5 collateral assets (WBTC, ETH, COMP, UNI, or LINK), and can borrow only USDC (the base asset) against their collateral. Note that there can be multiple markets on each network; each market is isolated, has its own set of collateral assets, and has a single base asset. For instance, there is a separate mainnet cWETH market where users can provide cbETH or stETH to borrow ETH.
As a consequence, Compound III markets have several notable and unique properties:
Collateral does not earn interest
There is only one Earn APR for each market. In the cUSDC example, this is the interest rate USDC lenders earn.
There is only one Borrow APR for each market. In the cUSDC example, this is the interest rate borrowers pay when borrowing USDC against their collateral.
Concept #2: Lend and Borrow rate curves are separate and independent.
Unlike Compound v2, the lender interest rate is not derived from the borrower rate. While it’s expected that earn and borrow rates should be highly correlated, the earn and borrow interest rate curves are independent and can be set separately.
What does this mean? Rate curves can be optimized for specific use cases or business cases. In fact, the mainnet cUSDC market has done this. cUSDC rate parameters are set such that the protocol subsidizes rates at low and/or high levels of utilization.
This is part of what enables lower USDC borrowing costs on Compound III vs. Compound v2, as analysis in this excellent report produced by Nethermind demonstrates.
The fact that rate curves are independent and highly configurable is a unique capability in Compound governance’s “toolbox”. Compound governance will likely adjust and experiment with interest rate curves as time progresses (see recent forum discussion for example).
Concept #3: Reserves accrue to protect against bad debt and to subsidize markets at high utilization.
A portion of the spread between the interest paid by borrowers and the interest earned by lenders is directed towards Compound III’s Reserves – funds stored in each pool that can protect users from bad debt and/or be withdrawn by Compound governance.
Reserves are expected to grow over time, but they can also be drawn from to cover bad debt or to subsidize rates (as mentioned in the section above).
For an example of what subsidizing rates look like, observe the cUSDC rate curves at high utilization levels (see screenshot below). As user monetsupply notes in this forum discussion, reserves are drawn upon when cUSDC utilization is below ~50% or above ~85%.
Reserves can also be withdrawn or used through the governance process (i.e., COMP governance can transfer assets to the DAO treasury).
This post will not dive into liquidation mechanisms, but reserves can also decrease or increase through the liquidation process. For a description of how Compound III liquidations work, the OpenZeppelin audit provides a concise, easy-to-follow overview of the process.
Other features worth noting
Compound III enables a buffer so that borrow positions may not be immediately liquidated. In Compound v2, when a position exceeded its collateral factor, it could be liquidated in the next block. Compound III makes max. collateral factor and liquidation collateral factor two separate parameters. This enhancement means that a potential buffer now exists such that positions can be topped up before they are liquidated.
Improved position management for users. Users can stack multiple lend+borrow transactions into a single transaction using the Bulker contract. Additionally, users can delegate another address to manage positions on their behalf.
Go-To-Market analysis
Compound III activity has grown steadily to >$800mm in TVL (deposits + collateral) and ~$250mm in borrows today (as of 1 June 2023).
Four Compound III markets have been deployed:
The cUSDC market on Ethereum mainnet accounts for >80% of TVL borrowing on Compound III deployments, as shown in the charts below:
Mainnet cUSDC
The cUSDC market on Ethereum mainnet was the first Compound III instance deployed and launched in Aug 2022.
The market enables users to borrow USDC against five assets: ETH, WBTC, UNI, LINK, and COMP. Lenders supply USDC and earn a yield on their USDC.
Unsubsidized USDC borrowing costs have been ~4%, but after adjusting for COMP rewards the borrow costs have consistently been <1% since launch.
The top three lenders supply >40% of USDC in this market, with the largest lender 7 Siblings (0x741) supplying ~$60mm USDC. Lenders have earned ~3% USDC yields on average and do not receive COMP rewards for lending.
Borrowing is also dominated by large addresses, with the top five borrowers comprising >50% of total borrows. Most borrowers are borrowing against Bitcoin and Ethereum collateral, with WETH and WBTC together making up ~90% of the collateral value (~$380mm of ~$420mm). LINK, UNI, and COMP collaterals generate <10% of borrowing activity.
Given Compound III’s ethos of simplifying and reducing risk, one wonders could be better if was limited to WETH and WBTC collateral only?
In an environment where Aave and even Curve’s crvUSD enable borrowing against long-tail assets, Compound III could stand out as the only market where lender positions are backed entirely by loans collateralized only by the majors – WBTC and WETH.
Mainnet cWETH
The cWETH market on Ethereum mainnet was deployed in Jan 2023 and enables users to borrow ETH against ETH liquid staking derivatives (LSDs): Lido’s stETH and Coinbase’s cbETH. Lenders supply ETH to the market and earn yield in return.
Earn yields for lenders have typically run ~2-4% (inclusive of COMP rewards) and have trended down since February. Borrowers typically paid ~3-5% to borrow ETH. COMP rewards have been provided to lenders, but not to borrowers.
The cWETH market has been used primarily as a venue for borrowers taking a levered long bet on cbETH. As shown below, cbETH typically comprises > 90% of collateral in this market.
cWETH Total Collateral Supply
Borrowing activity has been dominated by a few large users, who all exhibited similar behavior in borrowing ETH against LSDs. The behavior is as follows:
deposit cbETH/stETH as collateral → borrow ETH → swap ETH to cbETH/stETH → (repeat cycle)
The wallets following this strategy have begun closing out this trade, and it’s likely this market will continue to shrink. Most entered these positions when cbETH price was <1.00 ETH. They were betting cbETH price would restore towards its implied value (>1.03 ETH) after dropping to <0.99 ETH at times in Feb and March.
After the Shapella upgrade was announced and completed, cbETH price moved upwards to >1.02 ETH, and cWETH collateral supply shrunk considerably as several large borrowers repaid their debt and removed cbETH collateral (see 0xccfa’s 5 transactions withdrawing ~20k ETH on 3 April).
Taking a look at lenders supplying ETH to this market, again we find large lenders like 7 Siblings. The five largest cWETH suppliers account for >75% of the ETH supplied, with 7 Siblings supplying ~14k ETH (~35% of the total 40k supply).
Also worth noting is that ETH lenders in this market earn significantly less vs. what they could earn holding LSDs themselves. cWETH lenders earn ~2% total yield (~1.4% ETH yield, plus ~0.7% in COMP) while stETH and cbETH return between 4-8%.
This begs the question: why are these lenders taking on significant LSD exposure when they are earning well below staking yields1
It’s not clear where future growth for the cWETH market will come from. Historical borrow demand seems to have been driven by users expressing a position to go long on LSDs that were below the peg. Early lender demand came perhaps because of COMP farming subsidies on the earn side. Going forward, it may be tough to convince lenders to supply ETH to this market when they could earn >5% yields simply by holding stETH/cbETH/rETH or even staking with a centralized party.
Polygon cUSDC
The cUSDC Polygon market is the most recent deployment, launching in March 2023.
It has not won adoption overnight. Aave launched v2 and v3 markets on Polygon in early 2022 and still dominates the lending category on Polygon today. Compound v3 is a distant second, as shown in the DefiLlama ranking below.
Market dynamics for the cUSDC market on Polygon are currently heavily skewed by COMP incentives. As one can see, borrowing costs on this market are currently negative (borrowers earn ~4%) when taking COMP incentives into account.
Similarly, USDC earn rates are currently ~5% (inclusive of COMP rewards), which is high compared to similar opportunities available on Polygon (Aave v3 USDC = 2.3%, Aave v2 USDC = 0.7%). Given these observations, it’s surprising that Compound v3 hasn’t seen USDC supply grow higher than $15mm on Polygon. This implies that this pool is not well known and that DeFi markets on Polygon are not very efficient.
Given cUSDC’s higher yields relative to Aave, Curve, and Stargate on Polygon, it is reasonable to believe that the Polygon cUSDC market can grow substantially as awareness of this market grows.
Arbitrum cUSDC
Compound III launched on Arbitrum in May 2023. Users can borrow USDC against ARB, GMX, WBTC, and WETH collaterals via this market.
Arbitrum cUSDC presents more of a blue ocean strategy than previous Compound III deployments. By listing ARB and GMX as collateral Compound III is a first mover on Arbitrum, offering something that doesn’t already exist on the network. Of the two largest competing protocols on the market, Radiant does not list GMX as collateral and Aave accepts neither ARB nor GMX as collateral.
It’s probably too early to judge traction on this market. In the first two weeks ~$8mm USDC has been supplied, putting this market at a similar size to the Polygon USDC market. On the other hand, utilization is still <10% indicating that there isn’t much demand from borrowers, even at interest rates <2%.
Future opportunities
The Compound community has expressed its intention to deploy on many EVM-compatible chains and L2s by first launching “fragmented markets”, which could later be connected into cross-chain liquidity pools.
It has announced plans to deploy on Optimism after the upcoming Bedrock upgrade (ETA June 2023) and to deploy on Base later this year.
Beyond the cross-chain strategy, Compound III could explore other growth areas. The playbook used for deploying mainnet cWETH, which provided leverage against cbETH and stETH, could be replicated for other yield-bearing tokens.
Tokenized US treasuries and other tokenized securities are a growing area within DeFi, and Compound is well-positioned to serve these markets. As one idea, Compound III could create a cUSDC-USTreasury market where borrowers could borrow USDC against various tokenized treasuries collateral and lenders could easily tap into US Treasury yields. This could be a viable competitor (or perhaps collaborator) to Ondo’s Flux lending market; Flux’s fUSDC tokens are already being considered for collateral onboarding on Aave.
Beyond tokenized treasuries, Compound III could also serve long-tail tokens by launching markets focused on thematic opportunities or specific yield-bearing tokens. Potential examples could be markets where users borrow USDC against tokens like Yearn’s yvUSDC, borrowing ETH against yETH, and even borrowing USDC/ETH against more exotic tokens like AMM LP tokens.
Conclusion
Compound III is a unique infrastructure that offers borrowers arguably the “safest & most appealing markets”, while also providing lenders a simple, reliable experience. However, it’s unclear where Compound III fits into the large (and highly competitive) collateralized lending market.
Aave markets will typically be more capital efficient than Compound because the underlying collateral on Aave can be borrowed while Compound III collateral cannot. MakerDAO has battle-tested CDPs which have let users borrow DAI against isolated collateral for years. Spark Protocol has the potential to combine the best parts of Aave and MakerDAO, letting users borrow DAI directly from MakerDAO at market-leading rates. Curve crvUSD is another viable competitor, as its liquidation engine is friendly to borrowers (“soft liquidations”) and its underlying mechanics increase liquidity within the already robust Curve ecosystem.
Compound III can stand apart from competitors by offering simpler and safer markets than its competitors, and also, by offering exotic (but clearly isolated) markets that extend Compound’s reach. The Compound III primitive has not been adopted widely by the market today, but with experimentation and expansion into more EVM chains and collaterals, it could bring excitement back to the Compound ecosystem over time.
If you enjoyed this post, please share and subscribe!
The 7 Siblings wallet has a $550mm stETH position as collateral on MakerDAO. The wallet’s $80mm ETH position on Comp III is small in comparison, so perhaps this wallet and other large lenders use Compound III for the purposes of diversification (not putting all ETH directly into stETH) and farming COMP rewards.