Lesson 2 — Lending markets: Aave, Compound, Morpho
Lending protocols are interest-rate curves, collateral factors, and liquidation engines bolted together. Today: how each piece works and when positions break.
DeFi lending markets look simple from the outside — deposit collateral, borrow against it — and hide most of their interesting behavior in the parameters. Interest-rate models, oracle-driven valuations, liquidation thresholds, and isolation modes all decide whether your position survives the next market move. This lesson takes you from 'I have a position' to 'I can predict exactly when this position breaks.'
**Aave** is the largest pool-based lending protocol. Users deposit into a shared pool for each asset and receive aTokens that accrue interest in real time. Borrowers draw from the same pool, paying interest set by a kinked rate curve: low rates until utilization reaches the 'optimal' point (typically 80–90 percent), then sharply rising rates above that to incentivize repayment. Each asset has its own collateral factor (Aave calls it Loan-to-Value or LTV), liquidation threshold (the point at which the position becomes liquidatable, slightly above the LTV), and liquidation bonus (the discount liquidators receive on the seized collateral, typically 5–15 percent). Aave V3 added 'isolation mode' (riskier assets can only be borrowed against each other, not added to the main pool's collateral basket) and 'efficiency mode' (eMode — for correlated pairs like stETH/ETH, the protocol allows higher leverage because the price risk is lower).
**Compound** pioneered the pool-based lending model and operates similarly to Aave at the user level, with some structural differences. Compound V3 ('Comet') uses a single-borrowable-asset design: each market has one asset users can borrow (e.g., USDC) and many supported collateral assets. This simplifies risk modeling and lets the protocol be more capital-efficient with riskier collateral. Compound's interest rate model is also kinked but with different parameter curves; the practical difference is mostly that Compound's rates can be more volatile near the kink because of the way utilization is calculated.
**Morpho** sits on top of Aave and Compound in its 'Optimizer' design, matching peer-to-peer where possible to give both lenders and borrowers better rates than the underlying pool. Morpho's standalone 'Blue' design is a minimalist primitive: each market is a single loan asset, single collateral asset, fixed parameters, no governance over individual markets. Morpho Blue's safety case is the inverse of Aave/Compound: instead of having governance manage a large basket of assets, governance is removed from the market level and risk is isolated. Vaults built on top of Morpho Blue (managed by curators like Gauntlet, Steakhouse, Re7) handle the allocation decisions.
**The liquidation mechanic.** When the value of borrowed assets exceeds the liquidation threshold (collateral × liquidation-LTV), the position becomes liquidatable. Anyone — typically automated bots — can repay a portion of the borrowed debt and seize a proportional amount of collateral, receiving the liquidation bonus as profit. The bonus exists to incentivize liquidations during stress events when gas costs and slippage make liquidation expensive. Liquidation is not a single event for a large position; it's typically partial, with the liquidator closing only enough of the position to bring the health factor back to safety.
**The health factor.** Aave displays a 'health factor' on every position: (collateral value × liquidation-LTV) / borrowed value. A health factor of 1.0 is at the liquidation threshold; above 1.0 is safe; below 1.0 is liquidatable. A health factor of 2.0 means the position can sustain a 50 percent drop in collateral or a 100 percent increase in borrowed value before liquidation. Sophisticated users monitor health factor across all positions and rebalance when it drops below a chosen threshold (e.g., 1.5).
**Oracle dependence is the load-bearing risk.** Every lending market relies on an oracle to value collateral and debt continuously. A manipulated oracle — even briefly — can either liquidate solvent positions (oracle reports collateral lower than it is) or let underwater positions sit unliquidated (oracle reports debt lower than it is, letting the borrower accrue bad debt). The Mango Markets exploit in October 2022 ($117M) was exactly this: the attacker pumped the MNGO oracle by manipulating its underlying spot market, then borrowed against the manipulated collateral and walked away with the protocol's treasury. Lesson 5 covers oracle design in depth.
**Interest-rate manipulation risks.** Aside from oracle attacks, the more subtle risk in lending markets is interest-rate manipulation around the utilization kink. If an attacker can briefly push utilization above the optimal point, borrowing costs spike for everyone — including positions that were just under the liquidation threshold. This has been used in coordinated attacks: drain the pool's borrowable side via a large legitimate borrow, push utilization to 99 percent, sit there while other borrowers' positions become uneconomic and either repay (releasing liquidity at higher rates) or get liquidated.
Example
Walk through a concrete Aave position. User deposits 10 ETH worth $35,000 (price $3,500). They borrow $20,000 of USDC. ETH's LTV in eMode is 90 percent (because USDC is a stablecoin and Aave allows higher LTV for ETH-vs-stable in eMode); standard mode might be 82 percent. Liquidation threshold: 93 percent. Health factor at deposit: ($35,000 × 0.93) / $20,000 = 1.63. Now ETH drops to $2,400: collateral value falls to $24,000; health factor = ($24,000 × 0.93) / $20,000 = 1.12. Still safe but tight. ETH continues to $2,150: health factor = ($21,500 × 0.93) / $20,000 = 1.00 — at the liquidation threshold. A liquidator now repays, say, $10,000 of the USDC debt and seizes $10,500 of ETH collateral (the $500 difference is the 5 percent liquidation bonus). The position is now $11,000 collateral / $10,000 debt; health factor recovers. The user has effectively been forced to sell ETH at $2,150 minus the bonus — significantly below the spot price.
Common mistakes
- Borrowing close to the liquidation threshold. A 10 percent drop in collateral can trigger liquidation; staying below 50 percent LTV gives you a much larger buffer.
- Ignoring the liquidation bonus. The bonus is the implicit cost of being liquidated — typically 5–15 percent of the seized collateral.
- Conflating LTV and liquidation threshold. The LTV is the max borrowable; the liquidation threshold is slightly higher (gives the position a small buffer before liquidation). Both matter.
- Forgetting interest-rate risk on the borrow side. Stablecoin borrow rates can rise from 3 percent to 20 percent within hours when utilization spikes.
- Treating Morpho Blue vaults as equivalent to Aave/Compound positions. The risk model is genuinely different — the curator's allocation decisions are the load-bearing trust assumption.
- Assuming oracle prices are always accurate. They are usually accurate; the documented exploits are precisely when they weren't.
Check your understanding
You have an Aave position with $50,000 of ETH collateral and $20,000 of USDC borrowed. The asset's liquidation threshold is 85 percent. At what ETH-collateral USD value does your position become liquidatable?
Key terms covered
Sources & further reading
- Primary
- Primary
- Primary
- SecondaryMango Markets exploit post-mortem (October 2022)
Documentation of the oracle-manipulation lending exploit.
We prioritise primary sources. Where a topic moves quickly (regulation, security incidents), we re-check sources on the cadence shown by the page's "Next review" date.