How Does Compound V3 Work?

Comet architecture: each market has a single borrowable base asset (e.g., USDC) with multiple collateral types, isolating borrowing risk per market

Simplified from V2's multi-asset borrowing model. Single base asset per market reduces complexity and cross-asset contagion but fragments liquidity across Comet instances.

Decoupled supply and borrow rates with kinked utilization curve, where borrow rates spike sharply above optimal utilization

Improved from V2 with decoupled rates that maintain liquidity for lenders even at high utilization. Standard kinked curve design with governance-set parameters.

Absorption-based liquidation that gradually unwinds positions as they approach insolvency, rather than sudden full liquidation

Gradual liquidation reduces cascade risk compared to V2's sudden liquidation. However, in fast-moving markets, gradual unwinding may not keep pace with price declines.

COMP token governance with Governor Bravo, timelock, and on-chain binding proposals

Battle-tested governance framework but proven vulnerable to whale coordination. The 2024 Golden Boys attack demonstrated that low participation enables treasury extraction.

Compound governance guardian with veto-only power to block malicious proposals

Guardian can veto proposals but cannot initiate them. Guardian role is set to expire, transitioning to full community governance with reduced safety nets.

Chainlink price feeds for all collateral assets with per-asset liquidation thresholds

Standard Chainlink dependency. Oracle staleness or downtime could prevent timely liquidations, especially for volatile collateral assets.

Protocol reserves accumulate from spread between supply and borrow rates, with governance-directed spending

DAO treasury has been a governance attack target. Proposal 513 (Dec 2025) transferred $1.1M to the Compound Foundation, showing active treasury utilization.

Low voter participation combined with concentrated COMP holdings enabled the 2024 Golden Boys governance attack, extracting $24M from the treasury. Despite settlement, the structural vulnerability remains for future whale-coordinated proposals.

All borrowing demand concentrates in one base asset per market. A sudden spike in borrowing demand can exhaust the base asset, preventing lenders from withdrawing even if their collateral is healthy, creating liquidity freezes.

Gradual absorption may not unwind positions fast enough during flash crashes. If collateral value drops faster than the absorption rate, bad debt accumulates and is socialized across lenders.

Oracle update delays during high-congestion periods (e.g., gas spikes on Ethereum) can cause liquidation threshold monitoring to lag, allowing undercollateralized positions to persist.

As the governance guardian role sunsets, the timelock becomes the primary safety mechanism. Short timelocks may not provide enough reaction time for the community to organize against malicious proposals.