Compound: The Money Market Protocol
Executive summary
Compound's whitepaper (Robert Leshner and Geoffrey Hayes, February 2019) is a brief but historically pivotal protocol specification: eight pages that defined the algorithmic money-market template most subsequent DeFi lending protocols (Aave, Cream, Venus, Benqi, Sonne, Moonwell) inherit. The cToken accumulator pattern, the utilization-based interest-rate model, and the per-asset risk-parameter framework (collateral factor, reserve factor, close factor) are all originated or canonicalised here — though the paper's example rate curve is purely linear (2.5% + 20% × utilization); the now-standard 'kinked' jump-rate curve arrived later via Compound governance. The paper is also a model of operational under-specification — it specifies mechanism without rigorously analysing failure modes that became consequential in 2020-2022 stress events. DI rates the paper an eight on systems clarity and a five on risk analysis; its design has aged well, its risk treatment has not.
Background and motivation
Lending in early DeFi (2017-2018) was dominated by p2p matching protocols like ETHLend (later Aave), where lenders and borrowers were paired manually with bespoke terms, and the original MakerDAO single-collateral CDP system, where users borrowed DAI against ETH at protocol-set rates. Neither model scaled well: p2p matching was inefficient and required active counterparty discovery, and MakerDAO's single-asset, single-direction borrow flow was specialised to its stablecoin-issuance use case. Compound's contribution was to imagine permissionless money markets as a pooled-liquidity primitive, where lenders deposit into per-asset pools and borrowers draw from those pools subject to algorithmically-determined interest rates. The conceptual move was to replace counterparty discovery with a shared-pool aggregator, replace negotiated rates with algorithmic rates, and replace bespoke risk assessment with on-chain collateral mathematics. Compound v1 (2018) was a proof-of-concept; v2 (2019) was the production-ready version with the cToken design, the multi-asset markets, and the reference architecture that other protocols would copy. The intellectual lineage is partly TradFi (banks have offered pooled deposit-and-borrow products forever; the model formalised here is closest to repo markets with continuous rate adjustment) and partly the AMM literature (Compound's pooled-liquidity-with-algorithmic-pricing has structural similarities to constant-product AMMs, though the paper does not draw the parallel). The 2018 launch context — declining DeFi prices and falling activity — limited initial adoption, but Compound v2's mid-2019 launch coincided with renewed DeFi activity and the COMP governance-token launch in June 2020 catalysed the 'DeFi summer' that established lending protocols as a permanent DeFi category.
Core technical contributions
The cToken model is the paper's signature contribution. When a user deposits asset X, they receive cX tokens at the prevailing exchange rate. The exchange rate increases over time as interest accrues to the underlying market, so cX balance does not change but its claim on X grows. This design has three virtues. First, it is gas-efficient: interest accrual updates a single global state variable per market rather than per-position balances. Second, cTokens are themselves transferable ERC-20 tokens, enabling secondary-market trading of deposit positions without protocol involvement. Third, the model decouples interest accrual from user actions — lenders earn interest passively without needing to claim. The interest-rate model is a function of utilization (the fraction of pooled liquidity currently borrowed): the whitepaper specifies a governance-set demand curve and gives a purely linear example — Borrowing Interest Rate = 2.5% + utilization × 20% — with the supply rate implicit (the borrow rate multiplied by utilization). The now-universal piecewise-linear 'kinked' jump-rate curve — modest slopes below a target utilization (typically 80%) and steep slopes above it, preserving redemption liquidity for lenders during stress — is a later Compound-governance-era refinement, not part of the paper itself. Per-asset risk parameters — collateral factor (max LTV when used as collateral), reserve factor (protocol fee on interest), close factor (max share of debt liquidatable in one transaction, the deployed default being 50%) — give governance a small set of dimensions to tune market risk per asset. Liquidations are open to any liquidator: any party can repay a portion of an undercollateralised borrower's debt and seize their collateral at the market price minus a liquidation discount (deployed v2 set the liquidation incentive at 8%).
Methodology and rigor
The paper is engineering-spec rigor, not academic-paper rigor. The protocol's mechanisms are specified at sufficient detail for implementation, but the analytical justification for design choices is thin. The interest-rate curve is presented as a governance-set parameter without analysis of how to choose its shape — the paper's worked example is simply linear in utilization — and the kinked jump-rate model that later became standard, with target utilization converging empirically on 80-90% for most assets, arrived without theoretical grounding in the original text. The collateral-factor parameters are mentioned without a method for setting them; in practice, Compound's governance has set them by reference to historical volatility and liquidation depth, which is reasonable but ad-hoc. The paper offers no analysis of liquidation incentive sufficiency, no model of liquidation execution under congestion, and no discussion of oracle-attack scenarios despite oracle-driven liquidations being structurally central. The pooled-liquidity model means that depositors share counterparty risk across all borrowers in a market — a fact mentioned but not rigorously analysed for tail-risk implications. The Perez-Werner-Livshits empirical study (2020) and subsequent work showed that the protocol's behaviour in stress (particularly during the November 2020 DAI price spike to $1.30+ on Coinbase Pro that cascaded into liquidation-fueled bad debt on Compound) was not well-anticipated by the original design. The paper's silence on these dynamics is a real limitation; the protocol's strengths in normal market conditions came at the cost of fragility in tail scenarios that weren't analysed in advance.
Strengths
The cToken pattern is the paper's most durable contribution. By the time Aave (2020), Cream, Venus, Benqi, and others launched, all had adopted essentially the same exchange-rate-growth model; subsequent rebasing-token designs (cToken's main competitor) have generally been considered worse for composability. The utilization-driven rate-curve design has been similarly universal — virtually every algorithmic money market in DeFi prices borrowing as a function of utilization, as this paper specified, though the piecewise-linear kink-at-80% default that descendants standardised on was a post-whitepaper refinement. The pooled-liquidity model proved structurally efficient: total addressable depositors and borrowers face a single pool per asset, eliminating the matching friction of p2p designs and providing always-available liquidity. The per-asset risk-parameter abstraction (collateral factor, reserve factor) gave governance a minimal-but-sufficient toolkit for parameter tuning, and that abstraction propagated to subsequent protocols essentially unchanged. The paper's deliberate scoping was a strength: by not attempting to solve every lending use case (variable+stable rates, isolated markets, undercollateralised loans, etc), the v2 design remained tractable and shippable. Open-source contract architecture — Compound's reference codebase was widely forked — accelerated the broader DeFi lending ecosystem. The protocol's longevity (launched 2019, still operational and economically meaningful in 2026) is itself empirical evidence that the core design choices were sound.
Weaknesses and limitations
The most consequential weakness is that v2 has a single shared-risk pool per asset with no isolation. A risky asset listed on Compound shares lender capital with every other asset on the protocol; a bad-debt event in one market can drain reserves and impair lenders in others. Compound has also suffered separate smart-contract and governance-upgrade bugs — the September 2021 Comptroller reward-distribution bug that mistakenly over-distributed roughly $80M in COMP, and the August 2022 cETH price-feed upgrade bug that temporarily froze the cETH market — though these stemmed from contract errors rather than from the shared-pool design itself. Subsequent protocol designs (Aave v3 efficiency mode and isolated mode, Compound III itself, Morpho Blue's isolated markets, Euler v2 vaults) all moved toward isolated risk pools, which v2 did not anticipate. A second weakness is the close-factor design: a single transaction can liquidate up to 50% of a borrower's debt, which during stress can cascade as multiple liquidators front-run each other. Third, the oracle architecture (originally Compound's own price feed, later migrated to Chainlink) was not rigorously specified in the paper, leaving oracle attack as an under-considered surface. Fourth, the paper's reserve-factor mechanism provides limited dynamic protection — reserves accumulate only from interest, not from any direct skin-in-the-game by the protocol — which means protocol equity is at lender expense rather than additive. Fifth, the paper offers no analysis of governance-attack vectors, despite Compound being one of the first protocols to launch a governance token (COMP). The protocol has subsequently survived several governance-attack attempts but has had to add timelock and quorum protections that weren't in the original design.
Subsequent influence
Compound v2 is the most copied protocol architecture in DeFi after Uniswap v2. Direct forks include Cream Finance, Venus (BSC), Benqi (Avalanche), Sonne (Optimism), Moonwell (Base/Moonbeam), Tectonic (Cronos), and dozens of smaller chain-native lending protocols. Aave v1 launched after Compound v2 and adopted very similar mechanics with a stable-rate option layered on top; Aave's v2 (December 2020) and v3 (March 2022) further refined the design with isolated assets, efficiency mode, and supply caps but retained the kinked-rate-curve and aToken (Aave's analogue to cToken) design. The influence is structural: the 'algorithmic money market' as a category is essentially a Compound v2 descendant. Beyond the protocols themselves, the paper influenced the broader DeFi composability story. cTokens were one of the first widely-used 'yield-bearing tokens', and their composability into other protocols (lending cTokens as collateral on Maker, wrapping cTokens for leverage on dHedge, etc.) demonstrated that protocol-level interest accrual could be a primitive rather than a closed product. The COMP token launch (June 2020) was a separate but related event that initiated the 'governance-token-driven liquidity mining' era — Compound was the first major DeFi protocol to distribute a governance token to users. Subsequent academic literature (Perez et al 2020, Klages-Mundt formal stablecoin work, Werner et al's DeFi SoK) treats Compound as a canonical example of algorithmic money markets and uses it as a baseline for empirical analysis.
How it holds up in 2026
Seven years after the v2 paper, Compound itself is no longer the dominant lending protocol — Aave overtook it on TVL in 2021 and Morpho's optimisation layer captured a substantial share of efficient lending flow from 2022 onward. Compound III (launched 2022) significantly redesigned the architecture toward isolated markets and a single borrow asset per deployment, implicitly conceding that the v2 shared-risk-pool design had limitations. But the core design patterns from v2 — cTokens, utilization-based rate curves, per-asset risk parameters — remain the industry standard. By 2026, virtually every algorithmic money market in DeFi descends architecturally from this paper, even when the descendants have substantially diverged. The most important post-2019 development is the move toward isolated risk pools (Aave v3 isolated mode, Morpho Blue, Euler v2, Compound III), which addresses the v2 shared-pool weakness. A second important development is the rise of lending-rate optimisation layers (Morpho being the canonical example), which capture the spread between supplier and borrower rates that the pooled utilization-curve model leaves on the table — this is essentially a critique of the rate-discovery mechanism in v2 that the paper did not anticipate. Risk-parameter management has become a separate professional category (Gauntlet, Chaos Labs, BlockAnalitica), reflecting that the paper's offhand treatment of parameter setting was inadequate. DI's holistic view is that v2 was a foundational design that solved the core architecture of permissionless lending in DeFi and left the risk-management and capital-efficiency optimisation problems for the field to solve incrementally over the subsequent half-decade. As a piece of writing it is clear and shippable; as analysis it underspecifies the dynamic-risk dimensions that became consequential. Its lasting contribution is the abstractions — cToken, kinked curve, per-asset parameters — that are now infrastructure across DeFi.
Strengths
- cToken model: interest-via-exchange-rate is gas-efficient and composable
- Utilization-driven rate curve is now industry standard
- Per-asset risk parameters: minimal but sufficient governance toolkit
- Pooled-liquidity model eliminated p2p matching friction
- Open-source reference implementation accelerated DeFi lending ecosystem
- Block-by-block interest accrual without per-position storage cost
Weaknesses
- Single shared-risk pool per market — bad debt in one asset can impair others
- Close-factor design enables liquidation front-running cascades during stress
- Oracle architecture unspecified, leaving oracle-attack surface unanalysed
- Reserve-factor mechanism provides minimal dynamic protocol skin-in-the-game
- Governance-attack vectors unconsidered despite COMP-token-driven model
- Parameter-setting methodology not specified — left to ad-hoc governance
Key contributions
- cToken model: interest accrual via exchange-rate growth rather than balance changes
- Utilization-based interest-rate curve set by governance (linear in the paper; the kinked variant came later)
- Per-asset risk parameters: collateral factor, reserve factor, close factor
- On-chain liquidations open to any liquidator, at a fixed discount with a partial-liquidation clamp (close factor)
- Pooled liquidity model — depositors share counterparty risk across all borrowers in a market
- Block-by-block interest accrual without per-position storage cost
- Open-source contract architecture that became the reference implementation
TL;DR
Defined what an algorithmic money market looks like in DeFi — cTokens and utilization-based rate curves are now infrastructure, even as Compound itself lost share to Aave and Morpho.
Get DeFi Intel research in your inbox
Weekly long-form coverage of papers, incidents, jurisdictions, chains, tokens and the people building them. Free tier covers headlines; Pro adds the analyst-grade breakdowns.
Sources
External references gathered from the body of this brief. Last reviewed 2026-07-15.