How Does Marginfi Work?
A Solana lending protocol where you deposit crypto to earn interest or borrow against it, with $300M in deposits across multiple market types. It survived a close call in September 2025 when a bug that could have drained $160M was caught before exploitation. Its C+ grade reflects that near-miss, a 57% TVL crash during the Q1 2025 selloff, and acquisition uncertainty.
TVL
$40M
Sector
Lending
Risk Grade
C+
Value Grade
C-
Core Mechanisms
Lending/Over-Collateralized
Marginfi: overcollateralized lending with multi-market architecture (global, isolated, native stake)
Standard overcollateralized lending model similar to Aave/Compound but with a multi-market structure that separates risk across global interconnected markets, isolated per-asset markets, and native stake markets.
Lending/Isolated-Markets
Isolated lending pools for higher-risk long-tail Solana tokens with independent risk parameters
Each isolated market has its own collateral factors, interest rates, and liquidation thresholds, preventing toxic asset contagion to the core lending pool.
Lending/Interest-Rate-Curve
Kinked utilization curve with dynamic rate adjustment across multiple market types
Standard kinked interest rate model where rates spike above a utilization threshold to incentivize repayment and ensure withdrawal liquidity.
Lending/Liquidation
Partial liquidation with insurance fund backstop for bad debt
Positions are partially liquidated to minimize market impact. An insurance fund absorbs bad debt when liquidations fail to fully cover shortfalls.
Lending/Flash-Loan
Flash loan facility with account transfer capability (patched vulnerability in Sept 2025)
Flash loans allow atomic borrow-and-repay within a single transaction. The transfer_to_new_account instruction introduced a vulnerability that could bypass repayment checks, patched after Asymmetric Research disclosure.
Oracle/Multi-Oracle
Redundant oracle system with Pyth and Switchboard price feeds
Uses multiple oracle sources for price data to reduce single-oracle dependency. Oracle staleness checks prevent stale prices from triggering incorrect liquidations.
Incentive/Points-System
NovelPoints program with 1:1 migration promised to MRGN token under Project 0 acquisition
User activity earns points that will convert to MRGN tokens at TGE. The 1:1 migration promise under new ownership creates uncertainty about actual conversion terms and timeline.
How the Pieces Interact
Isolated markets for low-liquidity Solana tokens may experience failed liquidations during market stress, as liquidators cannot profitably close positions without sufficient DEX depth, leading to socialized bad debt.
The interaction between flash loans and account transfers created a vulnerability (Sept 2025) where liabilities could be shifted to disposable accounts, bypassing repayment. Future similar logic errors remain a risk in complex instruction sets.
During rapid price movements, oracle update latency on Solana can cause liquidation thresholds to be breached before bots can act, especially for volatile long-tail assets in isolated markets.
The insurance fund may be insufficient to cover simultaneous bad debt events across multiple isolated markets during a correlated Solana ecosystem downturn.
Acquisition changes may alter point conversion terms or timelines, causing point farmers to withdraw deposits preemptively and creating a TVL exodus before TGE.
What Could Go Wrong
- Flash loan vulnerability in September 2025 risked $160M in user deposits; patched before exploit but highlights smart contract risk
- TVL fell 57.6% in Q1 2025 amid market selloff, processing $1.7B in liquidations and revealing fragility under stress
- Project 0 acquisition creates governance uncertainty and potential delays to MRGN token launch for point holders
Cascading Liquidation Spiral in Correlated Downturn
ModerateTrigger: A sharp correlated decline in Solana ecosystem tokens (SOL, mSOL, JitoSOL) triggers simultaneous liquidations across Marginfi's interconnected and isolated markets
- 1.SOL drops 30%+ in 24 hours during a market-wide crash — Liquidation bots activate across all Marginfi markets simultaneously
- 2.Isolated markets for long-tail assets experience failed liquidations due to thin liquidity — Bad debt accumulates in isolated pools where liquidators cannot profitably close positions
- 3.Oracle price feeds lag behind rapid price movements on Solana DEXs — Liquidation thresholds are breached before positions can be closed, creating underwater loans
- 4.Lenders rush to withdraw from all markets as bad debt socialisation fears spread — Utilization spikes to 100%, trapping remaining lenders and preventing withdrawals
Risk Profile at a Glance
Overall: C+ (38/100)
Lower score = safer