Aave’s interest rate model is elegant on paper. Below a target utilization threshold - typically around 80–90% depending on the asset - borrow rates climb gradually. Cross that threshold and the slope turns near-vertical, theoretically shocking borrowers into repaying and pulling the pool back into a healthy range. The mechanism has worked often enough that it’s been copied across half of DeFi.

The problem is that the model assumes liquidity providers will behave symmetrically - that high rates will attract new deposits roughly as fast as rate spikes deter new borrows. In practice, that’s not what happens.

When a major asset hits 95% utilization on Aave, the borrow APY can jump from 15% to 80% or more within hours. That punishes existing borrowers immediately, especially those running leveraged yield positions that were calibrated at lower rates. Meanwhile, new supply-side liquidity takes longer to arrive: LPs need to bridge capital, assess whether the spike is temporary or structural, and decide whether the elevated yield justifies the lockup risk. The asymmetry is real and it shows up in withdrawal queues during stress events.

The stablecoin markets are where this gets most visible. USDC and USDT pools on Aave V3 have hit utilization ceilings multiple times during periods of elevated on-chain demand - particularly when Curve or other protocols experience yield compression and capital rotates into lending markets. At those moments, users who supplied capital to earn yield find themselves unable to withdraw, not because Aave is insolvent, but because there’s simply no available liquidity in the pool.

Aave has addressed this partly through its GHO stablecoin, which can theoretically absorb some demand that would otherwise strain third-party asset pools. But GHO’s peg has been inconsistent since launch, and its adoption as an actual borrowing alternative remains limited compared to USDC.

The deeper issue is that utilization-based rate curves are a reactive mechanism. They respond to imbalance; they don’t anticipate it. Morpho’s peer-to-peer matching layer and Euler’s reactive interest rate governors are both attempts to build something more adaptive - matching rates to actual capital availability rather than a smooth curve that breaks under extreme conditions.

Whether any of those approaches holds up better during a genuine liquidity crunch - not a theoretical one - is still an open question.