A borrower can leave a crypto loan unchanged and still face a sharp increase in its interest cost. In a utilisation-based lending pool, the price of borrowing depends on the state of the pool. Other borrowers drawing funds or suppliers withdrawing liquidity can change that state without any action by the original borrower.
That makes the displayed annual rate a snapshot. The institution needs to understand the curve that produces the rate, the events that move it and the time needed to repay or refinance. Our article on DeFi lending and vault economics explains the funding relationship. The borrowing desk needs a more specific model of repricing.
Utilisation measures how much liquidity is employed
In a simplified pool, utilisation is outstanding borrowed funds divided by the sum of outstanding borrowing and available liquidity. If borrowing is $80 million and available liquidity is $20 million, utilisation is 80%. The exact implementation can use additional accounting inputs, so production analysis should use the market's deployed calculation rather than substitute this simplified ratio.
A withdrawal of $10 million of available liquidity, with borrowing unchanged, raises the simplified ratio to about 88.9%. No new loan caused the increase. This is why supplier concentration matters to a borrower: a large lender leaving can change the cost of financing for everyone who remains.
Available liquidity also matters for suppliers trying to exit. A high displayed supply rate may coincide with little immediately withdrawable cash. Borrowers and suppliers therefore share a market-state dependency even though their economic objectives differ.
The kink changes the sensitivity
A piecewise rate curve applies one slope below a specified utilisation level and another above it. The change in slope is commonly called a kink. Its purpose is to make additional borrowing progressively more expensive when liquidity is scarce and potentially attract more supply or encourage repayment.
The Aave V3 interest-rate strategy documentation describes two slopes around an optimal usage ratio. It exposes parameters for the base rate, slopes and maximum rate. Those are market and reserve parameters, not universal constants for all assets carrying the Aave name.
A desk should obtain the applicable curve for the actual deployment, asset and observation date. It should also establish who can change the parameters and how changes take effect. A screenshot of a current rate cannot answer either question.
A hypothetical curve makes the risk visible
Assume a hypothetical pool uses a 2% base annual rate, an 80% kink, a 4 percentage-point increase across the region below the kink and a 60 percentage-point increase across the remaining region. This is an illustrative curve, not a quotation of live protocol parameters.
At 60% utilisation, the annual borrowing rate is 2% plus 4% multiplied by 60/80, or 5%. At the 80% kink it is 6%. At 90%, it is 6% plus 60% multiplied by 10/20, or 36%. A move of ten utilisation points above the kink produces a much larger repricing than ten points below it.
For a hypothetical $5 million loan, a simple seven-day interest estimate at a constant 5% annual rate is about $4,794.52 using a 365-day denominator. At a constant 36%, it is about $34,520.55. Those are scenario calculations, not forecasts, and actual protocol debt accrual may use compounding and implementation-specific timing.
The example shows why average utilisation is an incomplete budget input. Several days spent above the kink can dominate financing cost even when the monthly average appears moderate. The path and duration of utilisation matter, as does the starting balance being compounded.
Rate increases and collateral pressure can coincide
A market shock can encourage suppliers to withdraw, borrowers to seek liquidity and collateral prices to fall. A borrower may then face both higher interest expense and less headroom against liquidation. Treating rate risk and collateral risk as independent scenarios can miss the combination that matters operationally.
Accruing interest increases the outstanding debt measured under the protocol. Even with a stable collateral price, that can erode borrowing headroom over time. The significance depends on the debt balance, rate path and collateral rules. There is no universal rate level at which a loan becomes unsafe.
Our article on digital-asset collateral and haircuts addresses valuation protection. The financing model should connect that protection to debt accrual rather than assume a collateral buffer remains constant while borrowing costs rise.
Supply yield is a different number
A pool's supply rate is not usually the borrow rate paid to every supplier. In a simplified relationship, aggregate interest paid by borrowers is spread across supplied capital, with an agreed portion allocated to the protocol or another recipient. Utilisation and deductions therefore affect the relationship between the two rates.
A borrower should not estimate its cost using the supplier dashboard. A supplier should not interpret a high borrowing rate as the yield on all deposited capital. Incentive rewards, where present, add another component that should be separated from interest generated by borrowing.
This distinction becomes important in internal transfers. If one institutional team supplies liquidity while another borrows, the two positions do not automatically offset at the same rate. Fees, pool allocation, utilisation and the timing of withdrawals can leave a material spread and different availability constraints.
Repayment has an operating lead time
An intention to refinance does not cap the existing loan's rate. The borrower needs the correct repayment asset, available transfer routes, approvals and a transaction that completes. During that process, interest continues under the applicable rules. The alternative funding source may also reprice under the same market conditions.
A practical scenario model includes the time between detecting a rate threshold and reducing the debt. It should account for the asset's availability in custody, any bridge or exchange route and the institution's signing process. The loan balance at the start of the response may differ from the balance needed for full repayment when the transaction executes.
For partial repayment, model how the action affects both the borrower's debt and the pool's utilisation. A large borrower may materially lower the rate by repaying; a small borrower may have little influence. That feedback should be calculated rather than assumed to be either negligible or decisive.
Budget the curve, then monitor its inputs
A borrowing proposal should explain expected cost under a base case and the cost of explicit utilisation paths, including time above the kink. Include supplier withdrawals, borrowing demand and parameter changes as separate drivers. The purpose is to reveal the source of variability, not to attach false confidence to one annualised number.
Monitoring then needs more than a rate alert. Track available liquidity, distance to the kink, concentration of supply and the current debt balance. A threshold can trigger preparation before a high rate arrives. The institution is better placed when it knows what action is possible within its operating timetable, rather than merely recognising afterward that a cheap loan became expensive.