Crypto lending utilization vs borrow APR

Crypto lending utilization measures how much of a market's lending capacity is borrowed under that protocol's accounting rules. A variable borrow APR is the annualized rate produced by its interest rate model at that utilization. A kinked model changes the rate's sensitivity at a specified utilization threshold. Today's APR is a snapshot; your accrued borrowing cost depends on the debt balance, subsequent rates and the protocol's accrual method.

To understand a rate change, reconstruct the pool state and its curve before extending a percentage across your planned holding period. The same utilization can produce different rates in different markets, and the supplier rate need not follow the same formula everywhere.

Define the denominator before reading the curve

Consider a hypothetical, simplified pool holding 2 million units of an asset in available cash, with 8 million units outstanding as loans. Ignore reserves, unbacked claims, interest already accruing and other accounting adjustments. Its lending capacity is 10 million units, and utilization is 80%: borrowed units divided by available cash plus borrowed units.

Dividing 8 million by the 2 million still sitting in the pool produces 400%. That measures debt relative to idle cash, a different ratio. Likewise, adding the value of unrelated collateral assets to the denominator would create a different statistic. A market's total value locked label is insufficient evidence of its rate-model denominator.

The inspected Aave V3 Origin interest rate strategy calculates borrow usage as total debt divided by available liquidity plus debt. Its supply usage denominator also includes unbacked amounts. Those two usage ratios can therefore differ. Available liquidity in that implementation incorporates the virtual underlying balance and liquidity added or taken during the operation.

An inspected Compound III implementation provides another boundary. The Comet contract source divides present-value total base-asset borrowing by present-value total base-asset supply; its getter returns zero when supply is zero. It has separate functions and parameters for supply and borrow curves. Reconstruct the named market's fields rather than substituting a generic cash formula.

Keep utilization separate from your account's collateral ratio. Pool utilization concerns lending liquidity; LTV and liquidation thresholds concern how much an account can borrow and when its collateral position becomes eligible for liquidation.

Rebuild a hypothetical kinked rate curve

A kink is the point where the rate's sensitivity to utilization changes. It is not a promised operating level or a guarantee that liquidity remains available. Aave's V3 interest rate strategy documentation describes one slope below the optimal usage ratio and another between that point and full usage. Parameters belong to the particular reserve and market.

The following invented model illustrates the arithmetic. These are not current parameters for Aave, Compound or any available product:

• Utilization at the kink: 80%.
• Base borrow APR at zero utilization: 2%.
• Additional APR from zero utilization to the kink: 8 percentage points.
• Additional APR from the kink to 100% utilization: 60 percentage points.
• Reserve factor: 10% of borrower interest, under the supplier-allocation assumptions below.

Write utilization as a decimal U. At or below 80%, borrow APR is 2% plus 8% multiplied by U divided by 0.80. Above 80%, it is 10% plus 60% multiplied by the fraction of the upper segment used: (U − 0.80) / 0.20.

At 90% utilization, half of the upper segment is used. The rate is therefore 10% + 60% × 0.5 = 40%. Multiplying the 60% parameter directly by 90% would use the wrong slope convention. Other implementations can express slopes per unit of utilization instead of as the total increase over a segment; parameter names alone do not settle that distinction.

The supplier column uses the simplified interest allocation explained in the next section.

Below the kink, one additional percentage point of utilization adds 0.1 percentage point to borrow APR. Above it, the same utilization change adds 3 percentage points. The curve is continuous at 80%: either segment gives 10%. Its slope changes sharply, while the rate itself has no jump at that exact boundary.

The 100% row is a mathematical endpoint. It does not establish that another borrow or withdrawal could execute there.

Reconcile the supplier rate with borrower interest

For the table, assume all loans pay the same variable rate, total supplied units equal cash plus debt, and there are no unbacked amounts, defaults, incentives or extra income. Under this constructed allocation, supplier APR equals borrow APR multiplied by utilization and by one minus the reserve factor.

At 80%, that is 10% × 0.80 × 0.90 = 7.20%.

The denominator explains the difference. With 10 million supplied units and 8 million borrowed units, a frozen 10% borrow APR corresponds to 800,000 units of annualized borrower interest. The 10% reserve factor allocates 80,000 of those units to reserves, leaving 720,000 for suppliers. Dividing 720,000 by all 10 million supplied units produces 7.2%.

This is a frozen-rate annualization, not a prediction of a year's receipts. The 2.8 percentage point gap between borrow and supplier APR is not itself the reserve factor. It also reflects the 20% of capacity that is idle. Nor does a 10% reserve factor mean a yearly deduction of 10% of supplied principal.

The inspected Aave V3 strategy multiplies the variable borrow rate by its supply usage ratio and the complement of the reserve factor. Our simplified denominator makes supply usage equal borrow usage. Do not apply this shortcut universally: Compound Comet uses distinct supply and borrow curves, as its specification's interest-rate overview explains and its current source confirms.

APR also needs an accrual convention before it becomes an effective annual growth rate. The APR versus APY guide explains the compounding distinction; a lending position still needs its own protocol-specific calculation and balance records.

Your transaction can change the quoted rate

Return to the hypothetical pool with 2 million cash and 8 million debt. Before any action it has 80% utilization and a 10% borrow APR under our curve.

Suppose a permitted new borrow transfers 1 million units out of the pool. Cash falls to 1 million and debt rises to 9 million. Total capacity remains 10 million, so utilization becomes 90% and the model produces 40% APR. A calculation using the earlier 10% quote misses the effect of the borrow itself. In this invented variable-rate model, the new rate applies to outstanding variable debt going forward, not only to the final million borrowed.

An alternative event reaches a similar result without new borrowing. Starting again from the original state, suppose a supplier withdraws 1 million units. Cash falls to 1 million while debt remains 8 million. Capacity is now 9 million, utilization becomes 88.8889%, and borrow APR becomes approximately 36.6667%.

These are separate state transitions, not successive events. They show why rising utilization does not prove rising loan demand: withdrawals can shrink its denominator. Rate increases may encourage repayment or additional supply, but the curve cannot ensure either response arrives before someone needs liquidity.

Replace a three-day quote with a three-day ledger

Here is a separate hypothetical borrowing of 10,000 units. Assume the principal stays constant, interest is recorded separately without capitalization, each day uses one constant APR, and the year has 365 days. Exclude fees, rounding and every other cash flow. These simplifying rules isolate variable-rate exposure; they do not reproduce a named protocol's accrual implementation.

Each day's charge is principal multiplied by decimal APR divided by 365. Ending debt, including the separately recorded interest, is 10,020.547945 units. Extending the initial 10% across all three days would estimate only 8.219178 units of interest.

If instead this invented loan capitalized interest after each day, multiplying the three daily growth factors would give 10,020.560332 units. The difference is small over this example's short horizon; the assumptions still matter. For an actual position, reconcile the protocol's borrow index, timestamps, rounding and repayments rather than selecting whichever approximation is convenient.

The Comet source's index update uses per-second rates and elapsed time to update separate supply and borrow indices. Its rate getters explicitly do not accrue interest first. A raw getter value, a displayed annualized percentage and a current accrued account balance are different pieces of evidence.

An exchange loan can have a different clock again. The spot-margin borrowing-interest ledger covers charging periods and repayment evidence; those exchange policies cannot be imported into an on-chain market.

Keep a reproducible rate worksheet

Before comparing two lending screens, save enough information to reconstruct each one:

Recalculate the displayed utilization and rate at the same block. Then calculate the state after the intended borrow or withdrawal. Test a higher-utilization case and a changed-parameter case separately; a correct current quote does not validate an unchanged future curve.

Three failures deserve an explicit check. First, a dashboard may display an incentive-adjusted net rate while the debt still accrues at the gross borrow rate; keep rewards in a separate asset ledger. Second, mixing snapshots from different blocks can create a rate mismatch that never existed in one pool state. Third, a rising supplier balance does not establish immediately redeemable cash. The receipt-token accounting guide separates a claim's recorded growth from redemption liquidity.

For the next comparison, capture one market at one block, reproduce its denominator and both sides of its rate curve, and reconcile one actual accrual interval. That record turns a displayed APR into a calculation you can check.