The Only Number That Actually Matters

You deposit a stablecoin into a lending protocol and the borrow rate appears before you've finished blinking. No oracle. No price feed. No committee vote. The protocol calculated it entirely from its own internal state, in a single arithmetic step, before you even thought to ask.

That number is utilization.

Utilization is the fraction of a pool's supplied assets currently out on loan. If a pool holds 10 million USDC in total deposits and 6 million of those are borrowed, utilization is 60%. Everything else in the interest rate model flows from that one figure, with no external data required, no trust assumptions about third-party feeds, and no attack surface for price manipulation.

The design is more elegant than it first looks. Worth understanding properly.

Utilization as a Self-Contained Signal

It works because it captures exactly what the protocol needs to know: is money scarce right now, or is there plenty to go around?

When utilization sits at 20%, lenders earn thin returns and borrowers face little competition for funds. No reason to pay up for deposits. When utilization climbs toward 90%, the pool is nearly drained, borrowers are competing for the last dollars, and lenders deserve better compensation for the liquidity risk they're absorbing. The protocol doesn't need a Bitcoin price or an ETH/USD feed to know any of this. It only needs to look inward.

The formula is blunt:

``` Utilization (U) = Total Borrowed / Total Supplied ```

Compound Finance, one of the first protocols to formalize this approach, published its interest rate model in its whitepaper and on-chain code from its earliest deployments. Aave followed with a similar structure but added a specific refinement that became the standard: the kinked curve.

The Kink, and Why It's There

A simple linear model raises rates steadily as utilization climbs from 0% to 100%. That works, but it leaves a problem: at very high utilization, lenders can't withdraw because there's nothing left to withdraw. The pool becomes illiquid. A linear curve doesn't punish that outcome sharply enough, and in a system with no phone number to call, "not sharp enough" is a design failure.

The kinked (or two-slope) model solves this with a breakpoint, typically called the optimal utilization rate. Aave's rate strategy contracts define an `OPTIMAL_UTILIZATION_RATE` parameter, set at 80% for many stable assets in its V2 and V3 deployments. Below that kink, rates rise gently. Above it, they rise steeply.

In concrete terms, a model might look like this:

  • Below 80% utilization: Borrow APR = Base Rate + (U / Optimal U) × Slope1
  • Above 80% utilization: Borrow APR = Base Rate + Slope1 + ((U - Optimal U) / (1 - Optimal U)) × Slope2

Slope2 is far steeper than Slope1. In Aave's stable-asset pools, Slope2 values have historically been set dramatically higher than Slope1, sometimes by a factor of ten or more, specifically to create a near-vertical cost at extreme utilization.

Run the numbers. A USDC pool: Slope1 set to 4%, Slope2 set to 75%, optimal utilization at 80%, base rate at 0%.

At 60% utilization: Borrow APR = 0% + (0.60 / 0.80) × 4% = 3%

At 90% utilization: Borrow APR = 0% + 4% + ((0.90 - 0.80) / 0.20) × 75% = 4% + 37.5% = 41.5%

That jump, from 3% to 41.5% over thirty percentage points of utilization, is not an accident. It's the protocol screaming at new borrowers: there's almost nothing left in this pool, go somewhere else or pay dearly. Simultaneously screaming at lenders: your deposits are suddenly very valuable, add more.

Think of it as a pressure valve on a water main. Below the kink, normal operating pressure. Above it, the system is telling you the pipe is nearly dry, and the cost of drawing more rises fast enough to hurt.

What Lenders Actually Receive (And Why It's Always Less)

Borrowers pay the borrow rate. Lenders receive the supply rate. These are not the same number, and the gap is not a bug.

The supply rate is derived from the borrow rate but weighted by utilization:

``` Supply APR = Borrow APR × Utilization × (1 - Reserve Factor) ```

The reserve factor is a protocol-controlled percentage of interest siphoned into a treasury or insurance fund before lenders see a penny. Compound and Aave both use reserve factors, typically ranging from 5% to 20% depending on the asset.

Back to the scenario, at 60% utilization with a 10% reserve factor:

Supply APR = 3% × 0.60 × (1 - 0.10) = 3% × 0.60 × 0.90 = 1.62%

Lenders earn 1.62% while borrowers pay 3%. The spread compensates for the fact that not all deposits are working at any given moment and funds the protocol's risk buffer. No external oracle needed for any of this arithmetic. It all resolves from the single utilization figure.

The Part Most People Get Wrong About Oracle Independence

A misconception worth naming directly: "oracle-free interest rates" does not mean the protocol is oracle-free.

The borrow rate calculation is entirely internal. But almost every lending protocol still uses price oracles to determine collateral values and decide when a position should be liquidated. Aave uses Chainlink feeds for collateral pricing. Compound has its own price oracle infrastructure. Those are separate systems from the rate model.

The interest rate curve sidesteps oracle risk. Collateral valuation does not. Conflating the two produces a false sense of security about where the actual attack surface lives, and that's a mistake serious participants cannot afford to make.

A related confusion: people assume that because rates are algorithmic, they're immutable. They're not. Governance token holders in both Compound and Aave can vote to adjust Slope1, Slope2, the optimal utilization threshold, and the reserve factor. The curve's shape is governed; only the execution at any given utilization is automatic. When Aave's governance voted to adjust rate parameters for USDC markets following significant market volatility, it changed the slopes, but the utilization-based mechanism kept running without interruption. Governance sets the dial. The math runs itself.

Two Depositors, One Pool, Different Outcomes

Consider Maya and Daniel. Both deposit $50,000 in USDC to the same Aave pool on the same day. Utilization sits at 55%, and the supply rate is a respectable 2.8% APR.

Over the following months, a large protocol begins borrowing heavily from that pool to fund yield strategies elsewhere. Utilization climbs to 88%, past the kink. The borrow rate spikes. The supply rate follows, briefly touching 9% APR.

Maya notices the spike on her dashboard and withdraws half her position to deploy elsewhere. Daniel is on vacation. By the time he returns and withdraws, utilization has normalized back to 60% as new depositors rushed in chasing that 9% yield.

Both made money. But Maya captured the spike more fully, and Daniel benefited from the mean-reversion that the rate curve itself triggered. High rates attracted new supply, which pushed utilization back down, which reduced rates to equilibrium. No human intervention, no price feed, no central bank. The curve did all of it.

So ask yourself: if a system can reprice credit in real time, in every block, purely from one internal ratio, what exactly is the credit committee doing?

The Math Runs Itself

Every parameter in a borrow rate curve represents a deliberate policy choice: how aggressively to defend liquidity, how much to reward lenders versus borrowers, how large a reserve to build. Changing those parameters requires governance. Executing them requires nothing but arithmetic.

That's the actual innovation. Not the specific numbers, which vary by protocol, asset, and governance vote. The innovation is that credit pricing, one of the most information-intensive functions in traditional finance, can be reduced to a single internal ratio and a two-slope formula that never needs to phone home.

Want to audit it? The contract is on-chain. Want to predict it? Watch utilization. Want to game it? You'd need to move the entire pool, at which point you're providing the liquidity yourself.

For a system with no headquarters and no Bloomberg terminal, that's a hard arrangement to replicate. Most institutions spend billions trying.