You borrow 100,000 USDC on a quiet Monday. The interface quotes 6% variable — boring, serviceable, fine. Four days later the dashboard says 14%. Your first instinct is to hunt for a hack, a governance vote, an exploit. There is none. The pool simply got busier, crossed a threshold written into its rate contract, and your cheap loan rolled onto a different line of a piecewise curve.
Every major DeFi lending market — Aave, Compound, Morpho Blue and their peers — prices money this way. There is no loan officer and no rate committee. There is one ratio, two slopes, and an unforgiving bend. Once you can read them, rate spikes stop looking random and start looking predictable — often days before they hit your position.
TL;DR. Rates are set by utilization U = borrowed ÷ supplied, per asset pool. Below the kink (~85% on Aave, ~80% on Compound) borrow rates rise gently; above it the slope jumps to roughly 5–10x steeper, which is why a pool moving from 78% to 88% can double your rate. Suppliers earn borrow APR × U × (1 − reserve factor) — 8% borrow at 80% U with a 10% reserve factor pays 5.76%. At 100% utilization, withdrawals are blocked until repayments or deposits arrive, and stress rates can run toward 50–100%+. Before borrowing, check current U, distance to the kink and available liquidity — then model the cost.
Utilization is the entire market, compressed into one number
The utilization rate is the fraction of a pool's supplied capital currently on loan:
U = total borrowed ÷ total supplied, calculated separately for every asset pool.
A USDC pool with $100 million supplied and $82 million borrowed runs at 82% utilization. Next door, the ETH pool might sit at 45%. Rates never move as a platform-wide block; they move pool by pool, asset by asset, block by block. When you take a variable loan, you are not locking a price. You are taking a position on how crowded that specific pool will be tomorrow.
High utilization cuts two ways. It means suppliers earn more, because more of their capital is actually working. It also means the idle cushion — the cash sitting in the contract waiting for withdrawals — is thin. At 82% U, only $18 million of that $100 million pool is free. DeFi rate models exist, fundamentally, to keep that cushion from disappearing.
The kink: two gentle lines and one violent bend
The dominant model in DeFi is the piecewise-linear curve, also called the jump-rate model. It uses four parameters: a base rate, a first slope, a kink utilization, and a second, far steeper slope. The borrow rate R is:
if U ≤ kink: R = base + slope1 × U
if U > kink: R = base + slope1 × kink + slope2 × (U − kink)
Aave's contracts express the same shape normalized to the kink (U ÷ kink below it and (U − kink) ÷ (1 − kink) above it). The geometry is identical: a shallow ramp, then a wall. Below the kink, cheap money encourages productive borrowing. Above it, expensive money punishes new loans and pulls fresh deposits in. Aave's typical kink sits around 85%, asset-dependent and generally in the 80–90% band; Compound's classic calibration is around 80%. The second slope commonly runs 5–10 times the first — sometimes more.
Here is the curve with clearly assumed parameters: base 0%, slope1 8% per 100% utilization, kink 80%, slope2 300% per 100% utilization. Every row is arithmetic from the formulas above — an illustrative curve, not a live quote.
| Utilization | Calculation | Borrow APR |
|---|---|---|
| 50% | 0 + 8% × 0.50 | 4.00% |
| 80% (kink) | 0 + 8% × 0.80 | 6.40% |
| 90% | 6.40% + 300% × 0.10 | 36.40% |
| 95% | 6.40% + 300% × 0.15 | 51.40% |
| 100% | 6.40% + 300% × 0.20 | 66.40% |
Read the gap between the second and third rows twice. Ten percentage points of utilization — an ordinary week of deposit outflows — takes the rate from 6.4% to 36.4%, a 5.7x jump. Past the kink, every single percentage point costs 3% APR in this assumed curve. Your Monday-to-Friday spike from 6% to 14% was almost certainly a pool crossing this line, not a catastrophe. Catastrophe comes later, if it keeps climbing.
Where your supply yield actually comes from
Borrowers pay interest only on what they borrowed, and the protocol skims its share before suppliers see a cent. Everything else is distribution. The supply-side identity:
supply APR ≈ borrow APR × U × (1 − reserve factor)
Worked through: a pool quotes an 8% borrow rate at 80% utilization with a 10% reserve factor. Suppliers receive 8% × 0.8 × 0.9 = 5.76%. Two forces quietly dilute the headline. Idle capital earns nothing, so only 80% of the pool is generating interest, and the reserve factor routes 10% of the interest to protocol reserves — capital the DAO uses to backstop losses and fund operations.
The same identity works in a scare. On the illustrative curve at 95% utilization, borrowers pay 51.4%; suppliers earn 51.4% × 0.95 × 0.9 ≈ 43.95%. Eye-watering deposit APYs during stress are not free money from a generous protocol. They are the other side of someone else's emergency loan — and they vanish the moment utilization falls.
What 100% utilization really means
At U = 100%, every supplied token is borrowed. The idle cushion is zero, and suppliers cannot withdraw until a borrower repays or a new deposit lands. Your withdrawal transaction will simply revert — funds are contractually safe but temporarily inaccessible. In practice the wait is usually short, because the rate response makes repayment and depositing urgently attractive. In a genuine stress event, short is not guaranteed.
Say you supplied 20,000 USDC on a quiet Tuesday and tried to withdraw Thursday, after one large borrower pulled the pool from 72% to 100% utilization overnight. The interface shows your balance in full; the withdrawal keeps reverting. Hours later a fresh deposit unlocks redemptions pro-rata and you leave with most of it; the remainder waits for a repayment that arrives the next day. You lost nothing except timing. Now run the same scenario on funds you needed for a tax payment or a margin call elsewhere. That is the real lesson of utilization: availability is a pool-level fact, not an account-level promise, so money that must be liquid on demand should never sit in a single high-utilization pool.
This is where the jump slope earns its name. Rates run toward 50%, 100%, and beyond, and the pain propagates through the system in a predictable sequence. Leveraged borrowers — including the liquid-staking looping strategies that looked mathematically flawless at 3% borrow — watch interest costs explode and health factors shrink as debt accrues faster than the position earns. Some hit liquidation. Forced collateral sales pressure prices further. Suppliers who wanted out yesterday earn spectacular APYs they cannot access. The rate spike is the cure: it rations credit and auctions liquidity to whoever supplies it next. The cure works. It is also brutal to live through.
Same math, four different architectures
Knowing the curve shape is not enough in 2026 — you also need to know who sets each curve and what your risk is isolated from. The big venues genuinely differ:
- Aave V3 runs shared, cross-collateral pools with per-asset governance-set curves, plus isolation and efficiency modes for risk segmentation. Deep liquidity, conservative parameter changes.
- Compound V3 builds one base asset per deployment with separate supply and borrow curves — the supply identity is baked into market configuration rather than derived live.
- Morpho Blue uses permissionless, isolated markets: one collateral, one loan asset, one creator-chosen curve. There is no shared pool for a bad market to drain — but a badly calibrated curve or risky collateral is entirely on the participants in that market.
- Spark ties its rates to Sky ecosystem governance rather than pure market-clearing. The Sky Savings Rate was reported at 3.60% APY on September 16, 2026, administered and funded from protocol surplus — predictable, but not the same signal as a crowded pool.
| Protocol | Rate architecture | Typical kink | Approx. TVL |
|---|---|---|---|
| Aave V3 | Shared pools, per-asset curves | ~85% (80–90%) | ~$19.4B |
| Spark | Governance-linked to Sky (SSR) | Administered | ~$6.8B |
| Morpho Blue | Isolated markets, creator-set curves | Per market | ~$4.9B |
| Compound | Pooled V3, separate supply/borrow curves | ~80% classic | ~$2.7B |
Figures are approximate snapshots from April–September 2026 reporting and move daily. The deeper point: a 6% loan means different things on each row. On Aave it reflects pool demand against a known governance curve. On Morpho it reflects one market creator's choices. On Spark it reflects an administered rate. Side-by-side comparisons help — start with Aave vs Compound and Aave vs Morpho, then browse the best lending protocols.
Before you click borrow: the four checks
- Measure distance to the kink, not just the rate. A 6% loan at 55% utilization has runway. The same 6% at 84% is a pin waiting for a tap.
- Treat high-utilization cheap loans as spike risk. If a loan looks underpriced while the pool is near its kink, you are being sold the gentle segment of a curve that is about to bend.
- Depositors: check available liquidity before APY. A fat yield with 2% idle capital is a loan of your funds to the withdrawal queue. Liquidity data sits next to every market's rate.
- Model interest cost across scenarios. Run the rate today, the rate at the kink, and a stressed rate. If the position only works at today's number, it doesn't work.
Sources and further reading
- Aave V3 docs — Borrow interest rate strategy — utilization, optimal utilization and the two-slope model.
- Compound v3 docs — market architecture, per-market configuration and rate parameters (data aggregated through Sep 16, 2026).
- Morpho docs — isolated markets, vaults and permissionless market creation.
- Sky.money app — current Sky Savings Rate and sUSDS supply data.
Frequently asked questions
Why did my DeFi borrow rate jump from 6% to 14% in a few days?
Because the rate is a function of pool utilization — total borrowed divided by total supplied. When utilization crosses the kink (roughly 80–90% by asset and protocol), the curve switches to a slope 5–10 times steeper. A pool drifting from ~78% to ~88% can more than double your rate with no hack, no vote and no malfunction.
What is the kink in a DeFi interest rate model?
The kink is the utilization threshold where the curve bends. Below it, rates rise gently; above it, a jump slope punishes borrowing and attracts deposits. Aave's typical kink is around 85% (roughly 80–90% depending on asset) and Compound's classic calibration is around 80%. Aave's name for it, optimal utilization, is the same concept.
How is supply APY related to borrow APR?
Supply APR ≈ borrow APR × utilization × (1 − reserve factor). Interest is only earned on the borrowed fraction, and the protocol takes its reserve-factor cut first. At 8% borrow, 80% utilization and a 10% reserve factor, suppliers earn 8% × 0.8 × 0.9 = 5.76%.
What happens at 100% utilization?
All supplied capital is borrowed, so suppliers can't withdraw until repayments or deposits arrive. Borrow rates run up their steepest segment — 50–100%+ in severe stress — to pull capital in, which worsens leveraged positions and can feed liquidations. The spike is the mechanism restoring liquidity. It works, but you do not enjoy holding through it.
Do Aave, Compound, Morpho and Spark set rates the same way?
The math is related; the architecture is not. Aave and Compound use pooled, governance-calibrated markets, with Compound V3 running separate supply and borrow curves. Morpho Blue uses isolated markets with individually set curves and no shared pool. Spark is governance-linked to the Sky ecosystem — the Sky Savings Rate was reported at 3.60% APY on September 16, 2026 — rather than pure market-clearing.
Model interest cost before you borrow
Run today's rate, the rate at the kink, and a stressed scenario — then see exactly what each one costs over the life of the loan.
Open the Yield CalculatorRelated reading: Gas Fees vs Yield, Liquid Staking Looping, and the full lineup of lending protocol reviews.
A variable borrow rate is not a price. It is a starting position on everyone else's future behavior — depositors who might leave, borrowers who might pile in, liquidators who always show up. So check your pool's utilization, count the percentage points to the kink, and stress the loan before you take it. The curve will bend whether you watched it approach or not. What's the utilization on the loan you're holding right now?