How are DeFi interest rates set?

TL;DR. DeFi interest rates are set in one of three ways: by a formula in a lending pool (a program that holds many lenders' deposits), by a market that matches bids and offers, or by the parties to a single loan naming their own price. In pools such as Aave and Compound, the formula raises the rate as utilization, the share of the pool that is borrowed, rises, and steeply so past a point called the kink. Auction and offer markets usually fix the rate for the term, and in peer-to-peer loans each lender quotes a rate for one borrower. The mechanism decides whether your rate can change after you borrow and what the number is pricing.

Who actually sets the interest rate on a DeFi loan?

The interest rate on a DeFi loan is set by a formula, by a market, or by the borrower and lenders of a single loan. Lending pools such as Aave and Compound use a formula, auction and offer markets use bids, and peer-to-peer loans use a rate quoted for one loan.

A lending pool is a smart contract, a program on a blockchain, that holds many lenders' deposits of one token and lends them to borrowers who lock up collateral: assets that can be taken and sold to cover the debt if their value falls too close to the amount owed (a liquidation). APR, the annual percentage rate, is used in DeFi to mean the yearly interest rate before compounding and before any fees; rates here are APRs in that sense unless we say otherwise. For the steps of a loan, see How DeFi loans work.

We have designed loans at both ends of this table: the pool described in our first Field Notes post uses a utilization curve, and the crowd loan model we are building at Unlloo, in which many lenders fund one borrower's loan, is priced by lender offers.

What is the utilization rate, and why does it drive pool rates?

The utilization rate is the share of a lending pool's deposits that is currently borrowed: the borrowed amount divided by the supplied amount. Pools raise the borrow rate as utilization rises because a pool at 100% utilization has no tokens left for lenders who want to withdraw, and the rate is the main tool the contract has to attract deposits and encourage repayment.

A Banque de France blog post from April 2024 describes the borrow rate on DeFi lending platforms as an increasing function of utilization. An early academic review, Gudgeon and co-authors (2020), describes such rate rules as seeking to balance supply and demand for funds.

Utilization moves from both sides: it rises when someone borrows and when a lender withdraws. The Block reported on 29 July 2025 that, more than a week earlier, roughly 167,000 ETH had been withdrawn from Aave's lending pool, driving the borrow rate on wrapped ETH (ether in token form) from about 3% to double digits. Existing borrowers paid more because deposits had left. For what a full pool means for a lender, see Is it safe to lend crypto?

A utilization curve prices how scarce the pool's liquidity is. It does not price you.

In pools such as Aave V3 and Compound III, everyone borrowing the same token from the same pool pays the same rate at the same moment, whatever their history. Aave V4, launched on Ethereum on 30 March 2026, adds a User Risk Premium, which its documentation and an Aave Labs post of 24 July 2025 describe as an extra interest charge on top of the shared rate, based on the quality of the borrower's collateral. The borrower's history still plays no part.

What is the kink, and why do rates jump above it?

The kink is the utilization level, often 80% or 90%, where a pool's rate curve switches from a gentle slope to a steep one. Above the kink, borrowing gets expensive quickly by design, so that borrowers repay and new deposits arrive until the pool has spare liquidity again.

Aave's documentation calls the kink the optimal usage ratio, and the Banque de France post puts the usual target at 80% or 90%. One common form of the formula, used in Aave's contracts and written out in a poster presented at the ECB's Money Market Conference in November 2025, has four parameters: a base rate, the kink and two slopes. Each slope is the total rise in the rate across its segment: slope1 from zero utilization to the kink, slope2 from there to 100%.

• Below the kink: borrow rate = base + slope1 × (utilization / kink)
• Above the kink: borrow rate = base + slope1 + slope2 × ((utilization - kink) / (1 - kink))

Compound III writes the same shape with slopes multiplied directly by utilization, so its slope numbers are not comparable with Aave's.

The table runs a hypothetical curve through the formula above. Its parameters are invented and describe no real market. The supply rate is what lenders earn after the reserve factor, the share of interest the protocol keeps; the next section derives it.

From 50% to 80% utilization the borrow rate rises 1.5 percentage points. From 80% to 85% it rises 15.

The formula sets the rate, but people set the formula. On 28 March 2025 Chaos Labs, then a risk manager for Aave, posted on Aave's governance forum that it would lower the first slope on listed stablecoins (tokens built to hold a steady value, usually one US dollar) by half a percentage point, from 6.50% to 6.00% on most. It said it would use Aave's Risk Steward, described in a proposal of 15 January 2024 by BGD Labs, then a developer for Aave, as a tightly constrained contract that let Aave's risk providers adjust such parameters. BGD Labs and Chaos Labs have both since left Aave.

The lending protocol Morpho takes a different position. Its documentation says a market's rate model is chosen at creation from a governance-approved set and is then immutable: neither governance nor the market's creator can change it. The model instead shifts its own curve up or down over time to hold utilization near a 90% target.

Picking these numbers is a trade-off. The curve in our first post has a 5% floor, a kink at 80% and a 50% ceiling. A lower kink leaves lenders a bigger idle buffer and a more diluted yield. Set the ceiling too low and a nearly fully lent pool does little to draw in deposits or put off new borrowing; set it too high and whoever borrows last pays a rate meant only as a short-lived signal.

Why is the borrow rate higher than the supply rate?

In a lending pool, the borrow rate is higher than the supply rate because borrowers pay interest only on the borrowed part of the pool, while that interest is shared across every deposit, including idle ones, after the protocol takes a cut. The usual relationship is supply rate = borrow rate × utilization × (1 - reserve factor), where the reserve factor is the protocol's cut.

Take a hypothetical pool with a 10% reserve factor:

1. Lenders supply 100 tokens and borrowers take 80 of them at 6%.
2. Borrowers pay 80 × 6% = 4.80 tokens of interest over a year.
3. The protocol keeps 10% of that, 0.48 tokens.
4. The remaining 4.32 tokens are spread over all 100 supplied tokens: a 4.32% supply rate.

The same formula appears in the RareSkills walkthrough of Aave V3 and, with the protocol's cut called a fee, in Morpho's documentation. Compound III is an exception: its documentation describes separate rate models for supply and borrow.

Lending yield, then, comes from borrowers. A supply rate above the formula's is topped up from elsewhere, such as reward tokens: the Banque de France post puts the median extra rate from Compound's COMP token rewards at around 1.9% in 2023.

What is the difference between APR and APY on a DeFi loan?

In DeFi, APR is the yearly interest rate without compounding, and APY, the annual percentage yield, is the yearly rate with compounding. Because on-chain interest often accrues every second, a loan quoted at 12% APR costs 12.75% APY if it runs a full year with per-second compounding, and 40% APR becomes 49.18% APY.

Morpho's documentation defines borrow APY as e^(rate per second × seconds per year) - 1, the formula behind those two figures. On Aave V3, a Cyfrin Updraft lesson explains that borrower debt compounds every second while lender interest grows linearly between rate updates.

The two numbers are only comparable like with like: APY against APY on a pool loan where debt compounds, APR against APR on a fixed-term loan with simple interest.

Are DeFi interest rates variable or fixed?

DeFi interest rates are variable in lending pools such as Aave and Compound, where the rate can change with every transaction that moves utilization. Fixed rates exist where a loan has a set term and the rate is agreed at the start, for example in auction markets, offer-matching markets and peer-to-peer loans.

Aave's help centre tells borrowers that rates adjust as liquidity is borrowed or repaid. Older guides describe a "stable" rate mode on Aave. It is gone. New stable-rate borrowing was halted after a critical bug report on 4 November 2023, BGD Labs proposed removing the mode on 31 January 2024, and its Aave V3.2 upgrade, live on every active Aave V3 instance by 8 October 2024, removed the stable-rate logic from the code.

Fixed rates on-chain can come from a market. Term Finance, which describes itself as a fixed-rate protocol for collateralized lending, runs sealed-bid auctions: participants submit bids and offers that stay hidden until the market clears, the protocol computes one clearing rate, and every matched lender and borrower transacts at it. Morpho announced an offer-matching design on 12 June 2025 as Morpho V2: market-driven fixed-rate, fixed-term loans in which lenders and borrowers state their terms and are matched. It has since named its fixed-rate protocol Morpho Midnight.

A fixed rate can also start from a curve that locks each borrower's rate when they borrow, as in the design in our first post. The borrower gets certainty and the pool carries the rate risk: if utilization rises later, lenders hold loans priced for a quieter pool.

How are rates set when lenders make offers instead of a formula?

When lenders make offers in a per-loan design, the borrower states the highest rate they will pay, each lender offers an amount at a rate at or below that cap, and the offers the borrower accepts add up to a blended rate that belongs to that loan alone. The price reflects what lenders think of this borrower, not how busy a pool is.

Take a hypothetical request for 1,000 USDC, a dollar stablecoin, for 90 days with a 15% APR cap. Four lenders offer 1,200 USDC between them and the borrower accepts the three cheapest; the fourth offer stays refundable. Interest is shown as simple interest before any fees, for illustration; how interest accrues and what is deducted from it depend on the protocol.

The blended rate is (400 × 10% + 300 × 12% + 300 × 14%) / 1,000 = 11.8% APR, and each lender's interest is their amount × their rate × 90 / 365.

The example is pay-as-offered: each lender earns the rate they quoted, so a lender who quotes low to win a place earns less than the next one. That is a design choice: a single-clearing-rate auction such as Term Finance's pays every matched lender the same rate.

Compared with a pool, per-loan offers have costs:

• It needs lenders to show up. A pool lends instantly. A request waits for offers, and a thin market means a worse rate or no loan.
• Every lender does their own pricing. A pool lender deposits and takes the pool's rate. A lender making offers has to judge each borrower, and can judge wrong.
• An unaccepted offer earns nothing from that loan. It stays refundable.
• An accepted offer is committed until the loan is repaid. A pool lender can usually withdraw while the pool has unlent tokens.

A crowd loan is priced this way. What lenders look at when they price a person is on-chain reputation, a measure of a wallet's trustworthiness built from public blockchain data, which we take apart in The anatomy of on-chain reputation.

Why do loans without collateral carry higher rates?

Loans without collateral carry higher rates because the rate has to cover losses that collateral would otherwise absorb. With no collateral to liquidate, a lender prices the chance of default into the rate, on top of the cost of money over time.

Take a hypothetical lender who expects 5 of 100 similar loans to default with nothing recovered: the other 95 must each pay about 5.26% over the term for the lender to break even (1 / 0.95 - 1 = 0.0526), before any return for time and risk.

A loan backed by less collateral than the amount borrowed, or by none, is an under-collateralized loan; see the glossary entry for the lender's risk, crypto loans without collateral for the borrower's side, and what happens if you don't repay a crypto loan for default.

Do central-bank rates move DeFi rates?

Central-bank rates move DeFi rates only partly, on the evidence available, which covers the US Federal Reserve and dollar stablecoins. Research presented at the ECB's Money Market Conference in November 2025, using Aave V2 and V3 data on Ethereum for USDT, USDC and DAI from May 2021 to May 2025, estimates that about 0.3 of a Federal Reserve rate change reaches stablecoin deposit rates in the short run and 0.5 to 0.7 in the medium run.

The poster by Barbon, Barthélemy and Nguyen finds DeFi rates mostly driven by crypto-related factors and describes two channels pulling in opposite directions. A Fed rate rise draws lenders out of DeFi into US Treasury bills, short-term government debt, which pushes DeFi rates up. It also reduces the appetite for leverage, which a Warwick Business School article of February 2024 describes as traders borrowing stablecoins to buy more crypto, and that pushes DeFi rates down.

How do you work out what a quoted rate will actually cost?

To estimate what a quoted rate will cost, multiply the amount by the APR and by the days held divided by 365: 1,000 USDC at 12% APR for 90 days costs about 29.59 USDC in interest. If that interest compounds every second the cost is 30.03 USDC, and if the 12% was an APY it is 28.34 USDC.

With t as days divided by 365, the three columns are amount × rate × t, amount × (e^(rate × t) - 1) and amount × ((1 + rate)^t - 1).

A quoted rate also leaves out three things:

• Early repayment. If interest stops when you repay, the 90-day example repaid on day 45 costs 14.79 USDC of simple interest, not 29.59. Not every loan works that way.
• Fees on top. Origination fees, charged when a loan is opened, and other charges sit outside the rate.
• The worst case. On a fixed kinked curve, the rate at 100% utilization is the most a pool can charge. On a curve that shifts itself, such as Morpho's, that rate rises the longer the pool stays full.

This article is education, not financial, legal or tax advice. Which lending products you can use and how loan interest is treated vary by jurisdiction.

Sources

1. Banque de France, "Interest rates in decentralised finance", blog post no. 352, 9 April 2024. The borrow rate as an increasing function of utilization, the 80% or 90% target, and the median extra rate from COMP rewards in 2023.
2. Barbon, Barthélemy and Nguyen, "DeFi-ying the Fed? Monetary policy transmission to stablecoin interest rates", poster at the ECB Money Market Conference, 6 and 7 November 2025. The kinked rate formula, the data set (USDT, USDC and DAI on Aave V2 and V3 on Ethereum, May 2021 to May 2025), the finding that DeFi rates are mostly driven by crypto-related factors, the pass-through estimates of 0.3 and 0.5 to 0.7, and the two transmission channels.
3. Gudgeon, Werner, Perez and Knottenbelt, "DeFi Protocols for Loanable Funds: Interest Rates, Liquidity and Market Efficiency", arXiv, 11 June 2020 (version 3, 15 October 2020). An early academic review of how Compound, Aave and dYdX set interest rates, which describes the rate rules as seeking to balance supply and demand for funds.
4. Aave, "Interest Rate Strategy", Aave V3 documentation, undated. The two slopes on either side of the optimal usage ratio.
5. Aave DAO on GitHub, "DefaultReserveInterestRateStrategyV2.sol", Aave V3 contract code, undated. The form of the two-slope formula that Aave's contracts use.
6. Aave governance forum, "Chaos Labs Risk Stewards - Reduce Interest Rates for Stablecoins on Aave V3 - 03.28.25", 28 March 2025. People changing a curve parameter: the first slope lowered by half a percentage point on the listed stablecoins, from 6.50% to 6.00% on most, through the Risk Steward process.
7. Aave governance forum, "[ARFC] BGD. Risk Steward Phase 2: RiskSteward", 15 January 2024. What the Risk Steward is and that interest rate parameters, including the first slope, are within its limits.
8. The Block, "BGD Labs to cease Aave contributions after four years as governance tensions grow", 20 February 2026. BGD Labs saying it would end its work for Aave after 1 April 2026.
9. The Block, "Top Aave risk manager Chaos Labs exits amid governance dispute", 6 April 2026. Chaos Labs stepping down as Aave's risk manager, and the earlier departure of BGD Labs.
10. Compound III Docs, "Interest Rates", undated documentation. The utilization formula, Compound III's own way of writing the kinked formula, and the separate supply and borrow rate models set by governance.
11. Morpho Docs, "Interest Rate Model", undated documentation. The rate model chosen at market creation and then immutable, the adaptive curve that targets 90% utilization and shifts over time, and the supply APY and borrow APY formulas.
12. The Block, "Ethereum validator exit queue reaches highest point ever", 29 July 2025. The wrapped ETH borrow rate on Aave moving from about 3% to double digits after roughly 167,000 ETH was withdrawn more than a week before the report.
13. RareSkills, "The interest rate model of Aave V3 and Compound V2", originally published 1 December 2023, last updated 29 August 2025. The supply rate formula with the reserve factor.
14. Cyfrin Updraft, "Linear Interest", Aave V3 course, last updated 25 July 2025. Borrower debt compounding every second and lender interest accruing linearly on Aave V3.
15. Aave Help Centre, "Borrow Tokens", 9 August 2024. Borrow rates following utilization and adjusting as liquidity is borrowed or repaid.
16. Aave governance forum, "BGD. Full deprecation of stable rate", 31 January 2024. The 4 November 2023 bug report, the halt of stable-rate borrowing and the plan to move positions to variable.
17. Aave governance forum, "BGD. Aave v3.2: Liquid Emodes", 13 September 2024 (updated 8 October 2024). The removal of the stable-rate logic from the Aave V3 code, and the upgrade of all active Aave V3 instances.
18. Term Finance Docs, "Introduction to Term Finance", undated documentation. Term Finance as a fixed-rate protocol for collateralized lending.
19. Term Finance Docs, "Term Auctions", undated documentation. The sealed-bid auction format and the single clearing rate.
20. Term Finance Docs, "Auction characteristics", undated documentation. Bids and offers staying hidden until the market clears.
21. CoinDesk via Yahoo Finance, "Crypto Lending Platform Morpho V2 Brings DeFi Closer to Traditional Finance", 12 June 2025. The Morpho V2 announcement of market-driven fixed-rate, fixed-term loans.
22. Morpho, "The Morpho Effect: April 2026", monthly update for April 2026 (the page shows no day). The name Morpho Midnight for Morpho's fixed-rate protocol.
23. The Block, "Aave V4 launches on Ethereum mainnet with 'hub-and-spoke' architecture", 30 March 2026. The Aave V4 launch date.
24. Aave, "Aave v4", documentation, undated. The definition of the User Risk Premium.
25. Aave Labs, "Aave V4 Risk Premiums", 24 July 2025. A dated description of the risk premium as an extra amount on top of the base rate that depends on the risk of the collateral.
26. The Block, "Wildcat, the decentralized credit platform built by 'Crypto Twitter' mainstay Laurence Day, launches new version on Ethereum", 6 February 2025. Wildcat as an example of markets where the borrower sets the rate.
27. Code4rena, "The Wildcat Protocol Findings & Analysis Report", 24 October 2024. A Wildcat borrower's ability to reduce a market's APR and the reserve ratio change that follows.
28. Warwick Business School, "How are interest rates set on DeFi platforms?", 16 February 2024. Traders borrowing stablecoins to buy more crypto, which raises demand for stablecoin borrowing.