Every DeFi yield is paid by someone: borrowers, traders, liquidated borrowers or a protocol printing its own token. Here is how each rate is built, two worked examples with the arithmetic, how they compare with a Treasury bill, and the risks the headline number leaves out.
Market risk reading, cross‑asset context and an events calendar: see Exent Market intel →
From whoever is on the other side. A DeFi yield is real only if someone pays it: a borrower paying interest, a trader paying a swap fee, a borrower losing a liquidation penalty, or the protocol paying in tokens it creates. Bank for International Settlements researchers found that deposits into DeFi lending are driven mainly by the search for yield, while borrowing is driven mainly by speculation (BIS Working Paper 1183, May 2, 2024). So DeFi yields are, at bottom, the price speculators pay for leverage and trading.
Scope: this page explains how each kind of yield is built, with hypothetical worked examples, as of October 26, 2026. It does not quote live rates, recommend any protocol or token, or cover staking rewards paid by blockchains themselves or tax. For the basics of how DeFi works and its dated incidents, start with what is DeFi?; for the coins most DeFi runs on, see are stablecoins safe? It is education, not advice.
| Source | Who pays you | What sets the rate | What can go wrong |
|---|---|---|---|
| Lending interest | Borrowers, often leveraged traders | How much of the pool is lent out (utilization) | Pool fully lent out, so you cannot withdraw; bad debt if collateral is not sold in time |
| Swap fees | Traders swapping in a pool | Trading volume × fee rate ÷ pool size | Price moves cost more than fees earned (impermanent loss) |
| Liquidation bonus | Borrowers whose collateral falls too far | Protocol’s liquidation penalty | Earned by liquidators, not passive depositors; fails in congested crashes |
| Token rewards (emissions) | The protocol, in its own new tokens | Governance-set emission schedule | Reward token price falls as new supply is sold |
| Interest-like stablecoin rewards | A platform or issuer from reserve income | Treasury bill rates | Barred for US stablecoin issuers from January 18, 2027 under the GENIUS Act |
By a formula, not a banker. Most lending pools set the borrow rate from utilization, the share of deposits currently lent out. The rate rises slowly up to a target level (the “optimal” utilization) and steeply above it, to pull in deposits and push out borrowers. What depositors earn is the borrow rate, scaled down by how much of the pool is lent and by the protocol’s cut (the reserve factor). That is the formula in Aave’s published v3 rate contract (Aave v3 source code; Aave docs):
Worked example with hypothetical parameters for a dollar-stablecoin pool: base rate 0%, slope 1 6%, slope 2 60%, target 90%, reserve factor 10%. Real pools set their own values by governance vote and change them.
| Share of pool lent out | Borrow rate | Depositor rate | Deposits free to withdraw |
|---|---|---|---|
| 50% | 6% × 50/90 = 3.33% | 3.33% × 0.50 × 0.90 = 1.50% | 50% |
| 80% | 6% × 80/90 = 5.33% | 5.33% × 0.80 × 0.90 = 3.84% | 20% |
| 90% (target) | 6.00% | 6.00% × 0.90 × 0.90 = 4.86% | 10% |
| 95% | 6% + 60% × (95 − 90)/(100 − 90) = 36.00% | 36.00% × 0.95 × 0.90 = 30.78% | 5% |
Read the jump correctly: the highest rate appears exactly when the pool is short of cash, so the depositor being paid 30% may be unable to leave until borrowers repay or new deposits arrive. In normal utilization, the hypothetical pool pays roughly what a Treasury bill pays, with smart-contract, collateral and stablecoin risk on top. BIS researchers also found that retail deposits move with interest rates available in the real economy, so when bill yields rise, the gap that attracted deposits narrows.
Fees, minus what the pool gives up when prices move. In a standard two-token pool the product of the two balances stays constant, so when one token rises, traders buy it out of the pool and the depositor ends up holding less of the winner and more of the loser. In the original Uniswap design every swap pays 0.30%; since governance turned on protocol fees in December 2025, 0.25% goes to liquidity providers and 0.05% to the protocol on those pools (Uniswap docs, Fees).
Hypothetical: $10,000 deposited as 50 of token X at $100 plus 5,000 dollar stablecoins. Balance product 50 × 5,000 = 250,000.
| Token X rises 50% to $150 | Token X falls 50% to $50 | |
|---|---|---|
| Pool now holds | √(250,000 ÷ 150) = 40.82 X and √(250,000 × 150) = 6,123.72 stablecoins | √(250,000 ÷ 50) = 70.71 X and √(250,000 × 50) = 3,535.53 stablecoins |
| Pool position value | 40.82 × $150 + $6,123.72 = $12,247.45 | 70.71 × $50 + $3,535.53 = $7,071.07 |
| Value if simply held | 50 × $150 + $5,000 = $12,500 | 50 × $50 + $5,000 = $7,500 |
| Impermanent loss, before fees | $252.55 (2.02%) | $428.93 (5.72%) |
| Fees needed to break even, at a hypothetical $2.50 a day | $252.55 ÷ $2.50 = 101 days | $428.93 ÷ $2.50 = 172 days |
The $2.50 a day assumes the depositor owns 0.1% of a $10 million pool that trades $1 million a day at a 0.25% provider fee: $1,000,000 × 0.25% × 0.1% = $2.50, about 9.1% a year on $10,000 if volume never changed. Two things the example shows. First, the loss is the same for a halving as for a doubling, as the Uniswap documentation notes; the doubling case is worked in what is DeFi? Second, in the falling case the depositor is down about $2,929 on the original $10,000 before fees, more than the $2,500 an investor who just held would have lost. Fees offset the gap only if trading volume holds up.
Impermanent loss versus holding, before fees, from the formula 2 × √r ÷ (1 + r) − 1 where r is the price ratio (Uniswap docs, Understanding Returns). Newer pool designs let depositors concentrate their money in a price range: only in-range money earns fees, and both the fees and the loss per dollar get larger.
They pay liquidators, not ordinary depositors. When a borrower’s collateral value falls so that its health factor drops below 1, anyone may repay part of the loan and take collateral at a discount, the liquidation bonus, which the borrower pays (Aave v4 docs). Depositors benefit indirectly: liquidations are what keep loans covered. The mechanics and a worked liquidation are in what is DeFi?
The weak point is a crash. On March 12–13, 2020, as ether fell, congested blockchain transactions left liquidation auctions on the Maker protocol with almost no bidders; some collateral went for zero bids. Its governance forum recorded that the shortfall used up a roughly $500,000 surplus buffer and about $4.5 million more, which had to be covered by auctioning newly created governance tokens (Maker governance forum, March 2020). Lenders are only as safe as the liquidation system on its worst day.
Token rewards are new coins the protocol issues to attract deposits, usually its governance token, whose holders vote on the protocol’s rules (BIS Quarterly Review, December 2021). They are paid in the protocol’s own currency, so the yield is only worth what the token fetches when sold, and recipients selling the new supply pushes that price down. Hypothetical: a pool advertises 20% a year, made of 3% from swap fees and 17% in reward tokens.
| Reward token over the year | Fee part | Reward part | Fee and reward proceeds / initial deposit |
|---|---|---|---|
| Price unchanged | 3% | 17% | 20.0% |
| Supply grows 50%; assumed market capitalization unchanged; price ÷ 1.5 | 3% | 17% ÷ 1.5 = 11.3% | 14.3% |
| Price falls 60% | 3% | 17% × 0.40 = 6.8% | 9.8% |
| Price falls 90% | 3% | 17% × 0.10 = 1.7% | 4.7% |
In the last row, a 20% headline ends up close to the 4.22% bill yield, with all of DeFi’s risks attached. Reward systems are code too: the BIS review records an October 2021 error at the Compound lending platform that distributed rewards worth $90 million incorrectly, with no administrative tool to stop it.
A Treasury bill is the floor any DeFi yield has to beat after its risks. The Fed’s policy range is 3.75%–4.00%, raised on September 16, 2026 and held on October 28 (FOMC statement), and the 3-month bill yielded 4.22% on October 23, 2026 (US Treasury). A bill ETF such as SGOV showed a 3.76% 30-day SEC yield on October 22, 2026. Tokenized Treasury funds put the same bills on a blockchain, with an issuer and custodian behind them (see tokenized Treasury funds).
| Product | Yield shown here | Who pays it | What protects you |
|---|---|---|---|
| 3-month Treasury bill | 4.22% (Oct 23, 2026) | US Treasury | Full faith and credit of the US government |
| T-bill ETF (SGOV) | 3.76% 30-day SEC yield (Oct 22, 2026) | Treasury bills, minus a 0.09% fee | Registered fund; assets held by a custodian |
| Tokenized Treasury fund | Bill yield minus fund fee; quoted on different conventions | Treasury bills | Fund documents and custodian; eligibility limits |
| DeFi stablecoin lending (hypothetical model) | 1.50%–4.86% at 50–90% utilization | Borrowers | Collateral and liquidations; no insurance |
| DeFi trading pool (hypothetical) | About 9.1% in fees, minus impermanent loss | Traders | Nothing beyond the code |
| Emission-boosted pool (hypothetical) | 20% advertised, 4.7%–20% realized | New token supply | Nothing beyond the token’s price |
The rules are moving too. From January 18, 2027 the GENIUS Act bars US payment-stablecoin issuers from paying interest for simply holding a coin (GENIUS Act explained). The Senate Banking version of the CLARITY Act would extend a similar ban to platforms while allowing activity-based rewards, but it is not law (CLARITY Act explained). How stablecoin reserves link to the bill market is in stablecoins and Treasury bills.
Five kinds, each with a documented case. The yield is the same whether or not one of them happens to you:
| Risk | Dated example | Source |
|---|---|---|
| Smart-contract exploit | March 13, 2023: about $197 million taken from Euler V1 through a code flaw; assets were returned only after weeks of negotiation with the attacker | Euler Labs; Euler Foundation |
| Price manipulation with borrowed money | October 2020: a flash-loan attack on the Harvest platform caused $27 million of losses | BIS Quarterly Review |
| Liquidation failure | March 12–13, 2020: Maker collateral auctions cleared at zero bids; about $4.5 million of bad debt beyond the buffer | Maker governance forum |
| Stablecoin depeg | May 2022: TerraUSD, marketed with up to 20% yield, collapsed, wiping out $40 billion; March 2023: USDC fell to $0.88 | SEC; New York Fed |
| Bridge, key and oracle risk | Reported March 29, 2022: $620 million stolen from the Ronin bridge, attributed by the FBI to North Korean hackers | FBI |
Euler reported full recovery in this incident (Euler Labs, January 10, 2024); recovery after another exploit cannot be assumed. Counterparty risk also hides inside “decentralized” positions: a wrapped token depends on whoever holds the original, a stablecoin on its issuer, and a liquidation on the price feed (oracle) that triggers it. The Financial Stability Board found that DeFi inherits the weak points of traditional finance, leverage, liquidity mismatch and interconnection, often amplified (FSB, February 16, 2023).
Counterexample: a saver sees a stablecoin pool paying 30% and moves a $20,000 emergency fund into it. By the formula above, a rate that high usually means the pool is almost fully lent out, which is the moment withdrawals can stall, and the money also carries exploit, depeg and liquidation risk with no insurer. Money needed on a known date, or in an emergency, belongs in insured deposits or Treasury bills (see where to park large cash). Anyone who does use DeFi can limit the damage by sizing it as money they could lose entirely.
Market intel provides market context, cross-asset analysis, an events calendar and regular written notes. These can inform your review of an existing plan; they do not reliably predict crashes or guarantee better outcomes.
From borrowers paying interest, traders paying swap fees, borrowers losing a liquidation penalty, or the protocol paying its own newly issued tokens. BIS research found DeFi borrowing is mainly speculative, so most real yield is the price traders pay for leverage.
From utilization. In a two-slope model the borrow rate rises gently up to a target share of the pool lent out and steeply above it; depositors earn the borrow rate × utilization × (1 − the reserve factor). With hypothetical parameters, 80% utilization pays depositors 3.84%.
Usually because the pool is nearly fully lent out, a token reward is inflating the headline, or the yield is compensating for a risk. In the worked example, 95% utilization pays 30.78%, but only 5% of deposits can be withdrawn at that moment.
In a standard 50/50 pool, about 2.0% versus simply holding if one token rises 50%, and 5.7% if it falls 50%, before fees. On a $10,000 deposit that is $252.55 and $428.93.
Only if volume is high enough for long enough. At a hypothetical $2.50 a day in fees on $10,000, covering the loss from a 50% fall takes about 172 days, and the depositor is still down on the original stake.
Not on a risk-adjusted basis by default. The 3-month bill yielded 4.22% on October 23, 2026 with US government backing; DeFi yields carry exploit, depeg and liquidation risk with no insurance, so only the spread above the bill is extra pay, and it pays for those risks.
New tokens a protocol issues to reward depositors. Their value depends on the token price, which can fall as recipients sell; in the worked example, a 20% advertised yield with 17% in rewards realizes 4.7% if the token falls 90%.
Losing the principal, not a lower rate: code exploits, stablecoin depegs and failed liquidations can each wipe out deposits. A lending protocol lost about $197 million to an exploit on March 13, 2023, and TerraUSD’s collapse wiped out $40 billion.
US payment-stablecoin issuers may not pay interest for simply holding a coin once the GENIUS Act takes effect on January 18, 2027. Interest earned by lending a stablecoin in a DeFi pool comes from borrowers and carries that pool’s risks.
Market intel provides a Risk ON / Neutral / Risk OFF reading from four pillars, cross-asset analysis, an events calendar and regular written notes. Context for reviewing a plan, not a crash forecast.
Educational content, not investment advice. Examples are illustrative; past performance does not guarantee future results.