Ethereum lending, 2021 to 2026, in numbers
Lending on Ethereum went through two booms, a crash and a long shrink between 2021 and 2026. We rebuilt every loan on its main lending protocols, month by month, to see who lent, who borrowed and why the money left.
Corrected on 23 September 2026. The first version said liquidators kept about $900 million in bonuses. That sum counted Maker, Compound v3 and Liquity, where they keep none of the backing, so we took it out. We also fixed a price error that added $10 million to 2024's total.
- $35.1B
- 69%
- $6.3B
- 56%
- $0.55B
- $441M
01 · The questionIs DeFi lending bigger and healthier than in 2021?
At the end of 2021, lending on Ethereum was a four-way race. Compound and Aave each held a bit more than a quarter of all the money lent out. Maker had a quarter. Abracadabra, which lent its own MIM stablecoin against tokens that pay a yield, had most of the rest.
Today, Aave alone holds more than half the market. Compound is down to 3%. Abracadabra has almost nothing left.
DeFi lending protocols are pools of money with rules. Lenders deposit tokens and earn interest. Borrowers post collateral, usually ETH or a token that tracks it, and take out a loan worth less than what they posted. If its price falls too far, bots called liquidators repay part of the loan and take the collateral at a discount. Nobody checks a credit score. The code checks the price.
So is DeFi lending on Ethereum bigger and healthier than it was back then? To find out, Bitquery rebuilt every loan on the main lending protocols from our Ethereum data, month by month, for more than five years. That covers Aave, Compound, Maker and Sky, Spark, Morpho, Fluid, Euler, Liquity, Abracadabra and Maple.
We count real debt: what each borrower owes, up to what their collateral is worth. That leaves out dead markets where unpaid interest keeps piling up on paper. We covered those in our piece on Morpho's zombie markets.
02 · How big it gotTwo booms, a crash and a shrink
The chart falls into four parts. The dashed line is ETH's price.
The first boom. Real debt grew more than six times over in less than a year, to a peak of $23.2 billion in November. Much of it was stablecoins borrowed against other stablecoins.
The crash. Terra's UST stablecoin collapsed the next May, and ETH lost two thirds of its price by the end of the year. By December, real debt was down to a fifth of its peak, $4.4 billion.
The second boom. The climb back was slow, then fast. Borrowers learned to loop staked ETH, borrowing against it to buy more of it, and Ethena's USDe, a dollar token that pays a yield, gave the same trade to stablecoins. Real debt passed its old high and kept going, to a peak of $35.1 billion in August 2025.
The shrink. Then it went into reverse. A year after the peak, ETH had lost more than two fifths of its price, and real debt stood at $18.4 billion, lower than at the end of 2021.
Debt mostly moves with ETH's price. In the second boom it grew more than twice as fast as ETH. Borrowing outside loops kept pace with ETH, while loops grew more than twenty times over. In the shrink, debt and ETH fell almost in step.
The chain itself records more lending debt than that, $25.1 billion. Most of the gap is paper debt in dead markets that nobody is paying down.
03 · Who lendsThe lead changed hands, then Aave pulled away
For three years no lender stayed on top for long. The lead changed hands 11 times. Compound led at the start, and Aave and Compound swapped places through 2021. When prices crashed, Maker ended up on top, because its loans shrank less than the others'.
Aave took the lead for good in February 2024. At the peak it held almost three quarters of all real debt. Since then it has slipped to a little over half, as newer lenders grew. Morpho lets anyone open a lending market for any pair of tokens. Spark is Sky's own lending market. Maple lends to trading firms against collateral held off-chain. Each now holds about a tenth of the market.
These shares count Aave's older versions. Its new v4 held another $204 million at the end of August, which we count with the smaller lenders further down.
04 · Where Maker wentFrom user vaults to wholesale
In 2021, MakerDAO was a lender in the plain sense. People locked ETH in a vault and minted DAI against it. Those user vaults owed $5.4 billion at the end of that year. Today they owe about a tenth of that.
Maker, since renamed Sky, now lends mostly wholesale. For a while it made real-world loans through off-chain vaults, a book that peaked at $3.2 billion. Most of its stablecoin now goes to allocator vaults that lend it on. Two of them hold most of that money, one run by Spark and one called Bloom. Another $4.4 billion backs a swap that turns USDC into Sky's stablecoin one for one.
We leave that wholesale lending out of our totals, so no dollar is counted twice. Where it reaches a borrower, it shows up at the lender they borrow from. Spark alone now holds about a tenth of all real debt.
05 · What borrowing is forHalf of it is a loop
A loop is a loan against collateral of the same kind. Deposit staked ETH, borrow ETH, swap it for more staked ETH, deposit again. Each turn earns the staking yield once more. The debt is real, and the money stays inside the loop.
We merge each wallet's positions across every lender, so a deposit on one and a loan on another still count as one loop.
In the first boom, nearly all loops were dollars. Most were plain stablecoins, DAI and USDC, supplied and borrowed on Compound and Aave. UST, Terra's stablecoin, backed about an eighth of them.
The second boom was a loop boom. At the peak, more than two thirds of all debt was a loop. Most of it was ETH, looped through ether.fi's weETH and Lido's wstETH. The dollar loops ran mostly on Ethena's USDe and tokens built on it, many of them Pendle tokens that lock in a yield until a set date.
A year later, the loop share was back to half.
06 · The fallWhy lending almost halved
In the year after the peak, real debt fell by $16.7 billion, almost half. Two things did most of it.
The first is price. Much of the debt is ETH, and ETH's price fell 44% over the year. A loan of one ETH counts for fewer dollars when ETH is cheaper, even if nobody repays a thing. That alone took $6.3 billion off the total.
The second is people paying back, and most of them were loopers going home. Loops shrank by about $15 billion over the year, while all other borrowing fell by about $2 billion. The dollar loops built on Ethena's USDe almost vanished, from about $7 billion to under $1 billion.
Nine tenths of the fall came off Aave.
| Lender | Aug 2025 | Aug 2026 | Change |
|---|---|---|---|
| Aave | $25.2B | $10.3B | -$14.9B |
| Euler | $1.0B | $220M | -$797M |
| Fluid | $1.3B | $634M | -$638M |
| Compound | $950M | $588M | -$362M |
| Morpho | $2.5B | $2.2B | -$306M |
| Spark | $2.2B | $2.0B | -$188M |
| Maker/Sky | $799M | $628M | -$171M |
| Maple | $1.1B | $1.8B | +$762M |
| Others | $53M | $27M | -$25M |
| All | $35.1B | $18.4B | -$16.7B |
07 · When prices crashLiquidations, year by year
When prices fall fast, liquidations kick in. They are the market's safety valve. When a loan's backing loses too much value, a liquidator repays part of the loan and takes some of the backing at a discount. Three lenders do it another way. On Maker the backing goes to an auction, on Compound v3 to the lender itself, and on Liquity to a shared pool of LUSD holders. Since January 2021, liquidations have cleared $7.5 billion of loans on the main Ethereum lenders.
Most of the busiest days were market crashes, and the record still belongs to one in 2021.
| Day | Debt | Count | Top lender |
|---|---|---|---|
| May 19, 2021 | $441M | 1,854 | Aave v2 |
| Aug 5, 2024 | $386M | 3,252 | Aave v3 |
| Feb 3, 2025 | $328M | 4,833 | Aave v3 |
| Jan 21, 2022 | $269M | 999 | Maker |
| Dec 6, 2022 | $232M | 216 | Maker |
| Oct 10, 2025 | $207M | 1,262 | Aave v3 |
One day in the table had nothing to do with prices. That day, $232 million of Maker loans went through liquidation, all of them backed by stablecoins, most of them by USDC.
| Year | Count | Debt | Average |
|---|---|---|---|
| 2021 | 17,154 | $1.62B | $94K |
| 2022 | 33,527 | $1.90B | $57K |
| 2023 | 3,619 | $159M | $44K |
| 2024 | 13,950 | $938M | $67K |
| 2025 | 43,278 | $1.76B | $41K |
| 2026* | 27,276 | $1.09B | $40K |
| All | 138,804 | $7.47B | $54K |
More debt was liquidated in the crash year than in any other. The year of the peak had the most liquidations, though smaller ones on average. The year between the two booms was quiet. More than 2,000 addresses have liquidated a loan, and the ten busiest were behind more than a quarter of all liquidated debt.
08 · Losses on the booksBad debt that got booked
When a liquidation cannot cover a loan, the lender is left with bad debt. Some lenders write it off on-chain. Until Morpho's first real write-off in the spring of 2024, those write-offs came to less than $50 in all.
| Lender | Booked | First write-off | Biggest |
|---|---|---|---|
| Aave v3 | $125M | Mar 26, 2025 | rsETH, May 6, 2026: $124M |
| Compound v3 | $52.5M | Oct 10, 2025 | rsETH, May 9, 2026: $39.2M; deUSD, Nov 6, 2025: $12.9M |
| Morpho | $4.2M | Apr 24, 2024 | Many small ones |
| Euler | $8K | Jun 28, 2026 | Small |
| All lenders | $182M |
So far, the big lenders have booked $181.5 million of bad debt. Nine dollars in ten of it came in one week in May this year, on Aave and Compound, from loans backed by rsETH, a restaked ETH token. For a different kind of loss on an rsETH position, see how a Safe module drained one.
Booked losses are small next to the paper debt that nobody books. Dead markets carried $6.5 billion of it at the end of August, most of it on Morpho and Kashi.
09 · Who borrowsMore borrowers, and pools at the top
So who is doing all this borrowing?
More people borrow than in 2021. The number of wallets that owe at least $1,000 rose by more than half to a peak last September, and has eased since.
The top has spread out a little. The 100 biggest borrowers held more than two thirds of real debt at the end of 2021 and about three fifths now.
Many of the biggest borrowers are pools. Five of the top ten are vaults that take deposits from many people: Instadapp's Lite vault, ether.fi's Liquid ETH vault and three contracts named Subvault. Together they hold 18% of all real debt, out of the 24.5% the top ten hold. Three more borrow mostly from Maple. We checked each top borrower's own contract on-chain to tell a pool from a single wallet.
| # | Borrower | Lender | Share |
|---|---|---|---|
| 1 | Instadapp Lite | Aave v3 | 6.8% |
| 2 | ether.fi Liquid ETH | Aave v3 | 6.3% |
| 3 | Subvault 0x893a…0080 | Aave v3 | 1.7% |
| 4 | Wallet 0xf34d…a6da | Aave v3 | 1.7% |
| 5 | Subvault 0xcdfa…9c89 | Aave v3 | 1.5% |
| 6 | Subvault 0x3883…e4d7 | Spark | 1.5% |
| 7 | Wallet 0xb99a…bcf5 | Maple | 1.3% |
| 8 | Wallet 0xd938…48dd | Aave v3 | 1.2% |
| 9 | Wallet 0x198a…529a | Maple | 1.2% |
| 10 | Wallet 0x09b8…6b8a | Maple | 1.2% |
10 · The long tailDozens of smaller lenders
Beyond the big names sit dozens of smaller lenders. We mapped 116 of them to their contracts. The number with any borrowing or liquidations in a year nearly tripled between the two booms.
They are small. Together they held roughly $2 billion of debt at the end of August, about a tenth of the market. Another $2.6 billion sat in dead markets at nine of them, and we leave it out. Our count for this group is rough. It uses each lender's own recorded total, and only three in four of those figures could be priced.
11 · The answerBigger at the peak, smaller now
Is DeFi lending on Ethereum bigger than in the first boom? At its peak, yes, by half again. Today it is smaller, with $18.4 billion of real debt against $21.6 billion at the end of that boom. By head count it has grown by about a third.
Is it healthier? In some ways. The losses lenders have written off on-chain are small, and almost all of them came in one week. Crashes still hit hard, but the three biggest days since the crash left lenders with less than $1 million of booked losses between them.
Against that, one lender holds more than half the market. Half of all debt is a loop, and loops leave fast when prices fall. About $15 billion of them went in a year. And a few dead markets carry billions of paper debt with nothing behind it, which inflates every total that counts it.
More people borrow on Ethereum than in 2021. Fewer dollars do, and more of them sit on one lender and one trade.
12 · How we measuredWhat we counted, and what we left out
We rebuilt each loan from the protocols' own on-chain events, from each launch to late September. At every month end we checked our totals against each protocol's own on-chain numbers, 33,680 checks in all.
Real debt caps each borrower's debt at the market value of their collateral. Prices come from each protocol's own feeds first, then from DEX prices. Where a feed kept a collapsed token near $1, we used the price it traded at.
We leave out Maker's stablecoin swap and Sky's wholesale lending, which would count the same dollars twice. Aave's first version counts in the totals, but not in the borrower counts. Loops are judged each month by what each token traded like over the year before, so UST counts as a dollar in the first boom.
We did not measure interest rates over time, or lending on any chain other than Ethereum. For a lending attack on another chain, see the Tectonic exploit on Cronos.
13 · The recordThe contracts behind the numbers
| Contract | Address |
|---|---|
| Aave v3 Pool | 0x8787…a4e2 |
| Aave v2 LendingPool | 0x7d27…c7a9 |
| Spark Pool | 0xc13e…e987 |
| Morpho Blue | 0xbbbb…ffcb |
| Compound v3 USDC | 0xc3d6…cdc3 |
| Compound v2 Comptroller | 0x3d98…cd3b |
| Maker Vat | 0x35d1…492b |
| Fluid Liquidity | 0x52aa…e497 |
| Instadapp Lite ETH vault | 0xa0d3…be78 |
| ether.fi Liquid ETH | 0xf0bb…416c |
Ask about any lender in plain English
Every figure above came from data anyone can query. The Bitquery MCP server puts it behind an AI assistant, so you can ask how much a lender has out on loan, who its biggest borrowers are, or how many liquidations hit it last week, without writing the query yourself.
This article is provided for informational and educational purposes only and reflects analysis of publicly available on-chain data as of the dates indicated. It does not constitute legal, financial, compliance, or investment advice, and nothing in it is a recommendation to buy, sell, or hold any token or asset, or to lend to or borrow from any protocol.
The findings describe lending positions on Ethereum rebuilt from on-chain events and checked against contract state at month ends. Real debt caps each position at the market value of its collateral, using protocol price feeds where they track the market and traded prices where they do not; other price sources and other definitions of a loop would give different figures. Figures for smaller lenders use each protocol's own recorded totals and are approximate.
References to Aave, Compound, Maker, Sky, Spark, Morpho, Maple, Fluid, Euler, Liquity, Abracadabra, Ethena, Pendle, Lido, ether.fi, Instadapp, Terra and any other party describe contract state and on-chain activity. They are not statements about the intent, conduct or solvency of any company or person. Wallets are identified by address only.
Nothing herein should be relied upon as a definitive determination of fact. Readers should conduct their own independent verification before taking any action. The authors and publisher accept no liability for any loss or damage arising from reliance on this material. All trademarks and company names are the property of their respective owners.