InvestigationEthereumEthenaPendle

The $7 billion Ethena loop, and how it unwound

For a year and a half, the most popular trade in DeFi lending was a loop on Ethena's dollar tokens. We rebuilt every one on Ethereum, month by month, to find out what paid for it and what ended it.

At a glance
USDe is Ethena's dollar token, and sUSDe is the staked version that pays a yield. The loop was simple: post one of them, or a Pendle token built on them, borrow stablecoins against it and buy more. On Ethereum's lenders the loop grew from $125 million in March 2024 to $7.0 billion in August 2025, most of it on Aave. A year later $1.0 billion was left. The yield fell toward the cost of borrowing, and Pendle's payout dates gave people a moment to leave. One week in April 2026 took most of what remained.
$7.0B
USDe-based loops on Ethereum at the peak, August 31, 2025
$1.0B
Left a year later, August 31, 2026
86%
Share of the peak that sat on Aave
$38M
Loans backed by USDe tokens ever liquidated, all lenders
Key findingsEach line has its own share link
  1. $7.0B

    Of USDe-based collateral was backing loops on August 31, 2025. That was 81% of every dollar loop on Ethereum's lenders.

  2. $3.5B

    Left within a month of the Pendle payout on September 25, 2025. 915 wallets closed out.

  3. 12.8%

    Paid by the looped Pendle tokens in August 2025, against 6.9% to borrow. By October the gap was about one point.

  4. $0.982

    Aave's price for USDe during a dip on February 21, 2025. Loans worth $21.4 million were liquidated. Two weeks later Aave priced USDe off USDT.

  5. 15%

    Cost to borrow USDT on Aave on April 19, 2026, up from 3.5%, after stablecoin lenders pulled $2.5 billion in two days.

  6. $1,263

    Bad debt booked on loans backed by USDe tokens across the whole period. Forced sales came to $38 million.

01 · The tradeHow the Ethena loop works

At the end of August 2025, $7 billion sat in one trade on Ethereum's lenders. People posted Ethena's dollar tokens, borrowed stablecoins against them and bought more, round after round. A year later $1 billion was left. Almost none of it had been liquidated on the way down.

USDe is the dollar token from Ethena. Ethena keeps it near $1 by holding crypto and shorting the same amount on perps, and the income from that hedge goes to people who stake. Staked USDe is called sUSDe. It is worth a little more USDe every day.

Pendle splits a yield token like sUSDe in two. Its principal token, or PT, trades on-chain below $1 and pays a full dollar on a set date. Buying one locks in a fixed yield.

The loop is simple. You post sUSDe, USDe or a PT on a lender such as Aave, borrow USDT or USDC against it, buy more of the same and post that too. Each round adds more yield and more debt. It pays as long as the collateral earns more than the loan costs, and every extra round multiplies the gap, whichever way it goes.

Plain USDe earns nothing by itself. People looped it anyway, because Ethena also handed out points, called Sats, to USDe holders. Points never show up on-chain, so our numbers leave them out.

02 · The riseFrom $125 million to $7 billion

The Ethena loop on Ethereum, every month endUSDe-based collateral backing dollar loops, all lenders, in billions of dollarsPendle tokenssUSDeUSDeOther USDe tokensJul 2024Jan 2025Jul 2025Jan 2026Jul 2026$0B$2B$4B$6B$8BAug 2025: $7.0BAug 2026: $1.0BDec 2024: $1.7B

It started small. In March 2024 the loop held $125 million, all of it on Morpho. By December it had grown more than tenfold, and Aave held most of it for the first time.

It shrank in early 2025, then took off once Aave listed its first Pendle token on USDe at the end of April. From May on, Aave usually held more than 90% of the loop, and 86% at the August 2025 peak of $7.0 billion.

By then, about four in five dollars looped on Ethereum's lenders were built on USDe, and two thirds of the loop was PTs. Two of those PTs, both paying out on September 25, held $4.6 billion between them.

03 · What paid for itYield in, interest out

What the loop earned, and what it costMonthly averages, percent per yearPendle token yieldsUSDe yieldWhat loopers paid to borrowJul 2024Jan 2025Jul 2025Jan 2026Jul 20260%15%30%45%4.2%3.9%

The loop only pays while the collateral earns more than the loan costs. In its first months it did not. In March 2024, sUSDe paid about 30% per year and loopers paid more than 40% to borrow. The points were most likely the draw.

From the summer of 2024 the numbers worked. At the peak, the looped Pendle tokens paid almost twice what the loans cost. Then the gap closed. By October it was down to about one point, and through 2026 the yield and the cost both stayed close to 4%. One point of profit, looped ten times, is still thin. A single jump in borrowing rates wipes it out.

04 · Payout daysWhen the Pendle tokens came due

Thin margins made leaving easy, and Pendle's calendar made it natural. A PT has an end date. On that day a looper has to close the position or roll into a newer PT, so payout days are when money leaves.

Payout dateWalletsDueKept
Oct 24, 202462$92M54%
Dec 26, 202490$264M63%
Feb 27, 202543$135M40%
Mar 27, 2025202$480M50%
May 29, 2025347$465M60%
Jul 31, 2025488$1.4B67%
Aug 14, 2025280$735M21%
Sep 25, 20251,100$4.6B25%
Oct 16, 2025125$104M24%
Nov 27, 2025266$1.5B24%
Jan 15, 202687$57M38%
Feb 5, 202694$436M29%
Apr 2, 2026108$156M35%
May 7, 202642$124M66%
Pendle tokens that backed loops, by the date they paid out. Wallets and Due: at the month end before. Kept: those wallets' USDe loops a month after the maturity, as a share of what they had before. Closed out: kept less than 5%.

Until July 2025, loopers kept two fifths to two thirds of their positions going after a payout. From August to November they kept only a fifth to a quarter.

The big one came on September 25. That day the two Pendle tokens at the top of the loop came due, and a month later four in five of the wallets holding them had closed out, taking $3.5 billion with them. Most of the money that stayed moved into the next Pendle tokens, due in November.

05 · The runOne week in April

The last big exit had nothing to do with Ethena. On April 18, one wallet borrowed tens of thousands of WETH from Aave against rsETH, a restaked ETH token, that had left the rsETH bridge minutes earlier. Aave later wrote off more than $120 million on that position, as our data feature on Ethereum lending explains.

Within two hours, depositors began pulling stablecoins out of Aave, almost $1 billion in a single hour. Over two days about $2.5 billion of USDT and USDC left, net, and with that much less cash to lend, the cost to borrow both jumped more than fourfold, to 15% per year.

April 2026 on Aave: the run and the unwindBars: USDe and sUSDe collateral withdrawn each day, millions of dollars. Line: cost to borrow USDT, percent per year.USDe and sUSDe withdrawnUSDT borrowing cost$0M$200M$400M$600M0%4%8%12%16%Apr 1Apr 8Apr 15Apr 18Apr 22Apr 29Apr 18: rsETH-backed loan, then the run$537M

For a loop earning about 4%, that meant losing money on every borrowed dollar. Loopers repaid, and $1.8 billion of USDe and sUSDe collateral left Aave within a week. By the end of April the loop was less than a third of its March size. Very little of that was forced. Liquidations on USDe tokens that month came to $213,000.

06 · Who ran itThe fifty biggest wallets

Month endOver $1KOver $100KTop 10
Dec 202464740748.2%
Aug 20252,4421,26322.1%
Aug 202632412970.6%
Wallets with a USDe-based loop above each size. Top 10: the ten biggest wallets' share of the loop.

At the peak, the fifty biggest wallets held just over half the loop, and nearly all of them were plain wallets. No pooled vault made the list. That is unusual. In lending as a whole, half of the ten biggest borrowers in August 2026 were vaults. Our labels name almost none of these wallets.

TypeCountShare
Plain wallets4448.1%
Smart wallets44.0%
Multisigs21.4%
Top 505053.4%
The 50 biggest looping wallets on Aug 31, 2025, grouped by wallet type. Share: of all USDe-based loops that day. Smart wallets: ordinary wallets that have switched on smart-account code.

07 · The damageWhy so few loopers were liquidated

Across the whole period, lenders liquidated $38 million of loans backed by USDe tokens, two thirds of it on Aave. Bad debt booked on them came to $1,263. For a $7 billion trade that is close to nothing, and part of the reason is how Aave prices USDe.

Loops run close to the edge, so a small dip in the collateral's price can set off a liquidation. Until March 2025, Aave priced USDe with a market feed for USDe itself. On the afternoon of February 21, 2025, that feed dipped below 99 cents. Within five minutes, five sUSDe-backed loans worth $21.4 million were liquidated on Aave. Two weeks later Aave moved both tokens to a USDT price source, and it has priced USDe as if it were USDT ever since, however the market for USDe itself moves.

Time (UTC)Price sourceUSDe
Feb 21, 2025, 13:44Capped USDe / USD$0.9999
Feb 21, 2025, 15:44Capped USDe / USD$0.9824
Oct 10, 2025, day endCapped USDT/USD$1.0031
The price Aave's oracle gave USDe, read from the chain at each moment. On March 8, 2025, Aave moved USDe to a USDT price source.

The test came that October, on a day of heavy liquidations across crypto. Aave's price for USDe stayed at about $1. Aave liquidated over $100 million of loans that day, and almost none of it was backed by USDe tokens.

Few liquidations does not mean nobody lost. Anyone who stayed in while borrowing cost more than the yield lost money slowly, and we cannot see what the points were worth to them in the end.

08 · What ended itA trade that stopped paying

So what ended the Ethena loop? The yield that paid for it shrank until it barely beat the cost of the loans. Pendle's payout days then gave people a clean moment to walk, and April's run on Aave pushed out most of what was left. At every step people chose to go. Almost nobody was forced.

It still left a mark. In the year after the peak, loops of every kind on Ethereum shrank by almost $15 billion, as our look at Ethereum lending since 2021 shows. The Ethena loop was about two fifths of that, and most of it had sat on Aave.

09 · How we measuredWhat we counted, and what we left out

We rebuilt every loan on Ethereum's main lenders, month by month, from their on-chain events. A loop here is a stablecoin loan backed by USDe, sUSDe, a Pendle token built on them or a small wrapper of USDe. Loans in dead markets are left out.

Pendle yields come from each market's own on-chain rate, and borrowing costs are what loopers actually paid, averaged across lenders by how much they owed. We did not count the value of Ethena's points, or loops on other chains.

10 · The recordThe contracts behind the numbers

ContractAddress
Aave v3 Pool0x8787…a4e2
Aave Oracle0x5458…a0c2
USDe0x4c9e…68b3
sUSDe0x9d39…3497
USDe feed, to Mar 20250x55b6…cc17
USDe feed, from Mar 20250xc26d…4ff8
sUSDe feed, to Mar 20250xb37a…665a
sUSDe feed, from Mar 20250x42bc…86ec
Pendle PT-USDe Sep 20250x6d98…6af4
Pendle PT-sUSDe Sep 20250xa36b…a9f7
Pendle PT-sUSDe Nov 20250xb6ac…dbdd
rsETH-backed borrower0x1f4c…adef
Run it yourself

Ask about any loop 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 what a wallet has borrowed on Aave or Morpho, what a Pendle market pays today, or who pulled stablecoins out of a lender this week, without writing the query yourself.

Pull a wallet's loans and collateral on any lenderCheck a Pendle market's fixed yield over timeWatch deposits and withdrawals on Aave by the hourCompare a lender's price feed with the market price
Explore Bitquery MCP Figures measured September 23, 2026, against Bitquery's Ethereum data, month ends from March 2024 to August 2026. Written by Bitquery Research; AI tools ran the queries and drafted the text, and every figure was worked out again from the data before publishing.
Scope, limits and attribution

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 it is not a recommendation to buy, sell, hold or borrow against any asset.

The findings describe lending positions on Ethereum rebuilt from on-chain events and checked against contract state. Loop sizes are month-end snapshots; borrowing costs and yields are monthly averages. Ethena's points and anything held off-chain are not measured.

References to Ethena, Pendle, Aave, Morpho, Euler, Spark and any other party describe contract state and on-chain activity. They are not statements about the conduct or intent of any person or company. Wallets are described by the type of contract they are, not by who controls them.

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 the use of this article.