Anyone borrowing USDe in order to turn it into sUSDe has, since September 9, 2026, been paying more for the loan than the deposit yields. On the four Aave markets that can be measured from ou
Anyone borrowing USDe in order to turn it into sUSDe has, since September 9, 2026, been paying more for the loan than the deposit yields. On the four Aave markets that can be measured from outside, the borrow rate for USDe on this September 12 stands between 6.02 and 6.39 percent, while the yield on staked USDe is 5.01 percent. The difference is negative, which means every additional round of the leveraged loop worsens the result compared with simply holding. This piece shows where the two figures stand, how to recalculate them yourself, and at what point the arithmetic tips over.
The risk service provider LlamaRisk published a recommendation in Aave's governance forum on September 9, 2026 at 16:39 UTC that raises the base rate for USDe. It reads: "Increase the USDe base rate by 1% to 6% and decrease Slope1 by 1% on Core, Plasma, Monad, Mantle, and Avalanche." The base rate therefore rises from 5.00 to 6.00 percent, while the first slope factor falls by one percentage point. All five Aave V3 instances that list USDe as a reserve are affected. The full recommendation, including the underlying figures, is in Aave's governance forum.
The reasoning appears in the same post: the model "prices USDe borrowing in line with Ethena's native staking rate". That is exactly why the loop tips over as soon as the two rates touch. The rates LlamaRisk expected were 6.36 percent on Core, 6.53 on Plasma, 6.38 on Monad, 6.21 on Mantle and 6.87 on Avalanche. By its own account, those five reserves carry around $323.8 million in USDe debt against some $1.18 billion in supplied liquidity.
One point belongs with this, because it changes the implications: the change did not go through a vote of token holders but through the so-called risk steward route. The post says so openly: "We will move forward and implement these updates via the Risk Steward process." There was therefore no effective date you could have marked in advance. LlamaRisk discloses in the same text that it is partly funded by the Aave DAO.
Yield loop means this: you supply sUSDe as collateral to a lending protocol, borrow USDe against it, stake that USDe back into sUSDe and supply it again. Every round enlarges the deposit without you paying in any new capital of your own. The profit comes solely from the difference between what the deposit yields and what the loan costs.
Everything therefore hangs on two figures that move independently of one another. One is the yield on sUSDe, the interest-bearing wrapper around the synthetic dollar USDe. Where that yield comes from, and what role Germany's BaFin played in it, we took apart in Ethena USDe: where the yield comes from. The other is the variable borrow rate Aave charges for USDe. Various lending protocols and their terms can be found in the comparison of crypto lending platforms.
Important for understanding this: both sides are dollar-denominated assets. The loop is no bet on a rising price. The only bet is that an interest rate difference stays positive.
APY refers to the return on a deposit annualised and including compounding. Borrow APR is the annual rate a borrower pays, calculated without compounding. At Aave this rate is variable: it changes with every deposit and withdrawal in the same market.
The yield on sUSDe is set anew each week and fluctuates with the returns Ethena earns from its delta-neutral position. The protocol's documentation describes this mechanism in its section on staking USDe. For you that means one figure moves weekly and the other by the minute. A result you measure today is no promise for next week.
For this article we queried both sides on September 12, 2026 at around 00:50 UTC through DefiLlama's open data interfaces. The yield on sUSDe was 5.01 percent, measured on Ethena's pool on Ethereum with a supplied value of $1.31 billion. The borrow rates for USDe on Aave V3 stood as follows:
The difference from the sUSDe yield therefore comes to minus 1.19 percentage points on Ethereum, minus 1.38 on Plasma, minus 1.14 on Monad and minus 1.01 on Mantle. On each of these four markets, borrowing costs more than the deposit brings in. The fifth market, Avalanche, is not listed by this data source and is missing from our measurement; at LlamaRisk it stood highest at 6.87 percent.
Every round of the loop wraps more tightly around the same deposit. That is precisely why a small interest rate difference ends up hitting hard. Aave derives the borrow rate from three quantities. The base rate is the floor that applies even in an empty market. Slope 1 is the premium added linearly up to the targeted utilisation. The optimal utilisation is the point from which a distinctly steeper second premium takes hold, so that liquidity for withdrawals always remains. As long as the market sits below that point, the following applies:
Borrow rate = base rate + slope 1 × (utilisation ÷ optimal utilisation)
Insert the values for the Ethereum market: base rate 6.00 percent, slope 1 after the change 1.00 percent, optimal utilisation 90 percent, actual utilisation 17.8 percent. That gives 6.00 + 1.00 × (17.8 ÷ 90) = 6.198 percent. We measured 6.20. For Plasma, with slope 1 of 2.00 and optimal utilisation of 85 percent, 6.00 + 2.00 × (16.8 ÷ 85) = 6.395 comes out, and we measured 6.39. Monad and Mantle work out the same way.
This test is more than arithmetic for its own sake. The reconciliation shows that the recommendation of September 9 has in fact been implemented: a base rate of 6.00 percent is the only figure with which the four measured rates add up. On top of that, the formula lets you predict the rate for any future utilisation without waiting for an announcement.
Carry is the running difference between the return on a position and its financing costs. If it is negative, the position loses money as long as nothing else happens. Let us work that through with the measured values for the Ethereum market, so 5.01 percent return and 6.20 percent cost, each before fees and transaction costs.
Without leverage. You hold one unit of sUSDe and borrow nothing. Result: 5.01 percent a year.
With an 80 percent loan-to-value limit. One unit of equity becomes five units of deposit and four units of debt. Result: 5 × 5.01 − 4 × 6.20 = 25.05 − 24.80 = 0.25 percent. A good five percent has turned into a quarter of a percent, at five times the exposure.
With a 90 percent loan-to-value limit. Ten units of deposit, nine units of debt. Result: 10 × 5.01 − 9 × 6.20 = 50.10 − 55.80 = minus 5.70 percent a year. The position now costs you money even though no price has fallen.
One side condition carries this calculation: the sUSDe supplied at Aave earns no additional interest there. The data source shows a supply rate of zero for the sUSDe reserves on Aave. There is therefore no third source of return that would absorb the difference.
The general relationship can be stated in one sentence: as long as the difference between deposit yield and borrow rate is negative, every borrowed unit subtracts from the result. The formula behind it reads result = yield + borrowed units × (yield − borrow rate). With a negative difference the second term is always a deduction, and it grows with the leverage.
From this follows a threshold worth knowing: at minus 1.19 percentage points the result has fallen to zero at around 4.2 borrowed units, which corresponds to a loan-to-value of just under 81 percent. Above that it turns negative. But even below that threshold the loop sits under what simple holding yields. The threshold therefore marks only the point at which the loop slips into the red. It has already stopped being worthwhile against simple holding before that.
USDe is a synthetic dollar. Behind the token there are no bank deposits as with USDC; Ethena instead holds a spot asset and simultaneously sells perpetual futures in the same amount, meaning futures contracts with no expiry. This construction is called a basis trade, and its outcome is delta-neutral: if the price of the spot asset rises or falls, the gain and loss on the two sides largely offset one another.
The return comes from the funding rate, meaning the payment holders of long positions make to holders of short positions as long as the futures price sits above the spot price. If that payment turns negative, the yield shrinks. That is exactly why the rate moves: over the past two weeks it stood, according to the same data source, at 4.47 percent on September 4, 4.68 on September 8, 4.85 on September 11 and 5.01 percent on September 12.
For the loop that means the difference can close again from the yield side without Aave changing anything. The gap can also widen further. Anyone running a loop checks both figures regularly rather than once. Suitable tools for that are set out in the comparison of analytics platforms.
Utilisation is the share of supplied liquidity that is currently lent out. That makes it the only quantity in the rate formula that changes continuously, and the best leading indicator for your future rate. On Ethereum, September 12 saw 17.8 percent against an optimal utilisation of 90 percent. That is a long way from the steep zone.
The reverse holds too: if a large supplier withdraws its deposit, utilisation jumps and your rate rises with it. Above the optimal point the second, distinctly steeper premium takes hold. How steep is shown by an example from the same recommendation for a different market: there, a borrow rate of 24.50 percent applies at full utilisation after the second premium was halved; before that it would have been 44.50 percent.
At low utilisation, liquidity flows freely. The tighter the market gets, the more expensive each further borrowed unit becomes. The question cannot be answered with a single figure, but it can be broken into three parts. The first is smart contract risk: the protocol's code manages the collateral itself, and a flaw in it hits all positions at once. The second is oracle risk: Aave needs a price for sUSDe against USDe, and that price comes from outside. If it diverges from the market price, a healthy position can become unhealthy on paper.
The third part is the one the change of September 9 made visible. Parameters such as the base rate, slope factors and caps can be adjusted through the risk steward route without a vote, within set limits. That is intended as a safety mechanism, because a vote takes days and a risk does not wait. For you as a borrower it means the basis of your calculation can change while you sleep. The price of the protocol token we treat separately in the Aave price prediction.
A liquidation is the forced sale of your collateral by the protocol as soon as the value of the debt relative to the collateral crosses a threshold. Because both sides of the loop are denominated in dollars, it looks immune to price moves at first glance. In normal operation it largely is, and sUSDe even gains slowly in value against USDe, because the returns flow into it.
The construction is vulnerable elsewhere. If USDe's peg to the dollar loosens even briefly, or the oracle revalues sUSDe, the ratio shifts abruptly. How little room such interest rate difference positions usually carry is shown by an observation from the same LlamaRisk analysis of a different market: there, the median health factor of the largest suppliers was 1.03, so a good three percent above the liquidation threshold. That figure expressly applies to a different market and not to USDe, but it shows the order of magnitude in which such positions are run.
Three days lie between the recommendation and our measurement, and the figures reveal how the market has reacted. LlamaRisk put USDe debt across all five markets at around $323.8 million. On September 12 we measure $193.1 million across the four accessible markets combined. Part of the gap goes back to the missing Avalanche market, the rest to repayments.
At the same time supply has grown. The supplied value in the Ethereum market rose from $233.8 million on August 30 to $525.4 million on September 12. More supply with less demand means falling utilisation, and falling utilisation dampens the rate again. That is why the measured 6.20 percent sits below the 6.36 percent LlamaRisk expected. The floor of six percent is untouched by this; the rate can no longer fall below it, however empty the market becomes.
An independent analysis by CryptoSlate on September 11 reached the same finding and put the effect at 13 to 89 basis points. There, an sUSDe rate of 4.72 percent as of September 10 was used as the yield side, a lower figure than the one we measured on September 12. The direction agrees in both measurements; the size of the gap differs.
A yield loop generates more tax-relevant events than it looks, and the classification is contested in individual cases. Taking out a loan is as a rule not a disposal in itself. Swapping USDe into sUSDe and back, by contrast, can count as a disposal, and every round of the loop contains at least one such swap. The running returns from staking are income whose allocation depends on the specific structure.
Three questions therefore belong before the first round rather than after it: at what point in time and at what price is each swap valued? Under which category of income are the staking returns treated? And how do you document a chain of dozens of individual events so that it remains traceable? For the last point the tools from the comparison of tax and portfolio tools help. The first two are answered reliably only by a tax adviser who knows your case. This section is expressly not tax advice.
The measured values are a snapshot of September 12, 2026 and no forecast. Neither Ethena's staking rate nor Aave's borrow rate is guaranteed; both can move in either direction, and the difference may be positive again tomorrow. Fees, network costs and price slippage when swapping are contained in none of the three calculations, so they worsen the result further.
Also not included are incentive programmes. Protocols temporarily reward certain positions with their own tokens, and such a reward can offset a negative interest rate difference on paper for as long as it runs. Anyone counting on that should know that its value fluctuates and its term ends. Ethena's own earnings position we looked at separately in the fee switch and the ENA buyback.
(As of September 12, 2026. This article is not investment advice. Prices and fee structures change; check the terms with the provider before you buy.)
