Half of Aave’s debt sits in 9% of positions on one ETH trade

Galaxy’s analysis of Aave V3 Core found that a small group of borrowers carries an outsized share of the lending market’s risk. Fewer than 9% of loans hold about half of all outstanding debt, and that debt is concentrated in one Ethereum position: WETH borrowed against liquid-staking and restaking tokens. The cohort’s danger does not come from a fall in ETH’s price. It comes from those wrapper tokens trading at a discount to the ETH they represent.

The numbers behind the split

Galaxy’s Aug. 7 snapshot covered 19,073 Aave V3 Core loans after applying standard filters. The firm split them into two groups: borrowers using Aave’s E-mode setting and everyone else.

Borrower group	Share of positions	Share of debt	Debt-weighted LTV	Avg. health factor	Debt-to-equity
E-mode borrowers	8.91%	~50%	~90%	~1.06	~10.7x
Other analyzed borrowers	~91%	~50%	~49%	~1.79	~1.07x

The E-mode minority borrows close to nine-tenths of its collateral value at roughly ten times the debt-to-equity of the rest, with an average health factor only slightly above 1. The majority borrows about half its collateral value and sits well above the liquidation line.

What the E-mode cohort is holding

On the collateral side, Ethereum staking and restaking wrappers make up about 66.2% of what backs the E-mode loans. weETH alone accounts for roughly 42%, with rsETH and wstETH among the other wrappers. On the debt side, WETH accounts for about 73% of what the group owes.

Galaxy describes this as a concentrated bet on Ethereum’s staking basis, meaning the relationship between liquid-staking tokens and plain ETH. A borrower deposits a yield-bearing wrapper, borrows WETH, and typically converts that WETH back into more wrapper exposure. The trade earns the gap between staking or restaking yield and the WETH borrowing rate, amplified by leverage. The profit depends on that spread. The risk depends on the wrapper continuing to trade in line with ETH.

Why E-mode allows a 90% loan-to-value

E-mode, short for efficiency mode, lets Aave set higher borrowing limits when collateral and debt are expected to move together. Lending 90% against an unrelated asset would be reckless, because a modest price drop would leave the loan undercollateralized. If the collateral and the debt track each other closely, though, the same high LTV carries roughly the risk of a conservative loan between two independent assets.

Here is how that works for a weETH and WETH position. If ETH falls 10%, weETH falls about 10% with it, and so does the WETH debt. Both sides of the balance sheet shrink together and the health factor barely moves. A borrower in this setup can withstand a sharp ETH sell-off without coming close to liquidation.

What the position cannot absorb is a change in the exchange rate between the wrapper and ETH. If weETH, rsETH or wstETH starts trading below the ETH it represents while the debt stays fixed in WETH terms, the collateral loses value relative to what is owed. That can happen even if ETH’s price stays flat. Wrappers can drift from their underlying value when holders try to exit faster than the token’s redemption path allows, or when the market prices in risks specific to a staking or restaking protocol.

How large a discount it would take

Aave calculates a position’s health factor by multiplying collateral value by a weighted liquidation threshold and dividing the result by total borrowed value. A position becomes eligible for liquidation when that number falls below 1.

The cushion in the cohort’s average follows directly from the formula:

Starting health factor: 1.06
Collateral decline that brings it to 1: 1 − (1 / 1.06) ≈ 5.7%, assuming debt value is unchanged

The 5.7% applies to the whole collateral basket, and only about 66.2% of it is wrappers. Under a rough proportional model, where the remaining collateral holds its value against the debt, the wrapper discount needed to consume the cushion is about 5.7% ÷ 0.662 ≈ 8.6%. That is where the 8% to 9% estimate comes from.

The figure describes an average position, not a real account. Actual liquidations would depend on each account’s collateral mix, its specific liquidation thresholds, how Aave’s oracles price the wrappers, and what the debt consists of. Accounts below the cohort average would hit the line with a smaller discount.

Galaxy’s earlier May analysis gives a sense of the tail. It modeled a 10% weETH depeg and found Aave would be left with about $2.47 billion of debt against $2.42 billion of post-shock collateral, with 205 accounts pushed below a health factor of 1. The same work showed stress rising sharply once the depeg moved from 3% to 5%, not increasing gradually.

Wrapper/ETH discount	Approx. effect on E-mode risk	What it means
0% to 2%	Correlation mostly holds	E-mode works as a capital-efficient trade, no liquidation event
3% to 5%	Weakest accounts become sensitive	Borrowers may need to add collateral or repay WETH
8% to 9%	Average E-mode health factor approaches 1	Liquidation risk spreads across the cohort
10%	May model: 205 accounts below a health factor of 1	Forced deleveraging hits the ETH staking loop

What happens when a position crosses 1

A widening wrapper discount pulls health factors down gradually, and borrowers watching that decline can respond in two ways: deposit more collateral or repay part of their WETH debt. Either one raises the health factor.

A borrower who does neither and drops below 1 can be liquidated. Aave liquidation is permissionless: any liquidator can repay part of the debt and receive the matching collateral plus a bonus. In this cohort, the collateral being seized and sold is mostly staking wrappers. Selling those wrappers can add to the same discount that triggered the liquidations, which is why concentrated positions with thin cushions get attention even while aggregate leverage is falling.

Shrinking leverage, concentrated leverage

Galaxy’s second-quarter report describes crypto lending as contracting in an orderly way. Total crypto-related debt fell 15.08% from the prior quarter to $73.2 billion, the third straight quarterly decline. The firm compared the slow descent to taking the stairs, in contrast to the single-quarter collapse of 2022 when centralized lenders failed and leverage disappeared all at once.

Aave’s own composition complicates that picture. In April, the split between E-mode and standard debt was closer to 60/40 in favor of E-mode. By Aug. 7 it had narrowed to about 50/50, entirely because E-mode debt declined. The riskiest segment is shrinking, but it still carries close to half of all platform debt while making up fewer than one in ten positions.

On Aave V3 Core, the data suggests leverage has not spread more thinly but gathered into a small number of highly leveraged ETH-basis loans that still carry systemic weight.

What to watch

For this cohort, the main signal is the exchange rate between the major wrappers and ETH, not ETH’s headline price. weETH matters most because it makes up the largest share of collateral. A discount moving from the low single digits into the 3% to 5% range is where Galaxy’s work showed stress building quickly. The high single digits are where the average position runs out of room. Reaching that point would require weETH, rsETH and wstETH to trade at a high single-digit discount at the same time. The weakest accounts would be liquidated first, and forced selling would land on leveraged staking positions rather than on ETH spot holders generally. The direction of E-mode debt matters too: continued decline lets the concentration unwind without a liquidation wave, while new borrowing would rebuild it.