Aave borrowers faced up to 36% APR while a freeze blocked fresh collateral

Borrowers in Aave's frozen reserves kept paying interest while the restrictions prevented fresh deposits, including the extra collateral they might normally supply to protect a loan. Suppliers were still able to request their money back, provided enough tokens were available, so withdrawals could make borrowing more expensive even though new borrowing was blocked in those reserves.

That left Aave with a rate mechanism whose deposit incentive couldn't work in frozen reserves. More expensive borrowing encourages people to repay and offers better returns to anyone willing to lend, but the freeze prevented that second group from putting money in. Cutting rates could give borrowers some relief, while doing nothing by itself to return the tokens suppliers wanted to withdraw.

The restrictions were introduced during the Oct. 29 Binance hot-wallet incident, and the affected reserves had reopened by Oct. 31.

LlamaRisk's recommendations to reduce borrowing rates temporarily, then restore their earlier settings, explain how protecting a lending market can leave its existing users needing different things from the same pool of money.

Someone using Aave can supply crypto as collateral and borrow another token against it, obtaining funds without selling the original holding. If you want dollars while keeping your Ethereum, for example, you can borrow a dollar-linked token such as USDC.

Those borrowed tokens come from other users who supplied them to Aave's lending contracts, the blockchain programs that handle the transactions. Suppliers earn interest while borrowers owe it. However, the same person can be both a supplier and a borrower with different assets. Supplying ETH as security for a USDC loan doesn't mean you've supplied the USDC that someone else wants to withdraw.

During the Binance hack, a wallet linked to exploiters had already supplied about $60 million of USDT and borrowed 19,762 WETH, according to LlamaRisk's account. The provider described the position as overcollateralized, meaning its collateral exceeded its debt, and said it was subject to normal liquidation. The restrictions were intended to limit further use of Aave during an external compromise.

Freezing the affected reserves stopped new supply and borrowing while repayments, withdrawals, and liquidations were still enabled. Therefore, existing debt continued to accumulate interest, but borrowers couldn't supply extra collateral into those frozen reserves.

Repayment could still reduce the debt, provided the borrower could obtain the necessary tokens. If you've borrowed USDC against ETH, repaying requires USDC, while adding more ETH would ordinarily let you support the debt without reducing it. Losing the second option narrows how you can manage the position, especially if the collateral's market price is falling at the same time.

Under Aave's liquidation rules, the health factor divides collateral value, adjusted by its liquidation threshold, by total debt. Below one, the position becomes eligible for liquidation, allowing another participant to repay debt and receive collateral plus a bonus. Interest adds to the debt side of that calculation even if the collateral price doesn't move.

However, suppliers depended on a different balance: how much of the token they wanted was actually available to withdraw. In a simplified pool with 100 USDC supplied and 90 USDC borrowed, only 10 USDC is immediately unborrowed, even if every loan has ample collateral behind it. Collateral supporting those loans doesn't turn into additional USDC just because a supplier wants to leave.

If suppliers withdraw five of those available tokens while the debt stays at 90 USDC, the pool now has 95 USDC supplied and only five available. The borrowed share has increased from 90% to about 94.7%, despite nobody taking out another loan. This is what utilization measures, and it explains how withdrawals can make a frozen market more expensive for borrowers who haven't touched their accounts.

Aave's rate formula charges more as utilization increases, with a stronger response beyond a configured threshold. Its upper section is controlled by a parameter called Slope2. Normally, higher borrowing costs encourage repayment while better supplier returns can attract fresh funds, and both can replenish the tokens available for withdrawal. During the freeze, new deposits were prohibited, leaving repayment as the available channel among those two responses.

That's why LlamaRisk recommended reducing that upper rate response. Its rationale attributed utilization of 92% to 100% in major stablecoin reserves to supplier withdrawals during the freeze, with annual borrowing rates of 10% to 36% under the settings then in place. Those were annualized rates, so a few hours at 36% didn't add 36% to someone's debt.

To put the scale in dollars, a hypothetical $10,000 loan at a constant 36% annual percentage rate would accrue roughly $9.86 in one day, versus about $2.74 at 10%, ignoring compounding and fees.