36% and Locked Out: Inside Aave's One-Way Reserve Freeze

Thirty-six percent. That's the annualized borrow rate some Aave users were paying on positions they weren't allowed to defend.

Not 36% on a trade someone entered with full information and a clear exit. Thirty-six percent on a loan sitting in a frozen reserve, where the one person with the strongest incentive to fix the position, the borrower, couldn't add a single dollar of collateral.

Let me break this down, because the mechanics matter more than the headline number.

The Freeze

A reserve gets frozen. In this case the trigger was fallout from a Binance-linked exploit, and the response was the standard playbook. New deposits stop. New borrowing stops in those markets. Interest keeps accruing, because interest always keeps accruing.

Here's where it gets ugly. Suppliers could still request their money back, provided enough tokens were sitting in the pool. So they did. Liquidity drained. Utilization climbed. And on Aave's rate model, once utilization pushes past the kink around 90%, the borrow rate doesn't rise, it launches.

That's the 36%. A number generated by a market where only one side was allowed to act.

Run the math on a $100,000 position. At 36% you're paying roughly $3,000 a month for the privilege. Call it $98 a day. Except you can't call it anything, because you can't repay early, you can't top up, you can't do the two things any risk manager in the world would tell you to do first.

Freezing deposits isn't unreasonable on its face. When there's an exploit somewhere in the stack, you stop the bleeding and you stop inflows of tokens whose provenance you can't verify. Fine. But freezing deposits and letting withdrawals run freely isn't a freeze at all. It's a slow squeeze aimed at whoever's left holding the bag.

So the supplier who panics first gets out whole. The borrower who wanted to deleverage gets a bill that compounds. Same protocol, same block, completely different outcomes.

The Asymmetric Door

What the street is missing: this wasn't a governance failure or a malicious actor. Every individual action was rational. Suppliers withdrew because withdrawing was permitted and safer. The protocol locked deposits because locking deposits reduced exposure. The rate model did exactly what it was coded to do when utilization spiked.

Rational inputs, terrible output. That's the pattern crypto keeps rediscovering and keeps forgetting.

From a risk perspective, a one-sided freeze is worse than a full stop. If you're going to freeze a market, freeze both directions. Shut down the exit alongside the entrance and everyone absorbs the same pain on the same timeline. Allow only the entrance and you've handed early movers a free option priced against everyone still inside.

And yes, this is a bank run wearing different clothes. The whole point of deposit insurance and lender-of-last-resort facilities in TradFi is to break the loop where individually sensible withdrawals become collectively catastrophic. DeFi has no such backstop. It has a governance vote and a multisig, and neither one shows up at 3 a.m. when utilization is at 95%.

Here's the part that should bother you more. Borrow rates in DeFi get marketed as a price signal. A 36% rate that no borrower can respond to isn't a price. It's a penalty. What does a rate even mean when half the participants are administratively locked out of reacting to it? Nothing useful. It stops being information and becomes a mechanism for transferring value from people who can't move to people who already did.

Who wins here? Suppliers who exited early, obviously. Liquidators, if any positions tipped over, and you can bet some did at those rates. Anyone sitting outside the frozen pool with dry powder and a fresh appetite for cheap collateral. Who loses? Borrowers who did nothing wrong except hold a position in the wrong reserve on the wrong day. And honestly, the protocol's own credibility, because "we've risk managers" is a weaker pitch when the risk response amounts to locking the front door and leaving the back one open.

One more thing worth saying plainly. The risk frameworks on these markets are good. Gauntlet and Chaos Labs and the rest do real work with real modeling. But a model that recommends blocking deposits while permits withdrawals is optimizing for a number on a dashboard, not for the people with skin in the game. Risk management that protects the protocol by immiserating its borrowers is a short-term trade with a long-term cost.

What It Costs You

The fix is boring and cheap. During a freeze, allow repayments and allow collateral top-ups. Both of those reduce protocol risk. Both let borrowers defend themselves. Neither opens the door to whatever token the exploit touched, because you're not accepting new supply, you're letting people post assets you already whitelisted to protect loans you already booked.

Blocking the entire deposit side because you can't be bothered to separate two use cases is lazy design dressed up as caution.

So here's the takeaway, and it applies to anyone borrowing in a pooled market, not just Aave. Your collateral is only half your risk. The other half is the protocol's freeze policy, which you probably never read, and which will decide whether you get to act when things go sideways. Before your next borrow, go find the reserve freeze conditions. Ask what happens to your position if deposits stop and withdrawals don't. Ask whether the rate model has a kink you'll never be allowed to respond to.

Most people won't check. That's why the same lesson keeps costing someone 36% a year. Get it wrong once and you're not paying for a loan anymore, you're paying for a lesson.