In one weekend this past April, about $6.6 billion left Aave. Coindesk reported that the protocol’s total value locked fell from $26.4 billion to roughly $20 billion in less than a day. Nobody breached Aave’s own contracts. Attackers drained a different protocol, Kelp, then dumped the stolen tokens on Aave as collateral and borrowed against them.
I bring this up first because most explainers about Aave lead with the good news: biggest lender, deepest liquidity, years of uptime. All of that is true. But if you want to actually understand why capital keeps flowing back into Aave v3, you also need to understand the outflow. Crypto lending is just one of several ways to put idle crypto to work, and it happens to be the one with the deepest liquidity in the space.
So, this piece covers how the protocol works under the hood, what flash loans and risk tools actually do, and an honest look at why more than $13 billion still calls Aave home.
Key Takeaways
- Aave v3 is a non-custodial lending protocol built on smart contracts, not a company holding your funds. Aave V3 alone carries about $13.7 billion in TVL across 20+ chains.
- Suppliers earn interest by depositing into shared liquidity pools and receive aTokens that grow in their wallet automatically.
- Flash loans let anyone borrow without collateral, provided they repay within the same transaction, for a 0.05% fee.
- Isolation mode, eMode, and the health factor system exist specifically to contain the kind of risk that surfaced in April 2026.
What Is Aave v3?
Aave v3 is a non-custodial, on-chain lending protocol. You supply crypto, you earn interest, or you borrow against what you’ve supplied, all without a bank, broker, or company standing between you and the money. Nobody at Aave approves your loan. A smart contract does. And your loan gets approved by a smart contract, not someone working at Aave.

Aave itself traces back to 2017, when it launched under a different name before rebranding. Aave v3 was dropped in 2022 with a bunch of added features that matter most for this guide. Like the isolation mode for risky assets, eMode for capital efficiency, and cross-chain deployment. Today it runs on more than 20 chains, with Ethereum still holding the largest share of deposits at roughly 82%.
If you’ve used v2 before and you want to know what Aave v3 does, then this is the most convenient part. It does everything that v2 does, plus it isolates new or volatile assets so one bad listing can’t threaten the whole pool.
How Does Aave v3 Work?
Every asset Aave supports has its own pool, and everyone who supplies that asset drops into the same pot. Borrowers draw from the pot, not from any individual person’s deposit.
Interest rates move with how much of the pool is being borrowed. MixBytes’ breakdown of the Aave v3 architecture walks through this utilization model in detail. Like when a pool sits mostly idle, borrowing is cheap and supplying pays little. But when a pool gets heavily borrowed, rates climb on both sides at once, pulling in more supply while cooling off demand.
Liquidity pools and aTokens
When you supply an asset, Aave mints you an equivalent amount of aTokens, like aUSDC for USDC. These sit in your wallet and grow on their own as borrowers pay interest into the pool. Redeem your aTokens at any time, and you get back your principal plus whatever interest accrued, with no separate claim process required.

The smart contracts that run the protocol
Every supply, borrow, repayment, and liquidation on Aave v3 runs through open-source smart contracts, not a back-office team. Governance, run by AAVE token holders, sets the risk parameters; like which assets are listed, what the borrow caps are, and how aggressive liquidations get.
That governance layer is slower than a company making a unilateral call, but it means no single party can quietly change the rules on your position overnight.
Supplying and Borrowing: How DeFi Lending Works on Aave
Here’s what actually happens when you use Aave as a regular depositor, not a developer.
You supply an asset, for example ETH, and it becomes usable as collateral. Aave assigns every asset a loan-to-value (LTV) ratio, the maximum percentage of your collateral’s value you’re allowed to borrow against. Supply $10,000 of ETH at a 75% LTV, and you can borrow up to $7,500 in another asset, like USDC. As MixBytes notes, the same utilization curve that sets supply yields also drives what you’ll pay to borrow, so both sides shift together as the pool fills up.
That gap between what you supplied and what you borrowed isn’t arbitrary. It’s the buffer that protects the pool if ETH’s price drops before you repay.
Run the numbers on that same $10,000 ETH deposit. At a 75% LTV, your liquidation threshold would be a bit higher, maybe 80%, which is the point where your debt actually gets sold off rather than just capped. Borrow the full $7,500 and ETH only needs to fall around 13% before your health factor hits 1. That’s not a lot of room, which is exactly why most experienced borrowers stay well under their maximum LTV rather than maxing it out.
A few things worth knowing before you try it:
- Every loan on Aave is over-collateralized. You can’t borrow more than your collateral is worth, which is the opposite of how a mortgage works.
- If your collateral’s value falls far enough that your debt approaches the liquidation threshold, part of your position gets sold off automatically to protect the pool. More on that in the risk section below.
- Interest accrues continuously, not monthly. Check your position regularly if you’re borrowing against volatile collateral.
This is the part of Aave that most closely resembles traditional secured crypto lending. The difference is that code enforces the terms instead of a loan officer.
Flash Loans on Aave v3: Borrowing Without Collateral
A flash loan lets you borrow any amount of an asset with zero collateral, but on one condition; you’ve to repay it, with a fee, before the transaction finishes. But if you can’t, the entire transaction reverts as though it never happened. Aave’s own documentation sets the current fee at 0.05% of the borrowed amount, down from 0.09% in earlier versions.
This only works because of how blockchains process transactions. Everything inside a single transaction either all executes, or all gets undone. Aave uses that property to let you hold millions of dollars for a few seconds, as long as you hand it back in the same breath.
Two common uses:
- Arbitrage: Borrow a token, sell it where it’s priced higher, buy it back where it’s cheaper, repay the loan, and keep the spread, all in a single transaction.
- Collateral swaps: Pay off one loan with borrowed funds and switch your collateral type. Then open a new position, without needing the cash up front.
Flash loans have also been weaponized more than once. Several DeFi exploits have used flash-loaned capital to manipulate prices on thinly traded markets long enough to drain a separate, poorly designed protocol. That’s a real risk in this corner of DeFi, and it’s a big reason serious protocols spend heavily on price-oracle security instead of trusting a single exchange’s quote.
Aave v3’s Risk Management Tools
Aave v3’s biggest upgrade over v2 wasn’t a new asset or a new chain. It was a set of tools built specifically to stop one bad position from becoming everyone’s problem.
Isolation mode and efficiency mode (eMode)
Isolation mode lets Aave list a newer or more volatile asset without exposing the entire protocol to it. Isolated assets can only be borrowed against up to a fixed debt ceiling, and they can’t be combined with other collateral in the same position. If that asset implodes, the damage stays contained.

Efficiency mode does the opposite job for correlated assets. Borrowing ETH against staked ETH carries far less price risk than borrowing ETH against a random altcoin, so eMode raises the LTV accordingly. ChainScore Labs puts the ceiling as high as 97% LTV for correlated pairs under eMode, which is also what makes leveraged “looping” strategies possible: deposit staked ETH, borrow ETH, swap it back to staked ETH, redeposit, repeat.
What stands out to me is how much risk actually concentrates in that loop. Galaxy Research’s snapshot of Aave’s loan book found eMode loans carrying a debt-weighted LTV near 90% and an average health factor of just 1.05. Against roughly 47% LTV and a 1.88 health factor for regular, non-eMode loans.
Health factor and liquidations
The health factor is a single number that tells you how far a position is from liquidation. Above 1, you’re safe. At or below 1, part of your collateral becomes eligible to be sold off to repay your debt.
Say you’ve borrowed against ETH and your health factor drops to 1.1 because ETH’s price fell. You’re not liquidated yet, but you’re close, and a further price drop or accrued interest could push you under. Liquidators, not Aave itself, execute the sale and earn a bonus for doing it, which keeps the system self-policing instead of relying on a central party to catch bad debt.
Why $15B+ Stays Deposited in Aave v3
The reasons capital concentrates here aren’t complicated. Aave has the deepest liquidity of any lending protocol, years without a contract-level hack, audits from multiple independent firms including Cyfrin’s v3 review, and a native stablecoin, GHO, that’s grown to a $583 million market cap by minting directly against Aave collateral, not an outside issuer. It also generates real revenue: roughly $28 million in fees over the past 30 days, some funding AAVE token buybacks since April 2025.
Then came April 19, 2026. Attackers exploited a bridge used by Kelp, a liquid restaking protocol unrelated to Aave, and took 116,500 rsETH. They deposited the stolen tokens on Aave as collateral and borrowed roughly $236 million against them, leaving an estimated $123 million in bad debt, per Galaxy Research and Aave’s governance incident report. TVL dropped $6.6 billion within hours, and the AAVE token fell 16%.
Deposits came back for a specific reason. Aave’s founder confirmed the protocol’s own contracts were never compromised; only its collateral choices were exposed. The Safety Module, funded by roughly $184 million in staked AAVE, exists to absorb exactly this shortfall through staker slashing if governance approves it. A broader industry effort helped backfill the rest. That admission, plus a backstop that actually got used, is what brought large depositors back within weeks.
The lesson holds regardless: Aave’s risk isn’t confined to its own code. Whatever collateral it accepts becomes Aave’s problem the moment something else breaks. Suppliers holding stablecoins or blue-chip assets in isolation mode stayed largely insulated. Anyone looping correlated collateral through eMode carried exactly the exposure that broke in April, whether they realized it or not.
Aave v3 vs. Other Crypto Lending Protocols
Aave v3 and Compound v3 solve the same problem, but a bit differently. ChainScore Labs’ technical comparison frames it as a trade-off between flexibility and simplicity. Aave supports many assets as both collateral and borrowable positions in one pool. Compound v3 isolates each market around a single base asset, usually USDC, and doesn’t let that base asset double as collateral, which frees up more of it for borrowing but limits what you can actually do with a Compound position.
| Aave v3 | Compound v3 | |
| Collateral model | Multiple assets per position | Single base asset per market |
| Cross-chain liquidity | Native Portal bridge | Limited, market by market |
| Rate types | Variable and stable | Utilization-based variable only |
| High-risk asset listings | Isolation mode caps exposure | Not natively supported |
| Flash loans | Yes, 0.05% fee | No native flash loan product |
| Best fit | Multi-asset strategies, leverage, cross-chain apps | Simple, capital-efficient stablecoin markets |
If you want to borrow against a wide mix of collateral, run a leveraged eMode strategy, or move liquidity across chains, Aave is built for that. If you want the simplest possible exposure to a stablecoin market with fewer moving parts, Compound’s narrower design is arguably safer precisely because it does less.
Final Thought
Aave earned its position as DeFi’s largest lender the hard way, through years of audited code, conservative collateral standards, and a governance process that, however slow, has kept the protocol solvent.
The April 2026 exploit didn’t change that history, but it did prove that dominance in DeFi lending is never purely a function of your own code. It’s also a function of everything you agree to hold as collateral. Whatever comes next, whether that’s Aave v4, further growth in GHO, or a rethink of how liquid restaking tokens get priced, that lesson isn’t going away.
For more info on crypto and all things Web3, visit Blockverse.
FAQs
The protocol has undergone several independent hacks, and Aave v3 smart contracts have never been hacked. However, safe doesn’t mean risk-free: other protocols, such as April’s exploit involving rsETH, can cause losses to the pool.
Aave makes money by charging interest on borrowed funds, flash loan fees, and liquidation penalties. The fees totaled approximately $878 million per year by mid-2026, with a smaller portion, around $114 million, being generated as revenue for the protocol.
Aave v3 added isolation mode, eMode, cross-chain Portal transfers, and more granular risk parameters per asset. V2 is still live but now holds a small fraction of total Aave deposits.
Unlike the previous scenario, lenders are not at a significant direct risk unless there is a shortfall event such as the recent rsETH exploit in April, where slashing of the Safety Module funds becomes possible. If prices move against the borrower, then his collateral can be liquidated.
Mainly arbitrage and collateral swaps, both of which need large, temporary capital for a single transaction. They’re also occasionally used as a tool inside exploits against other, less secure protocols.
More than 20 chains, including Ethereum, Arbitrum, Optimism, Base, and Polygon, with Ethereum holding the largest share of deposits.
