How Aave's core works
Aave dominates decentralized lending with about $12-15 billion locked at peaks. The system is built on a simple idea: deposit collateral, borrow against it at variable or fixed rates. The LendingPool contract handles all the heavy lifting—deposits, withdrawals, borrowing, payback. Chainlink feeds keep collateral prices fresh on-chain.
When you deposit USDC, you get aUSDC tokens back. These tokens quietly accrue interest as the pool earns lending fees. The exchange rate between regular USDC and aUSDC climbs steadily. No claim transactions needed. You just hold it and watch it grow. Variable rates float with supply and demand. Stable rates lock in at borrow time—helpful if you want to predict costs.
The system uses loan-to-value ratios (LTV) to cap how much you can borrow. Stablecoins hit 75% LTV because they're safe. ETH maxes around 80%. Risky tokens might get zero. When your collateral falls short, liquidators jump in to clean up the mess. They repay debt and grab your collateral at a slight discount (usually 5-10%). This keeps the protocol solvent.
Flash loans and atomic arbitrage
Aave's flashLoan function lets you borrow any amount—collateral-free—as long as you pay it back in the same transaction. Borrow 10 million USDC, do something with it, pay it back plus 0.05% fee in one atomic block. Fail at any step and the whole thing reverts.
This created a cottage industry. Traders exploit price gaps across exchanges. Liquidators grab underwater positions and rebalance in one shot. Collateral swaps happen without liquidation risk. The fee structure dropped from 0.09% to 0.05% to stay competitive. Other protocols copied the model. Now Uniswap, dYdX, and others offer similar tools. Competition matters here.
GHO: the internal stablecoin
Aave launched GHO as a native stablecoin that lives on the protocol itself. You borrow GHO against approved collateral. Unlike USDC (backed by actual dollars), GHO has no reserve vault. It's algorithmic—supply grows when demand is high and shrinks when prices climb above $1. The protocol adjusts interest rates to manage the peg. Low peg? Cut rates to encourage borrowing. High peg? Raise rates to choke off new issuance.
The "bucket" system limits how much GHO you can mint from any single collateral type. This prevents one asset from dominating the system. Governance votes on ceilings. The Facilitation Manager lets approved partners issue GHO through alternate pathways beyond direct borrowing. It's basically MakerDAO's multi-collateral DAI with different mechanics under the hood.
Multiple chains mean fragmented liquidity
Aave runs on Ethereum, Polygon, Arbitrum, Optimism, Avalanche, Base, and Gnosis Chain. Each deployment is independent—separate pools, separate governance votes, separate risk parameters. Ethereum has the deepest liquidity. The other chains are cheaper to trade on but thinner. Arbitrage happens when interest rates diverge between chains. That USDC you deposit on Arbitrum earns slightly different returns than the same asset on Ethereum.
Cross-chain bridges help a little. Stargate moves stablecoins. Native Layer 2 bridges ship assets between chains. But moving money costs gas and time. The fragmentation is real. Aave's spread across multiple chains reflects user demand for lower costs and faster settlement, even if liquidity gets split up.
AAVE governance and the token
AAVE holders vote on protocol changes and treasury decisions. One token equals one vote in Snapshot voting. Total supply is 16 million. The Ecosystem Reserve holds about 3 million set aside for grants and partnerships. Governance unfolds in stages: forum discussion, Snapshot signal test, then on-chain vote through the Governor contract. Approved proposals execute after a time delay (usually 24 hours for standard tweaks, longer for sensitive changes).
This staged approach stops flash loan attacks where someone borrows voting tokens in a single block to game an election. AAVE appreciation brings more governance engagement. Price drops mean fewer people bother voting. The Risk Council proposes monthly parameter adjustments. Early holders and investors have outsized voting power, which rankles some decentralization purists.
Managing liquidation and collateral risk
The Risk Council reviews collateral constantly. What's the volatility? How deep is the liquidity? How many tokens concentrate in a few holders? LTV ratios vary wildly by asset—stablecoins get 75% because they're stable. Volatile garbage gets 0-10%. Liquidators keep things honest. When you're underwater, they buy your debt and seize your collateral at a 5-10% discount. That gap compensates them for speed and capital risk.
E-mode lets you use correlated assets (stETH and ETH are basically the same thing) with much higher LTV—up to 98.5%. You trade flexibility for concentration risk. If stETH tanks, that's a problem because Aave holds tons of it.
Isolated pools for institutions
Aave created isolated pools so institutions can set custom risk rules without endangering the main protocol. A hedge fund might deposit RWA tokens or illiquid stuff. That lives in its own sandbox. Even if it blows up, the core protocol survives. This attracted heavy institutional capital. Some pools hold over $500 million.
The regulatory angle matters too. Aave has engaged with regulators about stablecoin policy and GHO specifically. Governance controls which assets the protocol supports. No single party bears legal risk—that's spread across all governance voters. It's both a strength (decentralized) and a headache (who's actually liable?).
Interest rates follow utilization curves
Rates climb automatically as more capital gets lent out. The system has a target utilization (usually 80-85%). Below that, rates rise slowly. Above that, they spike hard. This encourages people to deposit more when the pool is strained and repay when it's congested. Variable rates update every block as conditions change. Stable rates lock in at borrow time.
Stablecoin lending gets cheaper base rates (0-2% at target) because it's safer. Volatile assets cost more (3-8%) because they're risky. Governance tweaks these numbers constantly. Professional risk analysts publish recommendations. It's become fairly sophisticated for what started as "supply and demand."
Protocol tokens and sustainability
Aave discussed a Protocol Token (APT) that would capture fees separately from AAVE governance tokens. This would let the protocol generate revenue without diluting governance. The specifics are still under discussion. Early incentive mining programs are winding down as the protocol matures. Growth mode is ending. Fee-based sustainability is coming.
The security council has emergency authority to pause the protocol during attacks—critical for quick response but it contradicts the pure decentralization ideal. Governance is slowly moving that power to multi-sig contracts to spread it around.
Security and incident recovery
Aave went through multiple audits (OpenZeppelin, Certora, Trail of Bits) and formal verification of sensitive code. The protocol survived the March 2020 flash crash and the May 2021 liquidation chaos without major fund loss. Bug bounties reach $500,000 for critical vulnerabilities. The security council responds quickly to threats.
Governance layers authority by sensitivity. Boring parameter tweaks get fast approval. Dangerous upgrades need more scrutiny. This is the messy compromise between decentralization and operational speed.
Building with other protocols
Aave's aTokens are standard ERC-20s. Anyone can build on them. Curve created leverage products. MakerDAO uses aTokens for DAI generation. Compound interest layers nicely. Lido (liquid staking) is tightly integrated. That tight coupling with stETH creates systemic risk if staking goes wrong. The protocol monitors this constantly.
Leverage protocols let you recycle collateral—deposit, borrow, redeposit, borrow more. It amplifies gains and losses. One bad cascade could trigger forced sells across multiple protocols holding the same collateral.
Where Aave is heading
Governance has floated Aave V4, a major rewrite to clean up technical debt. RWA integration would let real-world assets (bonds, invoices) back loans. Layer 2 native tokens might become collateral. Sky's MakerDAO rebrand parallels Aave's thinking about institutional positioning.
Both protocols are converging on the same big idea: stability, institutions, real-world asset integration. They'll probably serve different niches rather than destroy each other. Governance is exploring automated parameter adjustment (within bounds) to speed up routine decisions and better delegation systems to reduce voting friction. The goal is decentralization that actually works at scale without paralysis.