DeFi lending: a complete guide to decentralized money markets
Fi lending: a complete guide to decentralized money markets…
You've held a position through a brutal drawdown, told yourself you'd be patient, and now your capital is finally sitting at a comfortable multiple again. The instinct to take profits is real — but so is the opportunity cost of leaving that capital idle when you don't actually need to sell. That's the specific tension that pulls most people into DeFi lending protocols in the first place: how do you keep exposure, keep optionality, and earn a sustainable baseline of yield without taking on a leverage profile you can't sleep through the night with?
This guide walks through how decentralized lending actually works — not the marketing layer, but the mechanical layer underneath. We'll look at how money markets price credit, how collateral is structured, what triggers a liquidation, and which advanced utilities (flash loans and credit delegation) deserve a careful look before you use them. The goal is to give you a working framework so the next dashboard you open doesn't feel like guesswork.
The architecture of decentralized money markets: beyond traditional credit
The first mental shift worth making is that DeFi lending is not a lender-to-borrower transaction. There is no one on the other side of the table approving your credit history, your income, or your identity. Instead, almost every protocol you interact with is a pool-based or vault-based money market run by smart contracts. Lenders deposit assets into a shared pool, borrowers draw against posted collateral, and interest rates adjust algorithmically based on how much of the pool is currently being borrowed.
Two structural models dominate today. The first is the pooled approach used by Aave and Compound, where each market maintains liquidity for a base asset (often a stablecoin like USDC) and accepts a menu of other assets as collateral. The second is the vault model pioneered by Maker, where each collateralized debt position (CDP) is its own isolated vault against a single collateral type, minting DAI against the deposit. Both designs are overcollateralized by default — meaning you must post assets worth more than what you borrow — and that overcollateralization is the load-bearing wall of the whole system.
In DeFi lending, your collateral isn't a safety net — it is the loan.
Here's how the two compare in practice:
| Parameter | Pooled market (Aave, Compound III) | Vault model (Maker) |
|---|---|---|
| Borrowing unit | A market-level base asset (e.g., USDC) | DAI minted against the vault |
| Collateral | Multiple assets accepted per market | Typically one collateral type per vault |
| Liquidity source | Shared liquidity pool | Vault-specific; debt ceiling per type |
| Interest rate | Algorithmic, driven by utilization | Stability fee set by governance |
| Position structure | One position can use multiple collaterals | One vault per collateral type |
The practical implication of this architecture is that borrowing capacity and borrowing cost are functions of two things: how much collateral you post, and how heavily the pool (or vault) is already being used. There's no bank officer, but there is an algorithm constantly repricing risk. And one design detail worth flagging: in Compound III's market structure, the supplied collateral assets themselves do not earn or pay interest — only the market's base asset does, with a positive base-asset balance earning the supply rate and a negative balance paying the borrow rate. That distinction matters when you compare headline APYs across protocols, because the asset earning yield isn't always the asset you're depositing.
Interest rate dynamics: utilization curves and kinked models
Interest rates in DeFi aren't negotiated — they're computed. The most important variable in that calculation is utilization, which for Compound III-style markets is defined as TotalBorrows divided by TotalSupply. If a pool has $80 million borrowed against $100 million supplied, utilization sits at 80%, and that single number determines what borrowers pay and what suppliers earn.
Most protocols use a two-slope model with a kink. Below the kink, rates rise gradually as utilization climbs — the protocol wants to attract more suppliers. Above the kink, rates climb sharply, because high utilization means there's little liquidity left for new withdrawals, and the protocol needs to discourage further borrowing. In Compound III, the supply and borrow rate models depend on utilization, contain this kink, and accrue every second based on the block timestamp.
What this means for you as a lender is that the headline APY you see on a dashboard is a snapshot. The actual sustainable rate depends on where utilization sits relative to the kink for that market on that day. A pool running at 95% utilization may offer a very attractive APY, but the same number also signals that withdrawals may not be smooth — and that you're now deep into the high-rate slope where risk premiums are essentially being passed to you. The trick is not to chase the top APY; it's to find a market where utilization is healthy, the kink is sensibly calibrated, and the asset is one you're comfortable holding anyway.
For borrowers, the same curve cuts the other way. Borrowing at low utilization is cheap but rarely sustainable, because every additional borrower pushes the rate up. The "good" borrow rates on a protocol are usually the rates you see when utilization is moderate — high enough to be a real yield signal for lenders, low enough to make your leverage economics work.
Collateral management: distinguishing LTV from liquidation thresholds
This is where most new users get tripped up, and it's worth taking a minute to separate two concepts that protocols often display side by side but that mean different things.
Loan-to-value (LTV) is the maximum ratio at which you can open a new borrow position. The liquidation threshold is the collateral ratio at which your position becomes eligible for liquidation. Both are governance-configured per asset and per market, and both can change — but they govern different moments in a position's life. LTV is what determines whether you can borrow at all. The liquidation threshold is what determines whether you keep what you borrowed.
Aave's documentation makes this distinction explicit: LTV sets the maximum borrowing capacity against fresh collateral, while the liquidation threshold is the ratio that triggers liquidation if your collateral value falls relative to your debt. The same protocol may give an asset an LTV of 75% and a liquidation threshold of 82%, which means you can borrow up to 75% of the collateral's value initially, but if your debt-to-collateral ratio climbs to 82%, you're now liquidatable.
Compound III expresses this slightly differently, using a collateral factor rather than a published LTV. The documentation gives an example: a WBTC borrow collateral factor of 85% lets you borrow up to 85% of the supplied WBTC's USD value in the market's base asset. The same logic applies — you can borrow up to a stated share, and the liquidation mechanism activates when your position becomes undercollateralized by the protocol's threshold.
Maker's vaults take a third approach. A collateralized debt position measures collateralization as the USD value of collateral divided by the USD value of DAI debt. The protocol's example gives 150% as a liquidation ratio: collateral must remain at least 1.5 times the debt. Drop below that, and the position is in liquidation territory.
A useful way to think about it: in pooled protocols, LTV and the liquidation threshold are separate per-asset governance parameters; in vault protocols, the ratio is a single number governing both borrowing capacity and the liquidation boundary. Either way, your job is to know which number is which — and to keep your position meaningfully inside both lines.
LTV decides how much you can borrow; the liquidation threshold decides how much you can lose.
| Parameter | Aave | Compound III | Maker Vault |
|---|---|---|---|
| Initial borrowing cap | Loan-to-value (LTV) | Collateral factor (per asset) | Vault-specific minimum ratio |
| Liquidation trigger | Liquidation threshold | Liquidation threshold | Liquidation ratio |
| Health metric | Health factor (weighted) | Collateral factor vs. debt | Collateral/debt ratio |
| Governance-controlled | Yes, per asset | Yes, per market | Yes, per vault type |
Risk mitigation and the mechanics of position liquidation
No matter how careful you are, oracle prices move and governance parameters shift. Understanding the liquidation mechanic is what turns borrowing from a guess into a managed position.
Aave expresses position health through a single number called the health factor. The formula is total collateral value multiplied by the weighted-average liquidation threshold, divided by total borrow value. A health factor above 1 means your position is healthy; a health factor below 1 means it's liquidatable. Most experienced borrowers keep a personal buffer — staying above 1.5, sometimes higher — so a sharp price move doesn't push them into the danger zone overnight.
When a position is liquidatable, the protocol doesn't take all your collateral. Aave's documentation describes a close-factor mechanic: up to 50% of total debt can be liquidated if the health factor is above 0.95 and both the collateral and debt sides are at least $2,000; up to 100% may be liquidated when the health factor is at 0.95 or lower, or when either side falls below $2,000. Liquidators repay a portion of your debt and receive a corresponding share of your collateral at a discount — that's the liquidation bonus that incentivizes third parties to step in. The discount is the cost you, as the borrower, pay for letting your position drift into unsafe territory.
A few practical rules that have served me well over the years:
1. Treat any protocol's published thresholds as ceilings, not targets. If the maximum safe LTV is 75%, plan to operate at 50–60%.
2. Watch utilization, not just APY. A market at 95% utilization isn't "high yield" — it's a queue.
3. Diversify collateral across uncorrelated assets if your position is large enough that a single-asset drawdown could matter.
4. Set alerts at 1.5× health factor (or your equivalent buffer), not at 1.0×. By the time you're at 1.0, your options are already reduced.
5. Re-check the underlying oracle. A stale price is as dangerous as a falling one.
Advanced credit utilities: flash loans and delegated borrowing
Once the basics are clear, two utilities are worth understanding — both because they're genuinely useful and because they're often misunderstood.
Flash loans
A flash loan is a loan that exists only during a single transaction execution. Under the ERC-7399 specification, the principal plus the lender-defined fee must be repaid before the callback ends; otherwise the transaction reverts. That atomic structure is the point: a flash loan is uncollateralized because it doesn't need to be — the loan lives for exactly as long as one transaction, and the transaction either completes in full or unwinds entirely. Flash loans are the building blocks of arbitrage, collateral swaps, and self-liquidation strategies. They are not a way to borrow "for free" with no cost; the fee is real, and the smart-contract risk is real, especially on less-audited deployments. If you're sizing flash-loan strategies, factor in the premium the same way you'd factor in gas — as a non-negotiable line item.
Credit delegation
Credit delegation is structurally different. In Aave's design, a delegator calls approveDelegation on a debt-token contract for a specified amount, and a delegatee then accesses that credit by passing the delegator's address as onBehalfOf in a Pool borrow call. The debt is attributed to the delegator, not the borrower; the delegatee never has to post collateral of their own. This makes credit delegation useful for treasury operations, structured products, and any arrangement where one party wants to extend credit to another on-chain without giving up custody of underlying assets.
The crucial caveat is that credit delegation is not trustless unsecured credit. The on-chain mechanism handles debt attribution and the approval mechanics, but the off-chain relationship — who is borrowing, what they're borrowing for, what happens if they don't repay — is governed by a separate agreement between the parties. The same diligence you'd apply to any off-chain agreement applies here. As a general rule of thumb, anytime you sign anything in crypto — a smart contract approval, a delegated borrowing arrangement, or the contract terms a KOL sends over before a paid promotion deal — the protocol layer handles the on-chain mechanics, but the obligation is still yours.
Navigating the trade-offs
DeFi lending is one of the cleaner yield primitives in Web3 because the mechanics are explicit. Rates adjust by formula, collateral is transparent, and liquidations are deterministic. That same explicitness is also what makes it unforgiving: there is no grace period, no courtesy call, no override. If your health factor drops below 1, the market handles your position without asking.
For passive income seekers, the practical path looks something like this. Lend into markets where the asset is one you'd happily hold anyway, where utilization sits comfortably below the kink, and where the protocol has a track record you can verify. If you borrow, do it at well under the LTV ceiling, maintain a real buffer above the liquidation threshold, and remember that the cost of being right too late is exactly as bad as being wrong. The protocols will keep working whether or not you do — and capital efficiency, in the long run, comes from staying in the game rather than maximizing every cycle.