Crypto · DeFi and staking
Crypto lending and yield: where the interest comes from
How crypto loans are collateralized, how the health factor and liquidation work, where a quoted yield comes from, and how often ether-backed loans were liquidated.
On this page
- How a collateralized crypto loan works
- Health factor and liquidation
- How often it would have happened: ether, 2024 to 2026
- Where the yield comes from
- Lending through a company
- Protocol lending, company lending and a bank account
- How Finabase does it
- Frequently asked questions
- Summary
- Check your understanding
Crypto lending lets one person earn interest on coins by lending them, and another borrow against coins without selling them. It happens in two places: on lending protocols, which are smart contracts that hold the collateral and enforce the terms in code, and at companies that take customer deposits and lend them out as they see fit. The advertised number in both cases is a yield, often several times what a bank account pays.
A yield is a payment from someone, for something. The useful questions are who is paying, why they are willing to, and what happens to the lender if they stop. This lesson covers how collateralized crypto loans work, the arithmetic of the health factor and liquidation, the sources a yield can come from, and what two years of ether prices would have done to borrowers.
How a collateralized crypto loan works
A bank lends against income and a credit history, and can sue a borrower who does not pay. A lending protocol knows nothing about its users and can sue nobody. It relies on one thing only: collateral it already holds, worth more than the loan.
The flow has four steps.
- Suppliers deposit a token, say the dollar stablecoin USDC, into a pool for that token. They earn interest and can withdraw whenever the pool has unlent funds.
- A borrower deposits collateral, say ether (ETH), into the same protocol.
- The borrower takes a loan from the USDC pool, up to a set share of the collateral's value. That share is the maximum loan-to-value ratio (LTV).
- Interest accrues every block. The borrower can repay at any time and take the collateral back. There is no term and no repayment schedule.
Why would anyone post $20,000 of ether to borrow $10,000 of dollars they could have had by selling half the ether? Usually to keep exposure to the ether's price, to avoid a taxable sale, or to buy more crypto with the borrowed money, which is leverage. The first two only work if the collateral is never sold out from under the borrower, which is the risk the rest of this page measures.
The price feed
The contract has to know what the collateral is worth, and a blockchain cannot look up a price by itself. Lending protocols read prices from an oracle, a service that publishes market prices on chain. A loan is therefore only as sound as its oracle: a wrong or manipulated price can trigger liquidations that should not happen, or allow borrowing against collateral that is not worth what the contract believes.
Health factor and liquidation
Each collateral asset has a liquidation threshold: the loan-to-value ratio at which the position stops being safe. It is set above the maximum LTV for new borrowing, so a position does not become liquidatable the moment it is opened. The protocol Aave, whose documentation this section follows, combines the numbers into one figure.
A worked example
A borrower deposits 10 ETH when ether is $2,000 and borrows $10,000 of USDC. The collateral is worth $20,000, so the loan-to-value ratio is 50%. With an 80% liquidation threshold:
health factor = 20,000 × 0.80 ÷ 10,000 = 1.60
The debt stays at $10,000 (plus interest) whatever happens to ether. The collateral does not.
| Ether | Collateral value | Loan-to-value | Health factor |
|---|---|---|---|
| $2,000 | $20,000 | 50.0% | 1.60 |
| $1,800 | $18,000 | 55.6% | 1.44 |
| $1,600 | $16,000 | 62.5% | 1.28 |
| $1,400 | $14,000 | 71.4% | 1.12 |
| $1,250 | $12,500 | 80.0% | 1.00 |
| $1,100 | $11,000 | 90.9% | 0.88 |
The liquidation price is the debt divided by the collateral amount times the threshold: 10,000 ÷ (10 × 0.80) = $1,250, a fall of 37.5% from $2,000. A shortcut: the fall that liquidates a loan is 1 − (LTV ÷ threshold). At 50% LTV and an 80% threshold that is 1 − 0.625 = 37.5%.
What liquidation does
When the health factor drops below 1.0, any third party, called a liquidator, may repay part or all of the debt and take collateral in return at a discount. The discount is the liquidation penalty (Aave calls it the liquidation bonus), and the borrower pays it.
Suppose ether is at $1,200, which puts the health factor at 0.96, and the penalty is 5%, an example value (penalties differ by asset). Under Aave's version 3 rules a liquidator may repay up to half of the debt while the health factor is above 0.95, and all of it at 0.95 or below or when the collateral or the debt is worth less than $2,000. Here both are well above that.
- The liquidator repays $5,000 of USDC.
- The liquidator takes collateral worth $5,000 × 1.05 = $5,250, which is 4.375 ETH.
- The borrower is left with 5.625 ETH, worth $6,750, and a debt of $5,000. The health factor is back to 6,750 × 0.80 ÷ 5,000 = 1.08.
The borrower has lost 4.375 ETH to cancel $5,000 of debt, $250 more than the debt was worth, and it happened at a low price they did not choose. Nobody phones first. There is no margin call and no grace period: liquidators are bots that act within seconds, at any hour, because crypto trades around the clock.
How often it would have happened: ether, 2024 to 2026
The data: daily bars of ether from November 18, 2024 to November 18, 2026. For each of the 366 days from November 18, 2024 to November 18, 2025 we opened a hypothetical loan against ether at that day's close and followed it for 365 days, counting it as liquidated if any day's low reached the liquidation price. The threshold is the 80% example value and interest is ignored, which flatters the borrower slightly.
| Borrowed (share of collateral) | Fall that liquidates | Liquidated within a year | Average days until it happened |
|---|---|---|---|
| 30% | 62.5% | 82 of 366 (22%) | 227 |
| 40% | 50.0% | 208 of 366 (57%) | 132 |
| 50% | 37.5% | 268 of 366 (73%) | 113 |
| 60% | 25.0% | 326 of 366 (89%) | 87 |
| 70% | 12.5% | 334 of 366 (91%) | 39 |
These numbers come from one asset over one period, and the period was a bad one: ether ended it 43.8% lower than it began, after trading between a low of $1,383 in April 2025 and a close of $4,831 in August 2025. In a rising market the shares would be far smaller. The point is not that a loan at 50% is liquidated 73% of the time. It is that a 37.5% fall, which sounds remote when the loan is opened, was an ordinary event for this collateral.
Follow the three loans in the figure, all opened on November 18, 2024. The one that borrowed 70% was liquidated on February 2, 2025, eleven weeks later. The 50% loan lasted until March 4, 2025. The 30% loan survived the whole two years, though the low of April 2025 came within 15% of its liquidation price. Less than six months after the 50% loan was liquidated, ether closed 50% above the price at which the loan had been opened. Being right about the eventual price did not help a borrower who had been sold out at the bottom.
Where the yield comes from
On a lending protocol, suppliers are paid by borrowers and by nobody else. Rates move with utilization, the share of the pool that is lent out.
If a pool holds $100 million, $80 million of it is borrowed at 5%, and the protocol keeps 10% of interest, suppliers earn about 5% × 0.80 × 0.90 = 3.6%. The numbers are invented; the structure is not. Borrow rates rise steeply as utilization approaches 100%, to pull in deposits and push out borrowers. That also means the supply rate you see today is not a promise: it changes block by block. At very high utilization, suppliers may be unable to withdraw until someone repays.
More generally, every quoted crypto yield traces back to one or more of these sources.
| Source of the yield | Who pays | What can stop it |
|---|---|---|
| Interest from overcollateralized borrowers | Traders who want leverage | Demand for leverage falls; bad debt after a crash |
| Trading fees in a liquidity pool | Traders swapping tokens | Volume falls; impermanent loss exceeds fees |
| Staking rewards | The network, by issuing new coins | The rate falls as more is staked; the coin's price |
| Reward tokens ("incentives") | The protocol, by issuing its token | The token's price falls; the program ends |
| Uncollateralized lending by a company | Trading firms that borrow | A borrower defaults; the company fails |
| Reserve income on a stablecoin | Treasury bills held by the issuer | Rates fall; rules on paying interest |
| New deposits | Later depositors, in a Ponzi scheme | Deposits slow |
Rows two and three have their own lessons: Decentralized exchanges and liquidity pools: how AMMs price trades and Crypto staking: how rewards work and what can go wrong. The fourth deserves suspicion. A pool that advertises 40% a year, of which 2% is interest and 38% is paid in a freshly issued token, is paying depositors in something whose supply grows as fast as it is handed out. The practice is called yield farming.
A test, not advice: for any yield, write down in one sentence who pays it and why. Compare the rate with what a Treasury bill pays on the same day. The gap is the market's price for the risks in the right-hand column, and it is not always enough.
Leverage loops do not create yield
A common strategy deposits ether, borrows against it, buys more ether, deposits that, and repeats. Borrowing 50% each round converges on double the exposure: $10,000 of ether becomes a $20,000 position with $10,000 of debt.
If ether supplied earns 2% and the borrowed asset costs 4%, the position earns 2% of $20,000 and pays 4% of $10,000: $400 in, $400 out. The loop has created no income. It has doubled the exposure to ether's price, and a fall of 37.5% now liquidates a position in which the owner's $10,000 has already shrunk to $2,500. Loops pay only when reward tokens are added on top, or when the collateral itself yields more than the loan costs, and both conditions can reverse quickly.
Lending through a company
The other kind of crypto lender is a business: the customer transfers coins to it, the company promises a rate, and lends or invests the coins however it chooses. Three of the largest failed within five months of each other in 2022.
- Celsius Network paused all withdrawals on June 12, 2022 and filed for Chapter 11 bankruptcy on July 13, 2022. Its founder pleaded guilty to fraud and was sentenced to 12 years in prison on May 8, 2025.
- Voyager Digital filed for bankruptcy in July 2022. The Federal Trade Commission later charged that it had falsely told customers their deposits were FDIC-insured.
- BlockFi agreed in February 2022 to pay $100 million in penalties to the SEC and state regulators, which found its interest accounts were unregistered securities. It filed for bankruptcy on November 28, 2022.
The common pattern: customers believed they held deposits, while in law they had made unsecured loans to a company that lent on with little or no collateral. When borrowers failed and withdrawals surged, there was nothing like a liquidation engine to protect the depositors, and no insurance. Crypto held at a lender is not covered by FDIC insurance or SIPC. What a custodian's failure means for its customers is covered in Crypto wallets and custody: keys, seed phrases, hot and cold storage.
A protocol and a company fail differently. The protocol's loans are collateralized and its books are public, but its code or its oracle can be exploited. The company's loans are opaque, and its customers are creditors if it fails.
Protocol lending, company lending and a bank account
| Lending protocol | Crypto lending company | Bank savings account | |
|---|---|---|---|
| Who holds the assets | A smart contract | The company | The bank |
| Borrowers post collateral | Yes, more than the loan | Often not, or not enough | Varies; the bank bears the losses |
| Can you see the loans | Yes, on chain | No | No |
| Rate | Floats every block | Set by the company | Set by the bank |
| If borrowers default | Collateral is sold automatically | The company absorbs it, or fails | The bank's capital absorbs it |
| Main ways to lose | Contract bug, oracle failure, a failed peg | Insolvency, fraud, frozen withdrawals | Bank failure above insured limits |
| Government insurance | None | None | FDIC, up to the legal limit |
| Who can freeze withdrawals | High utilization; governance in some cases | The company, at its discretion | Regulators, rarely |
How Finabase does it
Finabase shows prices, not lending markets. The crypto list and the quote pages for coins such as ETH.X show price, change, volume and market cap, and include the tokens of lending protocols, for example AAVE.X. Finabase does not show supply or borrow rates, utilization, health factors or any quoted yield, and it does not offer lending, borrowing or interest on crypto.
Two tools are still useful beside this lesson. The ETH.X chart shows how far a collateral asset has fallen within any period you choose, which is the number a liquidation price should be set against. And the Treasury page shows what the US government pays to borrow, the baseline any other yield should be compared with. The loan statistics on this page were computed from Finabase's daily ether bars frozen on November 18, 2026, using each day's low. Holding a lending protocol's token is not the same as lending on it: the token is a claim on governance and whatever the protocol decides to distribute, and its price is what the quote page shows.
Frequently asked questions
Is crypto lending safe?
No form of it is free of risk, and none carries deposit insurance. A lender on a protocol depends on the contract's code, its price feed and the stablecoin or coin being lent. A customer of a lending company depends on that company staying solvent and honest. Borrowers face liquidation at a penalty after a price fall.
What is a good health factor?
There is no safe number, only a trade-off. A health factor of 1.6 with an 80% threshold is liquidated by a 37.5% fall in the collateral; 2.67 (a 30% loan) needs a 62.5% fall. Ether fell more than 50% from its starting level within the two years studied here.
Why are crypto loans overcollateralized?
Because the protocol cannot identify or pursue a borrower, the collateral is the only thing that makes the lender whole. Requiring more collateral than debt leaves room to sell it before its value drops below what is owed.
Where does the interest on a crypto savings product come from?
From whoever borrows the coins: traders seeking leverage on a protocol, or trading firms and other borrowers at a lending company. Some products add reward tokens. If a product cannot say who the borrower is and what secures the loan, the depositor is carrying a risk they cannot see.
Is interest earned on crypto taxable in the United States?
The IRS treats digital assets as property and requires income received in digital assets to be reported. It has issued a ruling on staking rewards, covered in Crypto staking: how rewards work and what can go wrong. We found no IRS guidance specific to interest from lending protocols as of November 2026, so this page does not state how a particular product is taxed.
Summary
Crypto lending replaces credit checks with collateral. A borrower posts more than they borrow, the contract tracks a health factor of collateral value times a threshold over debt, and a fall below 1.0 lets liquidators sell the collateral at a penalty within seconds. On ether's prices from 2024 to 2026, most hypothetical loans above 40% of the collateral's value were liquidated within a year. Lenders are paid by borrowers through a rate that floats with utilization, and every other yield in crypto traces to fees, new issuance, reward tokens or an unsecured borrower. The companies that paid interest on deposits in 2022 showed the last case at its worst: three of the largest froze withdrawals and went bankrupt, and their customers turned out to be creditors.
Check your understanding
5 questions on this lesson. Each answer comes with the reason, and nothing is scored.
A borrower posts 5 ETH worth $3,000 each and borrows $6,000. The liquidation threshold is 80%. What is the health factor?
- 2.50
- 2.00
- 1.25
- 0.50
Show the answer
B. 2.00 Collateral value × threshold ÷ debt = 15,000 × 0.80 ÷ 6,000 = 2.00. The position becomes liquidatable when that figure drops below 1.0.
For the same loan (5 ETH, $6,000 debt, 80% threshold), at what ether price can the position be liquidated?
- $1,200
- $1,500
- $2,400
Show the answer
B. $1,500 Liquidation price = debt ÷ (coins × threshold) = 6,000 ÷ (5 × 0.80) = $1,500, a fall of 50% from $3,000. The shortcut: 1 − LTV ÷ threshold = 1 − 0.40 ÷ 0.80 = 50%.
A lending pool has $50 million supplied and $30 million borrowed at 6%. The protocol keeps 10% of interest. About what do suppliers earn?
- 6.0%
- 3.6%
- 3.24%
Show the answer
C. 3.24% Utilization is 30 ÷ 50 = 60%. Supply rate ≈ borrow rate × utilization × (1 − reserve factor) = 6% × 0.60 × 0.90 = 3.24%. Suppliers are paid only on the part of the pool that is lent, less the protocol’s share.
A pool advertises 30% a year: 3% from borrowers’ interest and 27% paid in the protocol’s own newly issued token. What is the main thing to understand about the 27%?
- It is guaranteed by the protocol’s reserves
- It is paid in a token whose supply grows as it is handed out, so its dollar value depends on that token’s price
- It is interest paid by borrowers at a higher rate
Show the answer
B. It is paid in a token whose supply grows as it is handed out, so its dollar value depends on that token’s price Reward tokens are issued by the protocol, not earned from borrowers. If the token’s price falls, or the program ends, that part of the yield shrinks or disappears.
Customers of a crypto lending company that failed in 2022 generally found that they were what?
- Depositors protected by FDIC insurance
- Owners of segregated coins that were returned at once
- Unsecured creditors of the company, waiting on a bankruptcy
Show the answer
C. Unsecured creditors of the company, waiting on a bankruptcy They had transferred their coins to a company that lent them on. When it failed, they held a claim in bankruptcy, with no deposit insurance and no collateral of their own.
Sources and further reading
- Aave — Liquidations (health factor formula; how much of a debt can be liquidated)
- Aave documentation — Reserve concepts (loan-to-value, liquidation threshold, interest rate model)
- Aave — Frequently asked questions (supply rate, reserve factor, liquidation penalty)
- SEC — BlockFi agrees to pay $100 million in penalties and pursue registration of its crypto lending product (press release 2022-26, February 14, 2022)
- FTC — Action against Voyager Digital over false claims that customer deposits were FDIC-insured (October 12, 2023)
- US Department of Justice — Celsius founder Alexander Mashinsky sentenced to 12 years in prison (May 8, 2025)
- SEC Investor Bulletin — Crypto asset interest-bearing accounts (February 14, 2022)
- Finabase chart data — ETH.X daily bars, November 18, 2024 to November 18, 2026; loan arithmetic computed by Finabase
