Say you bought 2 ETH years ago at a fraction of today’s price. Selling them to pay for a car, a deposit on a flat or a tax bill would turn the whole rise into a taxable gain. There is another route, the one most DeFi borrowers take: borrow against crypto instead of selling it, by locking the ether as collateral on a lending protocol and taking dollars out. On Aave V3 Ethereum, the largest blockchain that runs exactly as written when it is called, holding and moving funds according to its code, without an intermediary able to change the outcome.">smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool, borrowing USDC cost 4.78% a year at a variable rate when we read the contract on 29 October 2026 at 09:49 UTC. The ether stays yours, on paper at least, and nothing has been sold.
Once people borrow against crypto, three uses come up again and again. The first is plain liquidity: cash without a sale. The second is leverage by looping, where the borrowed asset buys more of the collateral, which is deposited and borrowed against again. The third is the stablecoin carry trade: borrow a dollar token at one rate and park it somewhere that pays a higher one. Each is worked through below with the real rates of 29 October 2026, snapshots that move every day, next to what the tax authorities of the United States, the United Kingdom, France and Italy actually say: less than many crypto sites claim. Nothing here is advice to borrow, loop or carry; it is a description of how these positions work and where they break.
Two lessons come first and are not repeated. The health factor, the liquidation threshold and the liquidation price are computed step by step in our lesson on the health factor and liquidation price. Who liquidates a position, how, and what the borrower loses is the subject of the previous lesson, how DeFi liquidations work. The pool model, utilisation-driven rates and eMode are introduced in our guide to crypto lending and borrowing.
Borrow against crypto instead of selling: what you avoid and what you take on
A loan is not a sale, until something sells the collateral
Mechanically, a crypto-backed loan on a protocol such as Aave is simple. You deposit an asset the protocol accepts as collateral, for example WETH (ether wrapped as a standard token). The contract lets you borrow another asset up to a share of the collateral’s value, the loan-to-value ratio. On Aave V3 Ethereum Core, as read on-chain on 29 October 2026, WETH had an LTV of 81% and a liquidation threshold of 84%, and USDC borrowers paid a variable 4.78%. There is no repayment date. Interest is added to the debt block after block, and you can repay whenever you like.
What you have avoided is a sale. Your ether has not been exchanged for dollars; you hold a debt in dollars and a claim on your ether in the pool. What you have taken on is a price trigger. If the value of the ether falls far enough relative to the debt, anyone can repay part of the loan and take part of your collateral at a discount: a liquidation. That forced sale is the event the whole strategy was meant to avoid, and it happens at the worst moment, with a penalty on top (5% for WETH on Aave V3 Ethereum Core on our reading date).
A third thing changes, and borrowers often miss it: who controls the coins. On a DeFi protocol the collateral sits in a public smart contract whose rules anyone can read; the protocol cannot spend it except through those rules. On a centralised lender the terms of use decide, and they may let the company re-lend your coins. That difference matters for safety and, in the United Kingdom, for tax, as we will see.
Where crypto loans come from: pools, isolated markets and app front ends
Most DeFi borrowing happens in three kinds of venue. Shared pools such as Aave V3, Compound V3 and SparkLend put every lender of an asset in one pot. Isolated markets such as Morpho pair one collateral with one loan asset and fixed parameters. Vault systems such as Sky’s mint a stablecoin against your collateral rather than lending you someone else’s deposit; that design is covered in our lesson on CDP stablecoins. Rates differ between venues. On 29 October 2026 DefiLlama showed variable USDC borrowing on Compound V3 Ethereum at 5.09% and USDS borrowing on SparkLend at 4.32%, against the 4.78% we read directly on Aave.
Exchanges have also started to put DeFi loans behind their own apps. Coinbase lets customers borrow USDC against bitcoin held on Coinbase, with the loan run by Morpho on Base, Coinbase’s layer-2 network, according to Morpho’s announcement; a fixed-rate version, with a set rate and a maturity date, was announced on 22 September 2026 on Morpho Midnight, Decrypt reported, while Morpho put the older variable-rate product at more than 1.4 billion dollars of active loans. Fixed-rate borrowing has its own mechanics and is the subject of a later lesson.
Centralised crypto lenders that hold the collateral themselves are the older model, and their record is mixed. Celsius froze withdrawals in June 2022 and went bankrupt. On 4 January 2023 the judge in its US bankruptcy case, Martin Glenn, ruled that crypto in its interest-bearing Earn accounts belonged to the Celsius estate under its terms of use, as Decrypt reported, which left those customers as unsecured creditors. The Earn accounts held about 4.2 billion dollars of crypto shortly before the July 2022 bankruptcy filing. The story sits in our history of DeFi and its failures. For a borrower the point is narrow: the contract, not the marketing, decides who owns posted collateral and whether the platform may use it.
That question has a physical answer on DeFi: the collateral sits in a contract you can inspect on a block explorer, and your position is a record anyone can query. On a centralised platform it sits wherever the company’s books say.
A cautious example: 2 ETH and a loan of a quarter of their value
Take 2 ETH at an example price of 2,500 dollars, so 5,000 dollars of collateral (the price is a round figure for the arithmetic, not a market quote). Borrowing 25% of that value gives 1,250 USDC. With WETH’s 84% liquidation threshold, the health factor is 5,000 x 0.84 / 1,250 = 3.36. The liquidation price is the debt divided by the quantity of ether times the threshold: 1,250 / (2 x 0.84) = about 744 dollars, a fall of roughly 70% from the example price. At 4.78% a year the loan costs about 60 dollars a year, if the rate stayed where it was, which it will not.
Borrow 40% instead, 2,000 USDC, and the liquidation price rises to about 1,190 dollars, a fall of 52%. Borrow the full 81% LTV and you would be liquidated after a fall of only a few percent. The LTV is a ceiling set by the protocol. The gap between the amount a borrower may take and the amount the position can carry through a bad week is the whole margin of safety. The formulas and a full table are in the health factor lesson.
Is borrowing against crypto taxable? What the tax authorities say
The phrase “borrow, don’t sell” travels with a tax promise. The promise is half right in most countries: receiving borrowed money is not income and not a sale. The other half is where readers get caught. A liquidation sells your collateral, and some ways of repaying or using the loan create a taxable disposal too. And on DeFi loans specifically, most tax authorities have said very little. This section is general information, not tax advice; the rules below were checked on 29 October 2026.
United States: no IRS ruling on crypto loans
The IRS has issued no specific guidance on loans backed by crypto collateral. Its starting point is still Notice 2014-21, which says “virtual currency is treated as property” and that general tax principles for property transactions apply. The notice does not mention loans or collateral, and the IRS pages on digital assets do not address them either.
In that silence, the general reading among tax practitioners is that borrowing stablecoins against crypto you keep is a loan and not a sale, so no gain is recognised when the cash arrives. That is the practitioners’ reading, not an IRS position. Two events are much clearer under the general property rules. A liquidation disposes of your collateral: the protocol sells it to repay your debt, and you have a gain or loss against your cost basis. Swapping crypto into dollars or stablecoins to repay the loan is also a disposal of the crypto you hand over. Our guide to crypto taxes in the United States covers reporting, including the Form 1040 digital-asset question.
United Kingdom: HMRC’s collateral test
HMRC has gone further than any other authority in this lesson. Its Cryptoassets Manual has a section on DeFi lending and borrowing, and page CRYPTO61640 on collateral, last updated on 28 November 2025, says whether posting collateral is a disposal depends on the platform’s terms and conditions. If the platform can deal freely with the tokens, HMRC treats that as a strong indicator that the platform has acquired beneficial ownership, so the borrower made a disposal when posting them, at their sterling market value. If the platform is specifically restricted from using them, the borrower keeps beneficial ownership and there is no disposal; a later liquidation is then the borrower’s own disposal, with the platform treated as a nominee under section 26 of the Taxation of Chargeable Gains Act 1992.
Two details matter for borrowers. Tokens taken as a liquidation penalty are not an allowable deduction, according to the same page. And page CRYPTO61630 opens with a plain rule: “Where a person borrows tokens, they are making an acquisition.” Citing the High Court case Chaney v Watkis, HMRC treats the borrower’s obligation to hand tokens back as the acquisition cost of the tokens borrowed. A borrowed stablecoin such as USDC is a token too, so the rule is not limited to borrowing WETH. What HMRC does not say is which side of its collateral test a pooled protocol like Aave falls on: the assets supplied there are lent on to other borrowers by the pool’s code, not by a company acting under terms of use. That is unsettled. Our guide to crypto taxes in the United Kingdom covers rates and allowances.
France: an emprunt is not a cession, but DeFi is a grey zone
French readers face a general rule rather than a crypto-loan rule. Article 150 VH bis of the tax code taxes the cession of digital assets for euros or in exchange for goods or services; exchanges between digital assets without a cash balance are deferred. Borrowing euros or stablecoins against crypto is, on that general rule, neither: you have sold nothing. Converting the collateral into euros, or spending it on goods, is a taxable cession. The impots.gouv.fr page on declaring digital-asset gains, last updated on 17 July 2026, confirms the flat tax (PFU) of 31.4% on 2026 gains and the deferral for crypto-to-crypto exchanges, but says nothing about loans, collateral or DeFi. How a liquidation on a DeFi protocol is treated, and whether depositing collateral counts as an exchange, has not been settled by the administration.
Italy: 33% from 2026 and the same-function rule
Gains on crypto-assets realised from 1 January 2026 pay a 33% substitute tax, set by the 2025 Budget Law (Law 207 of 2024, article 1, paragraph 24). From 1 January 2027 the same rule sits in article 304 of the new income tax code, Legislative Decree 117 of 19 June 2026, which also keeps a reduced 26% for e-money tokens denominated in euros. USDC, a dollar token, is not among them. The rule on exchanges, by contrast, is in Agenzia delle Entrate Circular 30/E of 27 October 2023: the swap of crypto-assets with “eguali caratteristiche e funzioni” (the same characteristics and functions) is not a taxable event. The circular gives bitcoin for ether as an example of a neutral swap, and treats an exchange of a crypto-currency for an e-money token, which carries a claim at face value on fiat money, as taxable.
For a borrower this has a sharp edge. Borrowing USDC is a loan, not an exchange. But a liquidation that sells ETH into USDC, or a swap of ETH into USDC to repay, can be a taxable exchange under the circular’s reasoning, because USDC is an e-money token under MiCA. The same-function rule is carried over word for word into article 76 of the new code. The circular does not discuss lending, collateral or DeFi at all, so the treatment of the deposit itself is unsettled.
| Event | United States | United Kingdom | France | Italy |
|---|---|---|---|---|
| Borrowing a stablecoin against crypto you keep | No IRS guidance; practitioners generally read it as a loan, not a sale | Depends on the platform’s terms (CRYPTO61640); borrowing tokens is an acquisition (CRYPTO61630) | Not a cession under the general rule; no doctrine on DeFi | A loan; Circular 30/E silent on DeFi |
| Liquidation of the collateral | Disposal of property under general rules | Borrower’s disposal, or none if posting the collateral was already one | Taxable if into euros, goods or services; DeFi treatment unsettled | Taxable if into euros or an e-money token such as USDC |
| Swapping collateral into dollars or euros to repay | Disposal | Disposal | Cession at 31.4% PFU on 2026 gains | Taxable; 33% on 2026 gains (Law 207/2024) |
Rules as checked on 29 October 2026. This is general information, not tax advice: how a given position is treated depends on facts, such as the platform’s terms, that only a tax adviser looking at the case can weigh.
Looping: borrowing to buy more of the same
Looping uses the loan for something else entirely. Instead of spending the borrowed asset, you buy more of the collateral with it, deposit that, borrow again, and repeat. The classic version on Ethereum pairs staking means depositing a proof of stake coin with a protocol that stakes it for you and issues a transferable token (such as Lido stETH or rETH) representing the deposit plus rewards, which you can use or sell while the coins stay staked.">wstETH, Lido‘s wrapped staked ether, with WETH. wstETH earns the Ethereum staking yield through a rising exchange rate against stETH, as explained in our lesson on interest-bearing tokens. If you can borrow WETH for less than wstETH earns, every loop adds a little spread, multiplied by the leverage.
How an stETH loop is built on Aave
On Aave V3 Ethereum, wstETH cannot be borrowed (its borrowing flag was off on 29 October 2026) but WETH can, so the loop runs in one direction. The steps, described in words:
- Supply wstETH to the pool and turn on eMode category 1, “ETH correlated”, which on 29 October 2026 allowed an LTV of 93%, a liquidation threshold of 95% and a liquidation penalty of 1% for ether-linked assets (outside eMode, wstETH’s LTV was 79%).
- Borrow WETH against the wstETH.
- Swap the WETH for wstETH on a decentralised exchange, or stake it and wrap it.
- Supply the new wstETH and borrow WETH again, as many times as the target leverage requires.
Doing this by hand costs gas at each step, so most loopers use a front end that builds the whole position in one transaction with a flash loan, a technique covered in a later lesson. The economics are the same either way.
The leverage formula and the health factor at each step
If each loop borrows a share b of the collateral just added, the collateral after many loops tends towards 1 + b + b2 + … = 1/(1 – b) times your own stake. That is the maximum leverage: at the eMode LTV of 93%, 1/0.07 = about 14.3 times. For a position at leverage L, collateral is L and debt is L – 1 (in units of your own equity, all valued in ether), so the health factor is 0.95 x L / (L – 1). The net yield on your equity is L x the staking yield minus (L – 1) x the borrow rate.
We used two snapshots from 29 October 2026: a staking yield of about 2.3% (Lido’s 7-day average APR was 2.326%) and a WETH variable borrow rate of 1.94% on Aave V3 Ethereum Core. The table is an example calculation, not a forecast, and leaves out swap costs and gas.
| Leverage | Debt per unit of equity | Health factor (threshold 95%) | Fall in the wstETH/ETH price that reaches HF 1 | Net yield on equity at a 1.94% borrow rate | Net yield if the borrow rate rises to 3% |
|---|---|---|---|---|---|
| 2x | 1 | 1.90 | 47.4% | 2.66% | 1.6% |
| 5x | 4 | 1.19 | 15.8% | 3.74% | -0.5% |
| 8x | 7 | 1.09 | 7.9% | 4.82% | -2.6% |
| 10x | 9 | 1.06 | 5.3% | 5.54% | -4.0% |
| 14x | 13 | 1.02 | 2.3% | 6.98% | -6.8% |
Read the last two columns together. At 1.94% the spread is only about 0.36 points per unit borrowed, so ten times leverage turns 2.3% into roughly 5.5%. A rise of about one point in the WETH borrow rate wipes that out and then some: at 3% every leverage above about 4.3x loses money. The break-even borrow rate is the staking yield itself. Each extra point on the borrow rate costs L – 1 points of yield on your equity, which is why loopers watch the WETH utilisation rate more closely than the price of ether.
Why the price of ether in dollars barely touches this loop
A wstETH-WETH loop is a bet on a spread, not on ether. Both sides of the position are priced from ether: on 29 October 2026 Aave V3 Ethereum used a Chainlink ETH/USD feed for WETH and an oracle named “Capped wstETH / stETH(ETH) / USD” for wstETH, according to the oracle contract Aave’s pool pointed to when we read it on 29 October 2026. That oracle comes from BGD Labs’ price-cap adapters, whose published code computes the wstETH price from Chainlink’s ETH/USD feed and the wstETH/stETH ratio read from Lido’s contract, with a cap on how fast that ratio may grow (8.8% a year in the adapter’s settings on our reading date). If ether falls 30% in dollars, collateral and debt fall together and the health factor hardly moves.
So what can move the wstETH/ETH price in the table above? Mainly a fall in Lido’s own exchange rate, which would take a slashing of validators on a large scale, or a change of oracle by governance. A discount of stETH on the open market, like the one of June 2022, does not by this design feed into Aave’s price. That is a property of this oracle on this market on our reading date: other lending markets may use a market price, and the design can be changed by a governance vote.
What breaks a loop: borrow rates, exits, caps and code
The first risk is the one the table shows: the borrow rate. WETH’s rate on Aave follows utilisation, and loopers themselves push utilisation up. When many depositors withdraw WETH at once, the rate can jump within hours and stay high, eating the position from the inside: interest is added to the debt faster than staking adds to the collateral, the health factor drifts down, and a position that never sees a price move can still end up near liquidation.
The second is the exit. Unwinding means selling wstETH for WETH to repay, and at ten times leverage you sell ten times your stake. If stETH trades at a discount on exchanges at that moment, as in 2022, when it slipped below ether after Terra’s collapse in May and reached a record discount of about 8% on 13 June, with Celsius and Three Arrows Capital in trouble (the episode is told in our lesson on DeFi composability), the oracle may say your position is safe while the market makes leaving it expensive. Lido’s own withdrawal queue avoids the discount but takes time.
Third, liquidation still exists. The eMode penalty is only 1%, but it applies to the debt repaid and the collateral seized, and a position at 14x has almost no room for an oracle change or a burst of interest. Fourth, supply and borrow caps: when the wstETH supply cap or the WETH borrow cap is reached, new loops cannot be built and existing ones cannot be enlarged. Fifth, every layer adds code and governance risk: Lido, Aave, the swap route, any front end. On 18 April 2026 an attacker took rsETH, Kelp DAO’s restaked ether token, out of its bridge without the ether meant to back it, posted part of it as collateral on Aave and borrowed WETH against it. Galaxy Research found that WETH utilisation on Aave then stayed above 99% for 12.7 days, which for loopers meant a borrow rate at the top of its curve. Our guide to crypto lending tells that story in full.
The stablecoin carry trade: borrow dollars at X, earn Y
The carry trade is the dollar version of the same idea. You post collateral, often ether, borrow a stablecoin such as USDC, and deposit it in something that pays more than the loan costs: Ethena’s staked USDe (sUSDe), Sky’s savings token sUSDS, or a Pendle principal token (PT) that locks in a yield to a fixed date. The collateral keeps its own exposure, so this is not a hedge: it is a second position stacked on the first.
Here are the legs as they stood on 29 October 2026. Each rate is a snapshot taken that morning: Aave’s USDC borrow rate read from two sources less than an hour apart differed by about 0.16 points.
| Leg | Rate on 29 October 2026 | Source |
|---|---|---|
| Borrow USDC, Aave V3 Ethereum Core | 4.78% variable | On-chain read, 09:49 UTC |
| Borrow USDT, Aave V3 Ethereum Core | 5.22% variable | On-chain read, 09:49 UTC |
| Borrow USDC, Compound V3 Ethereum | 5.09% variable | DefiLlama |
| Borrow USDS, SparkLend | 4.32% variable | DefiLlama |
| Hold sUSDe (Ethena) | 5.06% variable | DefiLlama |
| Hold sUSDS (Sky Savings) | 3.80% variable | DefiLlama |
| Pendle PT-sUSDe, maturity 26 November 2026 | 4.73% implied, fixed to maturity | Pendle API |
| Pendle PT-sUSDS, maturity 26 November 2026 | 5.63% implied, fixed to maturity | Pendle API |
Against a 4.78% USDC loan on Aave, sUSDe paid 0.28 points more, sUSDS paid 0.98 points less, PT-sUSDe was 0.05 points short and PT-sUSDS was 0.85 points ahead. On 100,000 dollars borrowed into sUSDe, the spread is about 280 dollars a year before gas, if both rates held, and they rarely do. There is also an opportunity cost: a holder who already had USDC could have supplied it on Aave at about 4.05% that morning with no loan at all.
Why the spread vanishes: inversion, funding and maturity
The most common way the trade fails is rate inversion. Both legs are variable. The borrow rate rises with utilisation; the yield on sUSDe depends on Ethena’s revenue from staked ether and from the funding and basis of its hedging positions, and Ethena’s documentation explains that rewards stop when those turn negative. sUSDe also has a cooldown between the request to unstake and the return of USDe. The sUSDe contract set it at one day when we read it on 29 October 2026 (DefiLlama still listed seven days), and Ethena can raise it to as much as 90 days, so a holder cannot always leave the instant the spread flips, except by selling sUSDe on the market. How USDe works is explained in our guide to Ethena’s USDe, and sUSDS in our lesson on yield-bearing stablecoins.
Pendle PTs fix the yield, but only to maturity. The PT-sUSDS market maturing on 26 November 2026 had about 4.7 million dollars of liquidity on our reading date, so exiting a large position early moves the price; and a fixed yield paired with a variable loan only stays profitable while the loan rate stays below it. Some markets also accept sUSDe or PTs as collateral, which lets traders loop the carry itself; the leverage arithmetic above applies, with a stablecoin oracle and a larger liquidation penalty. We do not detail one here.
When the vault is the risk: curated strategies
Chasing a higher yield often leads to vaults run by third-party curators or fund managers that promise a few points more. In November 2025 Stream Finance suspended withdrawals after disclosing a loss it attributed to an external fund manager, and its xUSD token lost most of its value; lenders on several protocols that had accepted Stream’s tokens as collateral were exposed. The chain of losses is in our lesson on how DeFi protocols stack. In a carry trade built on such a vault, the extra points of yield come with a risk that no rate table shows: the curator’s judgement.
Questions to answer before a position, not after
None of this says whether to borrow. It does give a short list of facts that are worth knowing before any of the three strategies, and that are all readable on-chain or in the protocol’s documentation.
- Where the collateral sits and who may use it: a public contract with fixed rules, or a company’s balance sheet under terms of use.
- Which oracle prices the collateral and the debt, and whether it follows a market price or an exchange rate.
- Whether the rate is variable or fixed, and what the rate curve does at high utilisation.
- What the exit costs: swap depth, withdrawal queues, cooldowns and PT liquidity.
- Which event creates a tax disposal in your country, including a liquidation.
- What the liquidation penalty is and at what price, or ratio, it starts.
Anyone who has not used a lending protocol before can rehearse all of this with test tokens on Aave’s Sepolia testnet, as shown in the health factor lesson: supply, borrow, watch the health factor, repay. Test tokens cost nothing and teach the interface. On the main network even a small loop costs several transactions of gas, and the risks above apply from the first dollar.
The next lesson of the course moves from the borrower to a tool both loopers and attackers use: the flash loan, a loan that must be repaid in the same transaction.
