Looping is a way of manufacturing leverage inside a lending protocol. The borrower deposits an asset as collateral, borrows a second asset against it, converts that borrowed asset into more of the collateral asset, and deposits it again. Each cycle adds exposure. Automated implementations do the whole thing in a single transaction using a flash loan rather than iterating manually.
The arithmetic
Maximum leverage is one divided by one minus the loan-to-value ratio. At 80 percent LTV that gives five times the starting capital. At the 82.5 percent LTV documented for ether on a major lending market, it gives about 5.7 times. Under a correlated-asset mode permitting 93 percent, the theoretical ceiling is roughly 14.3 times.
The ceiling is theoretical because each loop adds less than the last and costs gas and swap slippage. Reaching four times exposure from an 80 percent LTV takes roughly seven to nine iterations, and running to the mathematical maximum leaves no buffer at all, so a position at the ceiling is liquidatable on the first adverse tick.
| Maximum LTV | Theoretical maximum leverage | Collateral fall that wipes out equity at max leverage |
|---|---|---|
| 50% | 2.0x | 50% |
| 75% | 4.0x | 25% |
| 80% | 5.0x | 20% |
| 90% | 10.0x | 10% |
| 93% | 14.3x | 7% |
The right-hand column is the part that matters and the part most often skipped. Leverage and fragility are the same number viewed from opposite ends. A position at 93 percent LTV is wiped out by a 7 percent move in the collateral against the borrowed asset, and correlated assets do move against each other by more than 7 percent.
Why anyone loops
Three motives account for most looping. The first is yield spread capture: if the collateral asset earns more than the borrowed asset costs, each loop multiplies the spread. A staking derivative earning 3 percent, borrowed against at 2 percent, produces a 1 percent spread that four times leverage turns into roughly 4 percent on the original capital, before costs.
The second is incentive farming. Points programs and token emissions usually scale with position size, so looping multiplies the reward for the same starting capital. The third is straightforward directional leverage, using a lending market instead of a derivatives venue.
The first motive is the one that determines whether the position survives, because a spread of one percentage point is thin. Borrow rates on lending markets move with utilization and can rise several points within hours when a pool tightens. Staking yields do not move to compensate. A loop that was earning carry in the morning can be paying it by the evening, and unwinding takes as many transactions as building did.
The pairs people actually use
- Liquid staking tokens against ether, the classic loop, capturing the staking yield above the ether borrow rate.
- Liquid restaking tokens against ether, adding restaking points to the same structure and adding a second layer of protocol risk.
- Yield-bearing stablecoins against ordinary stablecoins, capturing the spread between a synthetic dollar's yield and a plain dollar's borrow cost.
- Fixed-yield principal tokens against stablecoins, locking a known yield and financing it at a floating rate.
Every one of these pairs is a bet that two assets stay closely related. Lending protocols recognize this by offering correlated-asset modes that raise the permitted LTV for exactly these combinations. The higher LTV is what makes the loop worth doing and what leaves almost no room for the relationship to break.
What a loop costs to build and unwind
Each iteration involves a borrow, a swap and a deposit, and each swap pays the spread on the venue it routes through. On correlated pairs that spread is small but not zero, and it is paid twice: once building the position and once unwinding it. A loop expecting to earn a one-point spread has to cover perhaps 40 to 60 basis points of round-trip execution cost before it earns anything, which sets a minimum holding period.
Flash-loan implementations remove the accumulating gas cost of iterating but not the swap cost, because the position still has to be built out of real purchases. They also introduce a dependency on the automated contract being available and correctly configured when the position needs to be closed.
Unwinding is the part people underestimate. Closing a loop means repaying debt, withdrawing collateral, swapping back and repeating, in a market where the collateral is falling and the swap venue is thin. The cost of exiting a leveraged position is highest at exactly the moment exiting becomes necessary, which is why a plan to deleverage at a specific health factor is worth more than an intention to watch the position.
How looped positions fail
The failure mode is not usually the collateral falling in absolute terms. It is the collateral falling relative to the borrowed asset, which for correlated pairs happens when the market stops treating them as interchangeable.
In June 2022, staked ether traded at a discount to ether after large holders withdrew several hundred million dollars of liquidity from the main pool linking them. Positions that had borrowed ether against staked ether faced liquidation on a spread that the position's whole design assumed would stay near zero. In April 2024, a liquid restaking token fell more than 18 percent against ether within minutes following an unpopular token distribution announcement, again driven by holders exiting a crowded, heavily looped position rather than by anything wrong with the underlying staking.
Both episodes share a structure. Leverage concentrates many participants in the same trade. The exit is a single liquidity venue. When enough of them try to leave at once, the price relationship the trade depends on breaks precisely when everyone needs it to hold.
Both episodes also show why the size of the discount matters less than who is holding the asset. A four percent discount on a staking derivative is an inconvenience to an unleveraged holder and a liquidation event for a position running at ten times leverage on a correlated-asset mode. The same price move produces entirely different outcomes depending on the leverage stacked behind it, and the leveraged holders are the ones forced to sell.
Running one more carefully
- Size to a stress case, not the maximum. Decide what collateral-versus-debt move the position must survive, and set leverage from that rather than from the LTV ceiling.
- Model the borrow rate rising, not staying still. Ask what the position earns if the borrow rate rises three points, which utilization spikes routinely deliver.
- Check the oracle. Market price or redemption rate, and how often it updates.
- Know the exit. Which venue the unwind trades through, and what depth it has when everyone else is unwinding too.
- Keep repayment capital outside the loop. Deleveraging requires a transaction, and the moment it is needed is the moment gas is expensive and interfaces are congested.
Looping is a legitimate way to express a view about a spread. It is not a yield product, and packaging it as a one-click vault does not change the arithmetic in the table above.