A double digit APY on a staked ETH product should bother you a little. Staking ETH pays low single digits, stablecoin lending pays whatever borrow demand says it pays, and vaults cannot invent yield out of nothing. When the headline number is three or four times the base rate, the explanation is almost always one word buried in the strategy docs: looping. Deposit an asset, borrow against it, acquire more of the same asset, deposit that too, repeat until the protocol stops letting you. The yield is real, but the number on the dashboard is also, in a very calculable way, a leverage multiple wearing a costume.
The classic version uses staked ETH. You deposit stETH or a similar liquid staking token into a lending market, borrow plain ETH against it, stake the borrowed ETH to get more stETH, and deposit again. Each pass is smaller than the last because you can only borrow a fraction of what you deposit. Stablecoin loops work the same way, a yield bearing stable as collateral and a plain stable as debt. In both cases you are long and short nearly the same asset, so price risk looks tiny, and the whole trade lives or dies on the spread between two interest rates.
The math behind the headline number
The leverage converges to a clean formula. If you loop at a loan to value ratio of L, total exposure lands at 1 divided by (1 minus L) times your starting capital. Loop at 50% LTV and you hold 2x. Loop at 75% and you hold 4x. Some markets have special modes for correlated pairs that allow 90% or more, which is 10x and up. Flash loan tooling means nobody even does the passes manually anymore, one transaction gets you to full size.
Return on equity is then simple: leverage times the yield your collateral earns, minus (leverage minus 1) times the rate you pay on the debt. Take a hypothetical loop where staked ETH earns roughly 3% and ETH borrows at roughly 2%. At 4x you earn 3% on four units and pay 2% on three, which nets to 6% on your one unit of actual capital. Double the base staking rate, and exactly the kind of number that ends up on a dashboard.
Two things about that formula deserve more attention than they get. First, your real leverage is knowable right now from your own position: total collateral value divided by (collateral minus debt). People who believe they are running a bit of leverage do this arithmetic and find 5x or 6x more often than you would expect. Second, the debt leg is larger than your equity, often much larger, so small moves in the borrow rate produce large moves in your net. In the example above, the borrow rate rising from 2% to 4% takes you from plus 6% to zero, and 5% puts you at minus 3%. A three point move in one floating rate swung the return by nine points, which is what (leverage minus 1) does for you in both directions.
How the spread flips negative
Borrow rates on lending protocols float with utilization, and most markets use a curve with a kink. Below some target utilization the rate climbs gently. Above it the slope turns steep, sometimes brutally so, because the protocol needs to attract deposits and shake out borrowers to keep withdrawal liquidity available. That kink is the single most common way loops die.
The sequence tends to run the same way each time. A loop gets popular, usually because someone posts the APY. New loopers all borrow the same asset, utilization grinds up toward the kink, and the borrow rate creeps toward the collateral yield, so the spread compresses from comfortable to thin. Then something pushes utilization over the kink, a large borrower arrives, a depositor leaves, incentives end somewhere else and capital migrates, and the borrow rate stops creeping and jumps. Every loop in that market flips negative at the same time, and nothing about your position on screen looks different until you go check the rate yourself.
There is a clean breakeven you can compute in advance. The loop loses money outright once the borrow rate exceeds your collateral yield times leverage divided by (leverage minus 1). At 4x on a 3% collateral yield that is 4%. At 10x it is barely 3.3%, meaning a 10x loop on a 3% asset dies the moment the borrow rate passes the collateral yield by a rounding error. Higher leverage buys a bigger headline and a thinner margin against the exact rate move that crowding makes more likely. Incentive tokens muddy all of this, since plenty of published loop APYs only clear breakeven because of emissions, and emissions get cut by governance votes most loopers were not watching.
The unwind is where the damage happens
A negative spread on its own bleeds slowly. A percent or two annualized, you notice, you unwind, you move on. The real problem is that unwinding needs two kinds of liquidity at once, and stress removes both at the same time for the same reason.
To unwind you repay debt and withdraw collateral, in slices or in one flash loan transaction. Withdrawing requires the lending pool to hold idle liquidity, and the same high utilization that spiked the borrow rate means the pool is mostly lent out. You can be solvent, comfortably clear of liquidation, and still unable to pull collateral until borrowers repay or fresh deposits arrive. While you wait, the rate that made you want to leave keeps compounding against you.
With a liquid staking token there is a second door: swap the collateral for the underlying on a DEX, or take the protocol withdrawal queue. Both degrade under stress. The queue stretches to days when everyone shows up at once, and DEX depth thins exactly when the crowd is exiting, so the swap route means selling at a discount to fair value. In a loop that discount does double damage. It is a realized loss on the way out, and if your lending market prices the collateral off its market price rather than its redemption rate, the depeg also drags your health factor down while you are mid exit. Which oracle your market uses is worth finding out on a calm day, well before you need the answer.
Checking your own loop
Whether you run one of these directly or hold a vault that does, the whole thing reduces to a short list you can work through in ten minutes:
- Compute real leverage: total collateral value divided by collateral minus debt. If a vault will not show you both numbers, treat that as your answer.
- Compute the breakeven borrow rate: collateral yield times leverage, divided by leverage minus 1. Write it down somewhere you will see it.
- Compare the current borrow rate to that breakeven, then check where utilization sits relative to the kink. The gap between current rate and breakeven is your real cushion, and proximity to the kink tells you how fast it can close.
- Separate base yield from incentives. Rerun the breakeven with emissions at zero and see whether the loop still clears.
- Check the exits: idle liquidity in the lending pool, DEX depth for your collateral, and the length of the redemption queue. Size the position so at least one exit can absorb it on a bad day.
- Set an unwind trigger below the breakeven rather than at it, because unwinding takes time and the rate keeps moving while you work.
Rate watching is the step people skip, because borrow rates feel like background noise right up until the week they are the entire trade. I keep the rates on anything I am looped into inside the same alert stack I run on Blockcircle for everything else, with triggers at my unwind level rather than my breakeven, since an alert that tells you money is already lost has limited value.
None of this makes loops a scam. The spread is real, the mechanism is genuinely elegant, and a modestly levered loop on a correlated pair with a wide cushion to breakeven is a defensible trade. The headline APY just quietly prices in that you have sold something like an option on borrow rates and exit liquidity, and those two strikes tend to get hit together. Know the leverage, know the breakeven, and pick your exit before utilization picks it for you.