Perps vs. Looping: What a Leveraged Long Actually Costs

There are two ways to be three times long Bitcoin, and almost nobody picks between them on price.

Route A: open a 3x long on a perpetual futures exchange. One click, one screen, done in ten seconds.

Route B: deposit WBTC into a lending market, borrow stablecoins against it, buy more WBTC, deposit that too. Several transactions, a health factor to watch, gas to pay.

The price chart does exactly the same thing to both positions. Bitcoin goes up 10%, you make 30% on your money, either way. The only thing that differs is the bill you pay for holding — and when we pulled six years of funding rates and realised borrow costs, the gap turned out to be large enough that it should probably be the first thing you check, not the last.

At today's rates, holding 3x BTC costs 8.99% of your notional per year on the perp and 2.96% through the loop. On ETH — where your collateral earns staking yield the whole time — it is 9.36% versus 0.74%.

On $300k of exposure that is $18,000 and $26,000 a year respectively, for owning exactly the same thing.

Here is why that gap exists, why it is structural rather than a quirk of this week's market, and — importantly — the reason the cheap route is not automatically the right one.

The perp charges rent whether or not anyone is home

A perpetual future has no expiry date, so something has to keep its price glued to spot. That something is the funding rate: every eight hours, whichever side of the market is crowded pays the other side. Too many longs, longs pay. Too many shorts, shorts get paid. Elegant, self-correcting, entirely reasonable.

Except that is not quite the formula the major venues use. Hyperliquid — like Binance before it — computes funding as:

funding = premium + clamp(0.01% − premium, ±0.05%)

Read that second term slowly, because it is the whole article. When the perp trades close to spot the premium is roughly zero, the two premium terms cancel, and funding does not settle at zero. It settles at exactly 0.01% per 8 hours — which is 10.95% a year.

That is not a bug. It is a hard-coded assumption, inherited from traditional futures, that a long is borrowing dollars and ought to pay interest on them. The exchange is contractually obliged to invent a fee when the market has no opinion.

And the market has no opinion remarkably often. Across the 28,064 hourly prints since July 2023:

• BTC funding sat at exactly that floor — not near it, on it — 52% of the time.
• ETH, 50% of the time.
• The median print for both assets: +10.95% APR.
• The mean, dragged upward by the manias: +14.4% APR.

Every hourly funding print since July 2023, on a log scale. The distribution is a reasonable bell around zero — with one absurd red spike standing on top of it at exactly +10.95%. That spike is the clamp, and it is where half of all hours live.

So a perp long is paying something close to 11% a year just for showing up. Half the time that fee carries no information about crowding, positioning, or anything else. It is the flat rate the machine falls back to.

Borrow rates have no such floor. A lending market's rate is a published function of how much of the pool is currently lent out — nothing more. When nobody wants to borrow, the rate goes to nearly nothing, because there is no mechanism forcing it upward.

That asymmetry is the entire trade.

The loop bills you on the debt, not on the position

The second reason the loop is cheaper is subtler, and it is where most back-of-envelope comparisons go wrong.

A perp charges funding on your whole notional. Three times long means three times the funding.

A loop charges borrow interest on the borrowed part only — which at 3x leverage is two-thirds of the position, not all of it. And if your collateral happens to be a yield-bearing asset like wstETH, you are simultaneously earning on the entire stack, including the leveraged part.

One line captures it. Call your borrow rate b, your staking yield s, and your leverage L. The loop is cheaper than the perp whenever funding sits above:

f* = b × (L−1)/L − s

That is the break-even funding rate — the level below which you should just use the perp. Notice what it does not contain: position size, entry price, or anything about Bitcoin. Only three rates and your leverage.

Plug today's numbers in at 3x:

	perp funding	loop cost	break-even funding	verdict
BTC (WBTC collateral)	8.99%	2.96%	2.96%	loop, by 6.0 pts
ETH (wstETH collateral)	9.36%	0.74%	0.74%	loop, by 8.6 pts

For the perp to win on BTC, funding would have to fall below 3%. It would have to fall below 0.7% on ETH. Actual funding on both is hovering near 9%.

The ETH case deserves a second look. A 4.45% borrow on two-thirds of the position costs 2.96%, and 2.23% of Lido staking yield on the whole position claws most of that back. You end up paying 0.74% a year to hold triple exposure to ether. Borrow somewhere slightly cheaper and the number goes negative — the position pays you to exist. We will come back to that.

The gap widens exactly when you want leverage most

If this were a feature of one sleepy September it would not be worth an article. So we ran it across every phase of the last two cycles.

phase	funding (BTC / ETH)	borrow	loop cost	who wins
Early bull, 2020–21	+24.7% / +38.3%	13.1%	8.7%	loop
Mania / blow-off, 2021	+64.7% / +82.5%	15.7%	10.5%	loop, hugely
Deep bear, 2022	+3.9% / −0.7%	2.0%	1.3%	split — perp on ETH
Early bull (ETF run), 2024	+36.2% / +36.4%	8.6%	5.6% / 2.1%	loop
Mania / blow-off, 2025	+24.8% / +25.6%	11.1%	7.4% / 4.4%	loop
Bear, 2026	+3.1% / +3.9%	4.5%	3.0% / 0.5%	near-tie / loop
Bear-market rally (today)	+8.7% / +9.1%	4.4%	2.9% / 0.7%	loop

Look at the mania rows. In the 2021 blow-off, ETH perp funding averaged 82.5% a year while stablecoin borrow averaged 15.7%. Both rates are driven by the same underlying force — everyone wants leverage at once — but funding climbed roughly four times as fast.

The reason is who sets each number. Funding is bid by whoever is most desperate to be long right now, and there is no ceiling on desperation. A borrow rate is a published curve on a utilisation ratio, and when it spikes, capital floods in to supply the pool and pushes it back down within days.

So the loop's cost advantage is not constant. It is largest precisely in the market where you most want to be levered. The perp long who held through the 2021 top paid four-fifths of their notional in funding for the privilege.

Six years of the same question, asked every day. Orange is what the perp charged, blue is what the loop charged, and the blue shading is the loop's edge. Note the two things the chart does that a table cannot: how violently the orange line spikes in every mania, and how rarely it dips underneath the blue one.

Run it forward over the whole Hyperliquid era and the arithmetic becomes almost rude:

holding 3x through...	perp	loop	difference on $300k
The 2024–25 bull leg, BTC	39.6%	10.4%	$87,500
The 2024–25 bull leg, ETH	37.2%	4.3%	$98,500
Since July 2023, ETH	45.3%	4.8%	$121,500

Holding 3x ETH through the last bull leg cost 37% of notional in funding versus 4.3% through wstETH on Aave. That single line is the argument.

The win rate is not close either. Across the whole Hyperliquid era the perp was the cheaper route on 17% of days for BTC and 14% for ETH. Five days out of six, the loop.

Now the catch, and it is a real one

Everything above says the loop is cheaper. Nothing above says it is safer, and at the same nominal leverage it emphatically is not.

Here is what "3x" means on each venue once the price starts falling:

	liquidation buffer at 3x
Perp, BTC (1.25% maintenance margin)	−32.5%
Loop, WBTC on Aave (0.78 liquidation threshold)	−14.5%
Perp, ETH (2.00% maintenance margin)	−32.0%
Loop, wstETH on Aave (0.78 threshold)	−14.5%

The perp tolerates an 18-point deeper drawdown before it kills you, on either asset. That is not a technicality. It is the difference between surviving an ordinary bull-market correction and not surviving it.

We checked that against history — the worst peak-to-trough move inside each of the last fifteen cycle phases, measured against each buffer. A 3x loop on Aave would have been liquidated in eleven of those fifteen phases on BTC. The 3x perp, in four. In seven phases the loop died and the perp lived.

Including, notably, the 2024 ETF run — the single best stretch of the cycle to have been levered long. A 15.7% dip inside that rally is enough to close a 14.5% buffer. You would have had the right trade and still lost the position.

So the honest framing is not "perp 3x versus loop 3x". Those are different products wearing the same number.

Two things make it tractable, and both matter.

1. Compare at equal risk, not equal leverage. Dial the loop down until its liquidation point matches the perp's and you land at 2.11x on BTC and 2.13x on ETH — about 1.4x more equity for the same exposure. And even then:

at an identical −32% liquidation point	cost per year
BTC perp, 3x	+8.99%
BTC loop, 2.11x	+2.34%
ETH perp, 3x	+9.36%
ETH loop, 2.13x	+0.13%

The loop is still four times cheaper on BTC and effectively free on ETH, at genuinely identical downside. The cost advantage is large enough to fund the extra collateral several times over.

2. Liquidation means different things in the two places. Aave liquidates you partially — a 50% close factor with a 5–6% penalty — and you can top up or repay at any moment before it happens. A 3x perp liquidation takes essentially the entire margin, instantly, with nobody to call. The loop gets liquidated more often and hurts far less each time.

If you do loop, do not loop on the default

Aave mainnet USDC at 4.45% is the obvious venue and not the cheapest one. A sweep of stablecoin markets that accept BTC/ETH collateral, taken while writing this:

venue	stablecoin	borrow rate	break-even funding, BTC	break-even funding, ETH
Aave v3 (Ethereum)	GHO	3.00%	2.00%	−0.25%
Aave v3 (Ethereum)	RLUSD	3.15%	2.10%	−0.15%
Aave v3 (Ethereum)	USDG	3.52%	2.35%	0.12%
Aave v3 (Arbitrum)	USDC	3.70%	2.47%	0.22%
SparkLend	USDS	3.93%	2.62%	0.37%
Aave v3 (Ethereum)	USDC	4.45%	2.96%	0.74%

Treat every rate in that table as a photograph, not a contract. Borrow rates are a live function of pool utilisation and they move continuously — the Aave USDC row had already drifted by a fifth of a point between the snapshot and publication. The ranking is more durable than the levels, but check both on the day you open the position, and again while you are holding it.

And only bother with stablecoins that have real liquidity behind them. This is the filter most people skip. A cheap borrow rate is worthless if you cannot get out of the thing you borrowed: the loop requires you to swap the borrowed stablecoin into more collateral on the way in, and swap back to repay on the way out. So it has to trade in size against USDC or USDT without handing your entire edge to slippage. GHO, RLUSD and USDG all clear that bar — you can route them back to USDC or USDT and out again. A thin market advertising half a point less does not, and the round trip will cost you more than the rate saved. If the build-and-unwind cycle is unfamiliar, start with how leveraged looping actually works.

Those negative break-evens on ETH are not typos. Borrow GHO at 3% against wstETH and Lido's staking yield more than covers the interest. The loop carries positively: you are paid a little to hold 3x long ether. The perp cannot reach that number under any market condition, because its floor is 11%.

Two other venues deserve a mention. Morpho and Fluid run wstETH/stablecoin markets at roughly 86% LLTV against Aave's 78%, which widens your liquidation buffer from −14.5% to about −22.5% at 3x — a meaningful chunk of the risk gap closed for free. The trade-off is thinner liquidity and per-market curator and oracle risk. Do not confuse these with their 94–96% LLTV wstETH markets: those borrow ETH, which is a staking-yield trade, not a leveraged long.

What the cost model cannot price

On the loop, variable rates are the real exposure. Your 4.45% is not contractual. Aave's curve kinks hard above roughly 90% utilisation, and USDC borrow hit a 7-day average of 29.1% in February 2021 and 18.2% in March 2024. Those spikes mean-revert within days — but they arrive precisely when everyone wants leverage, which is when you are holding, and when unwinding is most expensive. Prefer markets running well below the kink.

Collateral has its own story. WBTC carries BitGo custody risk; cbBTC is the usual alternative at the same parameters. On the wstETH side there is genuinely good news: Aave prices wstETH as ETH price × the wstETH/stETH exchange rate, not the market price of wstETH — so a secondary-market depeg of the kind that broke people in June 2022 does not directly move your health factor.

On the perp, funding is unbounded in the other direction. It averaged 65–82% in the 2021 blow-off and printed above 100% APR at the 99th percentile. Venue risk is real too — Hyperliquid runs a small validator set, and the March 2025 JELLY episode showed the exchange will step into its own markets. And your margin sits in USDC earning nothing at all.

Both routes, in other words, have a tail the carry table does not show. The loop's tail is a rate spike into a liquidation cascade; the perp's is a funding blow-off and counterparty concentration. Neither is in the headline number.

The rule that falls out of six years of data

regime	typical funding	typical borrow	use
Early / mid bull	15–40%	6–9%	loop — large edge
Mania / blow-off	25–80%	11–16%	loop — never hold a perp long here
Post-top / correction	4–8%	5–7%	loop — moderate edge
Bear / capitulation	0 to negative	2–4%	perp — the one regime it wins
Holding under a week	—	—	perp — execution dominates carry

Compute f* = b × (L−1)/L − s, compare it to trailing funding, and take whichever side of the line you land on. It takes thirty seconds and it is the difference between paying 9% and paying 1%.

Right now, with funding near 9%, borrow near 4% and staking at 2.2%: use the loop. On ETH the case is close to overwhelming. On BTC the edge is six points, but the collateral earns nothing and the liquidation threshold is tighter, so if you expect more downside from here the perp is a reasonable call.

And whatever you do, do not run the loop at a nominal 3x. A 14.5% move kills it, and Bitcoin has produced a drawdown that size inside eleven of the last fifteen cycle phases. Run it at roughly 2.1x, match the perp's risk profile exactly, and still pay a third of the price. And read why circular lending goes wrong before you leave one unattended.

The convenient route is not usually the expensive one by eight percentage points a year. This time it is.

---

Data: Hyperliquid hourly funding since July 2023 (28,064 prints), Binance 8-hour funding since 2019 for the 2020–22 cycle, Aave v2/v3 realised borrow costs read from cumulative on-chain interest indices, Lido staking APR via DefiLlama, prices from Binance daily klines. Rates quoted as of 2026-09-12.

This article is educational and is not financial advice. Leveraged positions can be liquidated in full on either venue; never deploy capital you cannot afford to lose.