Is funding-rate arbitrage risk-free?

No. It is delta-neutral, which is a different thing. Three risks remain: the round-trip cost of roughly 0.38% has to be earned back before you profit, the spot-perp basis can be wider when you exit than when you entered, and funding can turn negative so that you pay instead of collect. The one that actually liquidates accounts is margin — the short perp leg showed adverse excursions of +116% to +400% in our tests, and it needs portfolio margin to survive them.

How much does funding-rate carry actually return?

Our measured result is +4.7% a year after costs, unleveraged, which is roughly comparable to Treasury bills. Gross funding on BTC has averaged 11.38% a year since 2020, but you do not capture that: costs and the periods where funding is too thin to enter take the difference. Anyone quoting double-digit unleveraged returns for this trade is either using 2021 data or ignoring fees.

Why is funding so low in 2026?

Because the leverage imbalance the trade is paid to absorb has largely gone. Funding is the price of that imbalance, and across all 708 BTC prints in 2026 it has not once exceeded the 0.01% per 8h base rate, while 29.4% of prints were negative. More participants are willing to take the short side of perps than there were in 2021, so the price of doing it has fallen.

Should I run carry on several coins to diversify?

It does not diversify the way you would hope. Funding behaves largely as one market-wide factor — we measured pairwise correlation of 0.54 across symbols — so a dry spell on BTC is usually a dry spell everywhere. Running several symbols protects you against choosing a bad one; it does not raise the expected return.

Is Funding Rate Carry Still Profitable?

We have tested a lot of strategies here and published the results whether or not they flattered us. Directional models measured no out-of-sample edge after costs. Fast intraday trading ran an expectancy of -1.09%. Funding-rate carry is the one thing that ever came back positive: +4.7% a year after costs, robust to every parameter we varied. This guide explains why it works, and then spends most of its length on the more useful question — whether it is worth doing today. Our own data says it mostly is not, and that is the part other write-ups of this trade leave out.

Funding carry calculatorPut your own funding rate, holding period and fee tier in, and see whether the position clears its round trip.

What the trade actually is

Buy spot and short the perpetual future in the same size. The two legs offset, so the position does not care where price goes. What remains is the funding payment, which perpetual shorts collect from longs whenever funding is positive — and on crypto perps it usually is, because leveraged longs persistently outnumber leveraged shorts. This is qualitatively different from every other strategy we run. It forecasts nothing. You are being paid to absorb a leverage imbalance, and the payment is observable in advance rather than estimated.

Why we trust this result more than our directional work

Two reasons. First, it is structural rather than statistical: there is a reason someone is paying you, and you can read the price of that service on the exchange before you take the trade. Second, it survived every robustness check we threw at it — it stayed positive at every entry threshold and every fee tier we tested, whereas our directional strategies flipped sign fold to fold. We also re-derived the headline figure independently for this article, pulling all 7,000 funding prints Binance has published for BTCUSDT since April 2020. Funding was positive in 86.0% of them. Our internal study, run separately, reported 86%. When a number reproduces from a different direction, it is usually real.

// BTCUSDT 8h funding, Binance /fapi/v1/fundingRate
// 7,000 periods, 2020-04-04 -> 2026-08-24 (6.39 years)
//
// mean funding      0.01040% per 8h  = 11.38%/yr gross
// positive periods  86.0%
// periods clearing our 0.01%/8h entry bar: 47.2%
//
// NOTE: 11.38%/yr is GROSS funding collected if you were always in the
// position. Our measured strategy return is +4.7%/yr, because costs and
// being out of the market when funding is thin take the rest.

The costs are quadruple, and they decide everything

Entering means a spot buy and a perp sell. Exiting means both again. Four taker legs, which at our modelled fee tier comes to roughly 0.38% of notional round trip. At a funding rate of 0.01% per 8h you collect 0.03% a day, so it takes 12.7 days just to pay for the trade. Everything after that is profit and everything before it is not. This is why the holding period, not the headline funding rate, is what actually determines whether a carry trade makes money — and why acting on a single high funding print is usually a mistake. Our model enters on a three-day trailing mean rather than the spot value, because one 8-hour print is noisy, mean-reverts, and will churn you through that 0.38% repeatedly.

Round trip = 2 x (spot_fee + perp_fee + 2 x slippage)
            = 2 x (0.100% + 0.050% + 0.020%)
            = 0.38% of notional

Breakeven at 0.01% per 8h:
  0.38% / 0.01% = 38 funding periods = 12.7 days

What has changed: the edge has been compressing for years

This is the part worth your attention. Carry has not stopped working in the sense of breaking — it has been paying less every year as the market has become better supplied with people willing to short perps. Broken out by calendar year, the gross annualised funding rate on BTC and the share of periods that cleared our own 0.01% entry threshold both fall away sharply. 2021 was an extraordinary year that a lot of carry write-ups are still implicitly quoting. 2026 is not that.

Year   Gross annualised   Positive   Cleared 0.01%/8h entry bar
2020         15.44%/yr        87.0%        77.8%
2021         30.61%/yr        92.7%        86.0%
2022          4.16%/yr        77.9%        30.8%
2023          7.87%/yr        89.9%        44.8%
2024         11.92%/yr        91.6%        61.5%
2025          5.13%/yr        87.1%        17.2%
2026          2.43%/yr        70.6%         5.5%   (to 24 Aug)

In 2026, BTC funding has never once exceeded the base rate

Binance's funding formula carries a default interest component of 0.01% per 8 hours, and the premium is what pushes funding above it when longs are crowded. Across all 708 funding prints so far in 2026, BTC funding has exceeded that base rate exactly zero times. It sat precisely at it in 5.5% of prints, came in below it in 94.5%, and went outright negative — meaning the carry position pays rather than collects — in 29.4%. The leverage imbalance that this trade exists to be paid for has, for the moment, essentially disappeared on BTC. That is not a claim about where price is going. It is a statement about what the trade currently earns, and it is checkable in about thirty seconds.

BTCUSDT funding, 2026 year to date (708 prints)

  above 0.01% per 8h ....    0 prints    0.0%
  exactly 0.01% ........    39 prints    5.5%
  below 0.01% .........    669 prints   94.5%
  of which negative ...    208 prints   29.4%

  latest print: 2026-08-24 16:00 UTC = 0.010000% per 8h

The risk that actually liquidates people is margin, not direction

Carry is delta-neutral in profit and loss, which is not the same as being safe. The spot long and the perp short sit in different accounts, and only one of them can go against you on paper. Over holds of 31 to 98 days we measured adverse excursions on the short perp leg of +116% on BTC and up to +400% on SOL. The spot leg offsets that in economic terms, but if the perp account is margined on its own at 1x, it is liquidated long before the offset is realised — and you are left short the loss and long a position that no longer hedges anything. Portfolio or cross margin, where the spot long collateralises the perp short, is a prerequisite for running this trade at all. Anyone describing carry as low-risk without mentioning that has not held one through a rally.

Running it across many coins does not diversify it

The intuitive next step is to run carry on ten assets instead of one. We tested that and it does not do what people expect. Funding is largely a single market-wide factor — pairwise correlation across symbols measured 0.54 — so when it dries up on BTC it is usually drying up everywhere at once, which is precisely what the 2026 figures above show. Multi-symbol carry is insurance against picking the wrong symbol, not a way to increase the return. Treat it as symbol selection, not diversification.

How to check whether it is worth doing today

You do not need us for this, and you should not take our word for it. Pull the funding history for the pair you care about, take the mean of the last nine prints (three days), and compare it to your own round-trip cost divided by the number of periods you expect to hold. If the trailing mean does not clear that bar with room to spare, the trade is paying you less than it costs. As of the latest print in this guide, BTC funding sits exactly at 0.01% per 8h — which clears our entry threshold by precisely nothing, and is the default rate rather than evidence of crowded longs. That is the honest read: not dead, not dormant, but not currently paying enough to be interesting either.

curl -s 'https://fapi.binance.com/fapi/v1/fundingRate?symbol=BTCUSDT&limit=9' \
  | python -c "import sys,json; r=json.load(sys.stdin); \
    m=sum(float(x['fundingRate']) for x in r)/len(r); \
    print(f'3-day mean: {m*100:.5f}% per 8h = {m*3*365*100:.2f}%/yr gross')"

Summary

Funding carry is real, and on our measurements it is the only positive-expectancy strategy in this codebase: +4.7% a year after costs, underwritten by funding being positive in 86% of eight-hour periods over six years. It is also, right now, paying close to nothing. BTC funding has not exceeded its base rate once in 2026 and has been negative in nearly a third of periods, while the round trip still costs 0.38% and still needs twelve days of carry to clear. The strategy has not broken; the market has stopped paying for it as generously as it used to, and it may well pay again. We would rather tell you that than sell you a backtest that ends in 2021.

Is Funding-Rate Carry Still Profitable? We Measured Six Years of It