RESEARCH

Crypto funding rates: 38% in 2021, 1.3% so far in 2026

Study details

Measured
2020 - 2026, 777 monthly funding files covering ten Binance USDT-margined perpetual contracts; 7,305 funding settlements on the BTC contract alone
Instruments
BTC, ETH, SOL, XRP, BNB, DOGE, ADA, LINK, LTC and AVAX perpetual futures; funding settles every 8 hours
Method
for each contract, sum the published funding rate across every settlement in a calendar year, then take the median across the ten contracts. Separately, count what share of settlements were negative. Nothing is excluded, and 2026 covers eight months and is not annualised
Result
the median contract paid 38.0% in 2021 and 1.3% across the first eight months of 2026. On BTC the figure fell from 30.6% to 1.7%, a factor of 18.0. Meanwhile the share of settlements that pay the other way rose to a median of 34.8%
Contents
Funding is the clearest thing in crypto to measure and one of the least often measured over a long window. Every eight hours a perpetual futures contract settles a payment between longs and shorts, the exchange publishes the rate, and the record of those rates goes back to the beginning. If you have ever seen a delta-neutral yield quoted, that number is where it came from. So we added it up. Ten contracts, 2020 through August 2026, 777 monthly files, nothing excluded.

1. What the number actually is

When funding is positive, longs pay shorts. A position that is long the coin and short the perpetual against it collects that payment and does not care which way the price goes. Summing the published rate across a year gives what that position would have received, before anything else. That last phrase is carrying weight, and section 5 is about how much.

2. Year by year, the median contract

YearMedian fundingContracts that were positiveMedian share of settlements that were negative
202014.1%8 of 1014.9%
202138.0%10 of 106.6%
20220.4%5 of 1036.9%
20238.1%9 of 1013.1%
202413.6%9 of 109.1%
20254.2%9 of 1022.7%
20261.3%8 of 1034.8%
The last row is eight months, not twelve, and is not scaled up. Two things move together down that table. The amount shrinks, and the share of settlements running the other way grows. In 2021 the median contract paid something on roughly nineteen settlements in twenty. In 2026 it is closer to two in three.

3. The same story on the largest contracts

YearBTCETH
202017.2%27.5%
202130.6%37.5%
20224.2%0.8%
20237.9%8.3%
202412.0%13.0%
20255.1%4.9%
20261.7%1.0%
BTC in 2021 was 18.0 times what BTC has paid across the first eight months of 2026. On ETH, the share of settlements that were negative was 4.1% in 2021 and is 32.8% in 2026. Neither of those is a crash. It is a compression, and it has been running for years with one clear interruption upward in 2024.

4. What a bad year looks like

The table above makes 2022 look merely flat at 0.4%. It was not flat, it was two-sided. That year half the contracts finished negative, meaning a position holding spot against a short perpetual paid for the privilege all year rather than being paid. SOL paid out 38.0% over 2022. BNB was negative in four consecutive years, 2022 through 2025. This matters more than the headline, because a carry position is usually sized as though the payment is the safe part. On these contracts the payment itself reversed for a year at a time. Short-term, 2026 is the second-most two-sided year in the sample: a median of 34.8% of settlements went the wrong way, and on XRP it was 51.3% and on SOL 50.5%. More than half of the settlements on those two contracts were paid by the carry position rather than to it.

5. This is gross, and gross is the smaller problem

Everything above is the published funding rate and nothing else. It does not include exchange fees on either leg, the spread paid getting in and out, the cost of the dollars funding the spot side, or the basis between spot and the perpetual at entry and exit. It also assumes the position is simply held, with no rebalancing of the hedge as the price moves. Any real version of this pays all of those. That is why the correct way to read the table is as a ceiling: the most that was available before costs, not what anyone received. A ceiling of 38.0% leaves room for a lot of friction. A ceiling of 1.3% does not, and that is the whole point of the measurement. The number did not merely get smaller; it got smaller than the costs it has to clear, which turns a trade into a subscription to those costs.

6. Why it compressed, which we did not test

The obvious explanation is that more capital arrived to collect it. Funding is a price, and it is high when there is more leveraged demand for long exposure than there is balance sheet willing to take the other side. Products that systematically take the other side have grown a great deal since 2021. That is a plausible mechanism and we have not tested it. Our measurement establishes the compression; it says nothing about the cause. If you see this chart attached to a confident causal story, the story is someone's opinion attached to our kind of number. Limitations Ten contracts is a small and deliberately chosen sample, and all ten exist today. Contracts that were delisted are not in it, which biases the picture toward survivors. The effect on funding specifically is unclear rather than obviously favourable, but it is a selection and it should be stated. One exchange, one margin type. These are USDT-margined perpetuals on a single venue. Coin-margined contracts, other venues and other quote currencies can and do differ. Funding settles every 8 hours on these contracts for almost the whole window. The SOL contract in 2022 shows a mixture of intervals in the published record, so its settlement counts for that year are not directly comparable to the others; the summed rate is unaffected. Summing the rate is not compounding a return. It answers how much funding accrued, which is the question people mean, but it is not a portfolio simulation and should not be compared directly to a compounded yield. The 2026 figure covers eight months. It is presented as an eight-month sum throughout and is never annualised, because annualising a partial year in a compressing series would overstate it.

Related reading

← All research

Originally published by TraderWe on September 26, 2026. You may quote and link to this page. Republishing the full text without a link back to the original is not permitted.

0 replies

Sign in to reply →