The Smallest Funding Rate on the Board Was the Most Expensive.

Three exchanges quoted the same funding rate on the same perpetual this afternoon: 0.0100%. A fourth quoted 0.0073% — a smaller number, on the same contract, at the same minute.

The smaller number cost roughly six times more.

Nothing is wrong with any of those quotes. They are all correct, and they are not comparable, because a funding rate without its settlement interval is an incomplete figure. This is the most common way a trader loses money to an instrument they understood perfectly well.

What the rate actually is

Funding is a payment between the two sides of a perpetual, not a fee the venue collects. When the rate is positive, longs pay shorts. When it is negative, shorts pay longs. It exists to tether a contract with no expiry to the spot price it is supposed to track.

The number a venue publishes is the rate for one settlement. How often that settlement happens is a venue's own design decision, and the venues do not agree:

A rate of 0.0100% every eight hours is paid three times a day. The same 0.0100% every hour is paid twenty-four times a day. The published numbers are identical; the cost differs eightfold. Comparing them as printed is not a small imprecision — it is reading the wrong quantity.

The fix is one multiplication. Annualise:

rate × (24 ÷ interval hours) × 365

Everything below uses that figure, and only that figure.

A live reading

Injective perpetuals, captured 23 September 2026 at 14:09 UTC:

Hyperliquid's rate is the lowest number in the column and the highest cost on the board. A long holding INJ there paid at an annual rate of 64.3% while a long on Binance paid 11.0% — a gap of 53 annualised points, on the same asset, at the same moment.

On a $10,000 position that difference is about $440 a month. It is not a rounding error and it is not a rare configuration. It is the ordinary consequence of one venue settling twenty-four times a day while the others settle three.

Most of the time, the venues are not disagreeing at all

Look again at the first three rows. Binance, Bybit and OKX printed the identical rate — not similar, identical.

That is not a coincidence, and noticing why it happens is worth more than the table. Each of those venues applies a default rate when the contract's premium sits inside a normal band. The default annualises to the same place regardless of interval: 0.0100% every eight hours and 0.0050% every four hours are both 11.0% a year. So on a quiet instrument the major venues converge on one number because they are all quoting the same baseline, not because they have independently reached the same conclusion about positioning.

Which means the spread you see is usually not three venues arguing. It is one venue using a different mechanism.

When they genuinely disagree, they disagree about direction

Across 28 perpetuals with at least two comparable venues, captured in the same pass:

A sign disagreement is the interesting case, because it is not a matter of degree. On WIF the same afternoon, Binance, Bybit and OKX each annualised to +11.0% — longs paying — while Hyperliquid annualised to −42.2%, shorts paying. A long was being charged on three venues and paid on the fourth, simultaneously, for holding the same exposure.

Nine of twenty-eight is roughly a third of the board. Whatever else funding is measuring, it is not measuring one thing that every venue observes equally.

And the effect is not confined to thin altcoins. Bitcoin, the most liquid perpetual there is, spread 8.3 points across venues in the same capture: Binance 2.6%, Bybit 6.9%, OKX 11.0%, Hyperliquid 11.0%, against a combined $18.2 billion of open interest.

The trade this suggests, and the reasons it is harder than it looks

The obvious idea is to be long where funding pays you and short where it charges you, collect the difference, and carry no directional risk. It is a real strategy with a real name — cash-and-carry, or funding arbitrage — and the numbers above are genuinely the raw material for it.

It is also where a clean table turns into an expensive lesson. Four things sit between the spread and the profit:

The rate is not a rate you are promised. It is the rate for the next settlement. Funding is recalculated continuously and the gap that justified the position can close before the first payment, or invert. Annualising a single print tells you what today would cost for a year; it does not tell you what a year will cost.

Both legs cost money to open and to close. Taker fees around 0.055% a side on each of two venues is roughly 0.22% to enter and exit a round trip. Against a 10-point annualised gap, that is about forty days of carry spent before the first cent is yours.

Two positions on two venues is two margin accounts. A move against the short leg can liquidate it while the long leg sits comfortably on another exchange. Delta-neutral on paper is not delta-neutral through a liquidation, and capital cannot be moved between venues at the speed the move happens.

The widest spreads sit on the thinnest instruments. WIF's $46 million of combined open interest and INJ's $80 million are not BTC's $18 billion. The spread is wide partly because the book is thin, and a thin book is where slippage and gaps live.

None of that makes the strategy wrong. It makes the spread a starting point for arithmetic rather than a conclusion.

How to read the comparison honestly

Three habits cover most of it.

Annualise before comparing, always. A venue's published rate is only meaningful next to its own interval. Any table that puts raw rates side by side is comparing different units and hiding it.

Ask which venue is the outlier and why. If three venues sit on the default and one does not, the question is about that one venue's mechanism and its book, not about a market-wide view of positioning.

Weight by open interest. A 64% annualised rate on $6 million says less about the asset than an 11% rate on $34 million. The Basis derivatives panel prints an open-interest-weighted rate beside the per-venue rows for this reason: the average that matters is the one money is actually paying.

Funding is one of the few numbers in crypto that is directly observable, settled in cash, and published by everyone. That makes it unusually trustworthy — right up to the moment it is read without its interval attached, at which point it becomes one of the easiest numbers on a screen to be confidently wrong about.

Figures captured from Binance, Bybit, OKX and Hyperliquid public endpoints on 23 September 2026 at 14:09 UTC. Funding changes every settlement; the cross-venue comparison in the Basis derivatives panel is live.