Funding rates explained

Why does a perpetual need a funding rate at all?

A dated future converges to spot by construction. As expiry approaches, the two prices must meet, because on the last day the contract settles against the index — and arbitrageurs enforce it long before then. That convergence is what stops a future from drifting away from the thing it represents.

A perpetual has no expiry, so it has no such anchor. Left alone, a contract in permanent demand would simply trade above spot forever, and it would stop being a proxy for the asset. Funding is the replacement mechanism: instead of forcing convergence at a date, it makes divergence expensive continuously. If the perpetual trades above its index, holders of long positions pay holders of short positions; if it trades below, the payment reverses.

Nobody has to act on that for it to work, though in practice people do. A persistent premium creates a paid trade — sell the perpetual, buy the spot, collect the payment — and the arrival of that trade is what pushes the perpetual back toward its index. Funding does not force the price anywhere; it pays someone else to.

Who pays whom, and when?

The sign convention is universal: a positive funding rate means longs pay shorts, a negative rate means shorts pay longs. The venue is usually not a party to the transfer — it moves between position holders — which is why funding is not a fee in the ordinary sense.

The payment is proportional to position notional, not to the margin behind it. This is the detail that most changes the arithmetic in practice. A trader with $10,000 of collateral holding a $100,000 position pays ten times the funding, in dollars, of a trader holding $10,000 unleveraged — and it lands against a much smaller capital base. Leverage multiplies the cost of holding as reliably as it multiplies the profit and loss.

Timing is venue-specific and matters more than it looks. The common arrangement is a snapshot: funding is exchanged at fixed timestamps, and only positions open at that instant pay or receive. Closing a minute before the timestamp avoids the payment entirely, which is why open interest sometimes dips just before settlement and rebuilds just after. Some venues instead accrue funding continuously, so holding time is what counts and there is nothing to dodge. Both designs are defensible; assuming the wrong one costs you real money.

Note Funding is realised cash, not unrealised mark-to-market. It hits your balance whether the position is up or down, and on a long-held leveraged position it compounds into a substantial share of the outcome.

Where does the number come from?

Nearly every venue builds the rate from two components. The first is a premium term: how far the perpetual’s price sits from its index, averaged over the interval. It is not a single reading at the end — it is typically sampled many times and averaged, often using an impact price built from the actual bid and ask at a realistic size rather than the mid, so that a thin quote cannot move the rate.

The second is an interest-rate term, representing the cost differential between holding the quote currency and holding the base asset. On most crypto venues it is a small fixed constant, and it is why funding tends to sit slightly positive in a neutral market rather than at exactly zero.

The two are combined, and the result is clamped: every venue caps how large a single funding payment can be, and some tighten the interval or widen the cap under stress. The exact formula, the sampling scheme, the constant and the cap are all published in the contract specification, and they differ between venues and sometimes between contracts on the same venue. Treat the spec as the source of truth and this paragraph as a map of what to look for.

One consequence worth internalising: the rate you see before settlement is an estimate that is still moving. Predicted funding converges to realised funding as the averaging window fills, and a spike in the last minutes of the window moves the final number far less than it moves the instantaneous premium.

What does funding tell you about positioning?

Read literally, funding tells you what leveraged exposure costs right now and who is paying. Persistently positive funding means the marginal participant is willing to pay to hold long exposure with borrowed money. That is a statement about crowding — about who is already positioned — and it is not a statement about what happens next.

It becomes considerably more informative when read next to open interest, because the pair distinguishes cases that funding alone cannot:

Two refinements make the read sharper. First, compare across venues: when one venue’s funding is far above the rest, it is that venue’s participants who are leaning, not the market — and the gap itself is an arbitrage that tends to close. Second, compare perpetual funding with the basis on dated futures. Funding is the leveraged, retail-accessible expression of the same premium; a market where dated basis is calm while perpetual funding is extreme is one where the leaning is concentrated in leveraged hands.

What funding cannot do is time anything. Rates can stay expensive for weeks in a strong trend, and paying to be long is perfectly rational when the trend pays more than the funding does. Treating an elevated rate as a short signal is the most common way to lose money with an otherwise good instrument.

If funding is positive, why doesn’t everyone just collect it?

They try, and the trying is what compresses the rate. The trade is cash-and-carry: buy the spot asset, sell the perpetual against it in equal size, and collect funding while holding no net directional exposure. It is the mechanical link between funding and the real cost of capital, and it is the reason extreme rates are usually temporary.

It is also not free money, for reasons that are entirely practical:

• The short leg needs margin, and a sharp rally consumes it. A carry trade that is delta-neutral on paper is still liquidatable in practice if the margin sits in the wrong place — see liquidations and cascades.
• Funding can flip. The position that was collecting becomes the position that pays, and it does so precisely when the market has turned against the crowd you were fading.
• The legs are not on the same ledger. Spot and perpetual can be on different venues with different collateral, and moving margin between them takes time you may not have.
• You inherit venue risk on both sides, for a return that is measured in percent per year rather than percent per day.
• It is crowded. When the trade is obviously on, the rate is already compressed toward the cost of capital, which is exactly the equilibrium the mechanism is designed to produce.

The useful takeaway is not the trade, it is the intuition: funding tends to mean-revert toward the cost of capital, and departures from it measure how much leveraged demand exceeds the arbitrage capital available to absorb it.

How do extreme funding regimes unwind?

There are exactly two exits, and it is worth knowing which one you are watching.

The slow exit is that funding does the work. Carry traders arrive, the premium compresses, positions that were paying to stay long find the cost no longer justified, and the rate drifts back toward its baseline with nothing dramatic happening to price. This is the common case and it is undramatic by construction.

The fast exit is that price does the work. Crowded leveraged positioning is, by definition, a large stock of positions with margin behind them; a move against that crowd converts holders into forced sellers, and forced selling into further moves. Funding does not cause this — it identifies the fuel. The mechanism that ignites it is in the margin engine, and it is covered in liquidations and cascades.

So the honest formulation of what extreme funding tells you is this: it does not raise the probability that the market turns. It raises the cost of being on the crowded side, and it raises the magnitude of the move if the market does turn. Those are statements about payoff, not about timing, and they are the ones you can actually use.

Why do intervals differ between venues, and what breaks if you ignore it?

An eight-hour interval — three settlements a day — is the most common arrangement across major venues. It is not universal: hourly settlement exists and is used by some venues, and several venues shorten the interval or adjust the cap for specific contracts or during stress. The interval is a contract parameter, published in the spec, and it changes from time to time.

The practical consequence is that a raw funding number is not a comparable quantity. A rate quoted per 8-hour period and a rate quoted per hour are different units, and the hourly one is necessarily a smaller number for the same annualised cost. Comparing them side by side produces a conclusion that is exactly backwards.

The interval also changes what a spike means. A venue settling hourly reacts to a premium sooner and in smaller increments, so its rate is jumpier and its extremes are shallower in raw terms. A venue settling every eight hours averages over a longer window, so a violent hour shows up muted. Two venues can disagree about “how extreme funding is” purely because of their sampling, with no disagreement whatsoever about the market.

The rule that avoids all of this: convert everything to the same period — per day or annualised — before you look at it, and label the axis. Every comparison, every percentile, every historical range should be computed on the normalised series. The funding cost calculator does that conversion, and applies it to a position size so the answer arrives in currency rather than in basis points.

What are the common mistakes?

• Comparing raw rates across venues with different intervals. The single most frequent error, and it inverts the conclusion.
• Reading funding as direction. It measures the price of leverage and who is paying it, not where price goes next.
• Forgetting it is charged on notional. At high leverage, funding can consume a meaningful share of margin over a week of flat price action.
• Taking predicted funding as final. It is a running average until the window closes.
• Reading one venue as the market. Funding is venue-local by construction, and divergences between venues are themselves the interesting observation.
• Ignoring the snapshot. On snapshot venues, whether you were open at the timestamp is the entire question; average exposure over the interval is irrelevant.

Funding is at its most useful as one of three positioning inputs, alongside open interest and the state of the book. Each covers what the others cannot: the book shows the immediate cost of trading, open interest shows how much exposure exists, and funding shows what that exposure is willing to pay. Read together, they describe a market’s posture; read alone, each is a way to be confidently wrong.

Questions fréquentes

Does the exchange keep the funding payment?

On the standard design, no — funding moves between long and short position holders, and the venue is not a party to it. Trading fees are separate and are the venue’s revenue. Check the contract spec, because designs vary.

Can I avoid paying funding by closing just before the timestamp?

On a snapshot venue, yes: only positions open at the settlement instant exchange funding. On a venue that accrues continuously, no. This is a per-venue mechanic, and it is one of the first things to check.

Is negative funding bullish?

It means shorts are paying longs, which tells you leveraged positioning is leaning short and that being long is currently subsidised. Whether that resolves upward depends on everything else. It is a crowding measurement, not a forecast.