Designing Funding Rates for FX Perpetual Futures
What we learned building FX perpetuals that account for interest rate differentials
We describe an interest-rate-inclusive funding formula for FX perpetual futures that makes them behave as synthetic FX forwards, bringing the carry trade onchain. The design follows from a simple observation: offchain, FX forwards embed the interest rate differential between two currencies in their price. A dateless contract cannot, and must instead express that differential continuously through funding. We discuss the structural consequences for trader economics, liquidity, and product design.
Notation and conventions
Throughout the paper we use FX market conventions. For a pair written BASE/QUOTE (e.g. USD/JPY), the base is the asset being priced and the quote is the numeraire in which the price is expressed. USD/JPY = 159 means one USD costs 159 JPY. We write \(r_{\text{base}}\) and \(r_{\text{quote}}\) for the risk-free rates in the two currencies, and \(x_t\) for the spot rate.
Our platform settles all positions in USDC, which we treat as equivalent to USD for pricing purposes. The contract type therefore depends on which side of the pair USD sits on. Following Ackerer, Hugonnier and Jermann (2024), we call a contract linear when the margin currency coincides with the quote and inverse when it coincides with the base. For pairs where USD is the quote (EUR/USD, GBP/USD, AUD/USD) the contract is linear; for pairs where USD is the base (USD/JPY, USD/CHF, USD/CAD) the contract is inverse. Cross pairs without a USD leg (EUR/JPY, GBP/CHF) are formally quanto contracts and lie outside the scope of this paper as we will not list them at launch.
We use USD/JPY as the running example throughout.
Perpetual futures and the funding mechanism
Perpetual futures account for over 90% of crypto derivatives volume (TD Securities, 2025). Their core innovation is simple: replace the expiration date with a funding rate that continuously nudges the contract price toward spot.
The logic is straightforward. When the perp trades above spot, longs pay shorts. When it trades below, shorts pay longs. Traders are incentivized to close the gap, and the contract stays anchored. No rolls, no expiry.
Three families of funding formula are commonly used across existing protocols:
(i) Continuous. The funding rate of change is proportional to skew, and the rate itself is the integral over time:
(ii) Discrete. A weighted sum of the price premium and the open interest skew, sampled at intervals (typically every 1 or 8 hours):
(iii) Direct skew. The minimal version. Funding is proportional to skew:
These three formulas work well for BTC, ETH, and other crypto-native assets, but the asset class is a very particular one. There is no observable, arbitrage-enforceable risk-free rate in either leg of a BTC/USD pair. Central banks do not set a "BTC risk-free rate", and DeFi lending rates exist but are endogenous to the same speculative flows that drive the perp itself, so plugging them into funding would be circular. Perp protocols therefore omit the interest rate differential from funding and let the premium and skew terms absorb everything. FX is the opposite situation: both rates are exogenous and enforced by real-world money-market arbitrage, which changes the design problem entirely.
The FX problem: interest rates exist
In traditional finance, FX forwards embed the interest rate differential between two currencies via covered interest parity (continuous-time form):
A 1-year USD/JPY forward is not just the spot rate; it folds in the roughly 3-5% rate gap between the Fed and the BoJ. Arbitrage anchors this, imperfectly: a mispriced forward lets an arbitrageur borrow in the cheap currency, lend in the expensive one, and lock in the forward to hedge. Since 2008, dealer balance-sheet frictions have left a persistent residual, a "CIP basis" of order 20 to 30 bps in G10 at benchmark tenors, but the basis is small relative to the underlying differential, which remains the first-order driver of forward pricing.
A perpetual future has no expiry, so the differential cannot be embedded in a maturity-based price. It has to be expressed continuously through funding. Ackerer, Hugonnier and Jermann (2024) formalize this: for a linear contract on currency pair base/quote (quoted, margined, and settled in the quote currency), the funding payment per period on a long position is:
where \(x_t\) is spot, \(f_t\) is the perp price, \(\kappa_t > 0\) is the premium rate, and \(\iota_t\) is the interest factor. The only choice of \(\iota_t\) that keeps the perp price anchored to spot at all times is:
For inverse contracts, AHJ's Corollary 2 gives \(\iota_{\text{AHJ-I},t} = (r_{\text{base},t} - r_{\text{quote},t}) / (1 + r_{\text{quote},t})\) for a long with PnL \((1/f_{I,t+1} - 1/f_{I,t})\), the side opposite our standard-convention long. The no-arbitrage anchor is the same; the equalizing \(\iota\) on our long is then the negative:
Perps as synthetic forwards: the carry trade onchain
An IR-inclusive perp is a synthetic, continuously-rolled FX forward. It replicates the exposure of a rolled forward strategy. This is what enables the FX carry trade onchain. The carry trade, long the high-yielder and short the low-yielder to collect the differential, is one of the most studied and most persistent phenomena in international finance. It works in FX forwards because forwards price in the differential. It cannot work in a perp that zeroes out the interest component, because the funding rate is then driven entirely by speculative imbalance and bears no reliable relationship to the rate differential. A carry strategy cannot be built on an instrument that does not price carry.
An IR-inclusive perp fixes this. A trader can hold a USD/JPY long for six months, continuously collecting the Fed-BoJ differential via funding, exactly as they would with a rolled sequence of one-month forwards, without rolls or maturity management.
Onchain forwards are poorly suited to onchain liquidity structures. Each maturity needs its own order book and its own liquidity providers; onchain FX is a nascent market, and splitting thin liquidity across four or five maturity buckets would give every tenor unusable depth. Passive liquidity providers also need risk characteristics they can price. Perps offer stationary inventory risk. Dated contracts do not work this way, their price rolls down deterministically toward spot as expiry approaches, and the optimal quote moves with time, so the risk and reward of the position shift across its life. Traditional FX forwards settle by physical delivery on the value date at the outright price agreed at trade time, not against a benchmark. A cash-settled onchain analogue still needs a credible end-of-day mark for each tenor, which is a non-trivial oracle problem distinct from running a continuous spot feed. Finally, every expiry forces traders to close and reopen positions, incurring gas, slippage, and position discontinuities that break automated hedging strategies.
Perps sidestep all of this: one book, stationary risk, no rollover friction, and the carry is expressed continuously through funding rather than embedded in a settlement price.
Our design
Our funding formula layers the interest factor on top of the standard premium and skew terms. For linear contracts:
For inverse contracts, the premium and imbalance terms are unchanged; the interest component takes the form \(\iota_{I,t} = (r_{\text{quote}} - r_{\text{base}}) / (1 + r_{\text{quote}})\) instead.
Longs pay shorts when \(r_{\text{fund}} > 0\); in the USD/JPY example with the Fed above the BoJ, the interest component flows from shorts to longs.
The premium and skew terms are the standard perpetual-contract machinery: \(\alpha\) controls how aggressively funding reacts to perp-spot price gaps, and \(\beta\) controls how aggressively it reacts to open-interest imbalance. These are the balancing mechanisms. The interest component is something else: it is a genuine economic cost (or payoff) that exists offchain regardless of what any protocol does. More on that distinction below.
Position-level cost of carry
Consider USD/JPY at a 3% annualized rate differential (Fed at 3.75%, BoJ at 0.75%), about the midpoint of the 2023-2025 range. Converted to a per-day cost:
A perpetual is a leveraged product, however, and the funding charge is computed on notional, not equity. On equity, the cost scales linearly with leverage:
| Leverage | 1 day | 1 week | 1 month |
|---|---|---|---|
| 1x | 0.008% | 0.058% | 0.247% |
| 5x | 0.041% | 0.288% | 1.233% |
| 10x | 0.082% | 0.575% | 2.466% |
| 25x | 0.205% | 1.438% | 6.164% |
Carry is noise for a scalper but meaningful for anyone holding a leveraged directional view across days or weeks. The asymmetry is part of the design: the product is materially more attractive for carry-positive directional trades, where funding accrues to the trader over time, and materially less attractive for carry-negative ones.
Structural asymmetry and the competitive question
IR-inclusive funding creates a persistent directional bias that pure-crypto perps do not have. Every hour, longs on USD/JPY receive roughly 3% annualized, shorts pay it, independent of market conditions.
A natural objection is that traders could defect to a venue that does not price the rate differential. Onchain FX is a fairly new market, and venues handle interest rates in different ways. Our approach, described in the previous sections, makes the rate differential explicit in the funding formula via the interest factor. Others bundle the differential into a rollover-style charge derived from underlying financing rates, closer to a traditional FX-broker swap mechanic. Others omit it entirely, using funding driven only by open-interest imbalance and volatility. A trader on the third kind of venue does indeed pay no carry.
But the rate differential does not vanish because a protocol chooses not to price it. In an efficient market, arbitrageurs would long the IR-inclusive perp to collect carry, short the non-IR perp to hedge, and push the non-IR perp's funding to reflect the differential anyway. The cost shows up either way, explicitly in the formula or implicitly through price deviations and unpredictable funding spikes. In practice this arbitrage is incomplete: cross-venue capital deployment, bridging risk, and borrowing costs eat into the spread, so non-IR FX perps can misprice for extended periods.
The structural risk on an IR-inclusive venue is that the carry pulls flow in one direction by design. Carry seekers and hedgers whose underlying exposure aligns with the favorable side cluster there, while the unfavorable side draws a thinner pool, populated mainly by hedgers running the other way. The premium term has to widen funding on that side to attract counterparties, and the book can drift toward a lopsided equilibrium where one direction is well-priced and the other is a carry farm with no counterparties.
Managing this asymmetry is an open problem. But the same cross-venue arbitrage pressure that forces non-IR venues to reflect the differential also creates natural demand for the unfavorable side on IR-inclusive venues: arbitrageurs need somewhere to put the unfavorable leg, so the problem may be bounded in practice rather than eliminated in principle.
Who benefits from IR-inclusive funding
Speculators holding intraday positions barely notice 3% annualized carry. Carry traders and hedgers, on the other hand, cannot use a non-IR perp for their purpose: without the interest component, neither the carry exposure nor the hedge reliably prices correctly over their holding period. Market makers care less about the interest component itself (it nets out in their book) and more about whether funding is predictable, which is actually a point in favor of IR-inclusive design: the interest component is a known, slow-moving input, which makes residual funding less volatile than it would be if the full rate differential had to be absorbed by the premium term.
Carry traders and hedgers bring two-sided natural flow. A Japanese exporter hedging USD receivables by shorting USD/JPY is a structural counterparty for a carry seeker going long. Matching that flow is exactly what market makers need to run tight books, and it reduces the toxicity of order flow compared to a pure-speculator venue.
Future work
We chose IR-inclusive funding to support the use cases described above: carry trading, hedging of real-world FX exposure, and the two-sided natural flow these users bring. The alternative either breaks those use cases entirely or forces the differential to show up in different ways.
This paper does not address several design questions that deserve separate treatment: weekend, holiday, and off-hours handling when reference markets are closed; oracle design and fallback rules for interest-rate feeds and the asymmetric-liquidity problem on the unfavorable side of the carry. We will write about each of these separately as the design matures and deployment data becomes available.
References
- Damien Ackerer, Julien Hugonnier, and Urban Jermann, "Perpetual Futures Pricing," NBER Working Paper No. 32936, September 2024. Ackerer, Hugonnier, and Jermann derive no-arbitrage pricing formulas for perpetual futures contracts, including linear, inverse, and quanto futures.
- Wenxin Du, Alexander Tepper, and Adrien Verdelhan, "Deviations from Covered Interest Rate Parity," The Journal of Finance 73, no. 3 (June 2018): 915-957. Du, Tepper, and Verdelhan document large, persistent, and systematic deviations from covered interest rate parity and link them partly to balance-sheet and regulatory frictions.
- Reid Noch, Peter Haynes, and Scott Baker, "Perpetual Futures: The Missing Link in Tokenized Equities," TD Securities, November 6, 2025. Noch, Haynes, and Baker explain how perpetual futures could enable continuous, leveraged trading in tokenized equities while raising liquidity, leverage, and regulatory concerns.