The setup: what funding arb promises
Cross-exchange funding arbitrage is conceptually simple: go long on the exchange where you pay the lower rate, short on the exchange where you receive the higher rate — for the same asset. If you hold both legs, you're delta neutral (price direction doesn't matter) and you collect the net spread every 8 hours.
On paper, the maths look compelling. When PEPE funding was 0.30% on Binance and 0.08% on MEXC, the spread was 0.22% per 8h — or 240% annualised. A $5,000 notional position on each leg would collect $11 every 8 hours.
In practice, we found that the vast majority of those opportunities were fully consumed by the actual costs of execution.
The real cost structure
Before any profit calculation, every funding arb trade carries these mandatory costs:
| Cost item | Taker-taker | Maker-maker | Notes |
|---|---|---|---|
| Entry fees (both legs) | 0.08–0.12% | 0.00–0.04% | Taker ~0.04–0.06% per leg |
| Exit fees (both legs) | 0.08–0.12% | 0.00–0.04% | Same as entry |
| Bid-ask spread slippage | 0.02–0.15% | 0.01–0.05% | Wider for low-liquidity alts |
| Capital opportunity cost | ~0.01%/8h | ~0.01%/8h | Margin locked, earning ~4.5%/yr elsewhere |
| Total round-trip cost | 0.18–0.39% | 0.01–0.13% |
For a taker-taker trade, the break-even spread is roughly 0.18–0.39% of notional — which needs to be earned back from the funding spread before the trade closes. On an 8-hour funding schedule, you need the spread to persist for multiple periods. On a 1-period trade, you need a spread of 0.18–0.39% just to break even.
The withdrawal fee trap
Cross-exchange arb also requires capital on both exchanges simultaneously. You can't withdraw from exchange A to cover a margin call on exchange B without a delay of 10–60 minutes (and a flat fee of $1–$5 for USDT on non-TRC20 chains, or $15–$25 for ERC-20).
We excluded withdrawal fees from the table above because many traders hold idle margin on both exchanges. But if you need to transfer capital during the arb, the withdrawal cost can wipe a small trade's profit entirely.
30-day spread audit: what actually survived
We tracked the Binance vs Bybit funding spread every 8 hours for 30 pairs from June 15 to July 14. We flagged each period where the spread was positive (arbitrageable) and measured how many consecutive periods it sustained above the taker-taker break-even threshold.
| Pair | % of 8h periods with spread >0.02% | Max sustained periods above 0.05% | Profitable after taker fees? |
|---|---|---|---|
| PEPE | 34% | 11 periods (3.7 days) | Yes — marginally |
| WIF | 28% | 8 periods (2.7 days) | Yes — marginally |
| DOGE | 21% | 5 periods (1.7 days) | Borderline |
| SOL | 12% | 3 periods (1.0 day) | No |
| ETH | 8% | 2 periods (0.7 day) | No |
| BTC | 6% | 2 periods (0.7 day) | No |
| AVAX | 9% | 2 periods (0.7 day) | No |
| Most others | <5% | ≤1 period | No — cost > spread |
Of 30 pairs tracked, 2 were marginally profitable after taker fees on a taker-taker basis (PEPE and WIF). The rest had spreads that either appeared for too few periods to cover the round-trip cost, or the spread was simply too small.
The anatomy of a spread that looked good but wasn't
The SOL case is instructive. On July 7, the Binance-Bybit SOL funding spread hit 0.045% — the kind of number that looks great at first glance. Here's what the actual trade looked like:
The spread was 0.045% when measured. After fees, you need it to persist for 6 periods to break even. It lasted one. The "opportunity" was noise.
What actually worked: maker order execution
The two profitable pairs (PEPE, WIF) were only profitable in the maker-maker scenario — placing limit orders on both sides and waiting for fills rather than hitting the book immediately.
| Scenario | PEPE net 30d ($10k/leg) | WIF net 30d ($10k/leg) |
|---|---|---|
| Taker entry + taker exit | −$34 | −$18 |
| Maker entry + taker exit | +$28 | +$11 |
| Maker entry + maker exit | +$91 | +$63 |
The catch: maker orders aren't guaranteed fills. In several instances, price moved away before one leg filled, leaving a directional exposure — the opposite of what you want.
The three categories of spread
Based on the 30-day audit, we grouped all observed spreads into three buckets:
- Noise (<0.02% spread): 73% of observations. Not worth acting on under any fee structure. The spread is within normal market microstructure variation.
- Borderline (0.02–0.05%): 21% of observations. Profitable only with maker orders AND the spread must persist for 4+ periods. Rarely the case.
- Real opportunity (>0.05%, persisting 4+ periods): 6% of observations. Consistently clustered around major narrative events (meme rallies, CPI days, major protocol launches). With maker orders, these were the only trades that cleared costs comfortably.
Practical filter: Only act when the spread has been >0.05% for at least 2 consecutive periods AND the pair is a high-volume memecoin (tighter spreads, more persistent sentiment). Use limit orders on both legs. Set a mental exit when the spread drops below 0.03%.
Pairs to watch for the next opportunity
Based on historical pattern — high funding, liquid enough for tight spreads, with sentiment extremes — the pairs most likely to show actionable arb opportunities:
- PEPE: Highest base funding, largest narrative swings. When exchange funding diverges, the spread can widen to 0.15%+ during a pump.
- WIF: Second-most volatile funding. Bybit sometimes lags Binance during rapid moves.
- DOGE: Larger notional markets mean tighter spreads but more liquidity — favourable for bigger positions.
- New listings in the first 48h: The biggest arb windows we observed were on newly listed pairs where one exchange was adding it days before another. Rates diverged dramatically before normalising.
Bottom line
Cross-exchange funding arb is real, but it's not a passive money printer. The honest summary:
- ~6% of spread observations were genuinely actionable after real costs
- Maker execution is almost mandatory — taker fees erode most spreads
- Capital requirements are meaningful — you need idle margin on both exchanges at all times
- Execution speed matters — spreads collapse within 1–2 periods as arbitrageurs pile in
- Not risk-free — exchange risk, margin call risk, and withdrawal delays are all real
If you're going to trade this, our live arb scanner shows current spreads with estimated break-even periods. The funding pain calculator lets you model the cost on your specific position size.
FAQ
What is cross-exchange funding arbitrage?
Funding arb means going long on the exchange charging the lower funding rate and short on the exchange charging the higher rate — for the same asset. If the spread exceeds your all-in cost, you earn the difference delta-neutral.
Why do most funding arb opportunities disappear after fees?
Taker fees on both legs cost 0.08–0.12% for entry and the same for exit. A spread that doesn't persist long enough to earn back 0.16–0.24% in funding is unprofitable — and most spreads revert in 1–2 periods.
Is funding arbitrage risk-free?
No. Delta neutral doesn't mean risk-free. Exchange counterparty risk, margin calls on one leg during a sharp move, withdrawal delays, and spread collapse before break-even are all real risks.
Does maker vs taker order type matter?
Significantly. Maker fees are typically 0–0.02% vs taker at 0.04–0.06%. Using limit orders for entry and exit can halve your fee cost, turning an unprofitable taker-taker arb into a profitable maker-maker one.
Which pairs had the most durable arb opportunities in June–July 2026?
PEPE and WIF showed the most sustained positive spreads. Both were only profitable with maker order execution. BTC and ETH had narrower but more stable windows — not enough to overcome taker fees without exceptional timing.