Bot trading · Cost · Updated June 2026

Arbitrage bot fees 2026: the hidden margin killer — and how to survive it

Arbitrage looks like free money: spot the same asset priced differently in two places, buy low, sell high, pocket the gap. The catch is that you pay a fee on every single leg, and arb legs are almost always taker orders that have to fire instantly. On a strategy whose edge is measured in basis points, that fee stack is frequently larger than the spread itself. Here's the real cost math for the three common arb types — and the three levers that decide whether your book lives or dies.

The short version

Arbitrage is the most fee-sensitive strategy in crypto because the edge per trade is tiny and the trade count is enormous. Three facts drive everything:

Every leg pays a fee. A two-venue arb is two fills; a cash-and-carry is two legs plus funding; a triangular arb is three fills. Fees multiply with legs.
Arb is taker-heavy. The gap closes in milliseconds, so you cross the spread with market orders and pay the taker fee — the expensive side of the book.
Your break-even is the sum of all leg fees plus slippage. If the spread you see is smaller than that number, the trade loses money even when the screen says "profit".

The fix isn't a better signal — it's a lower cost base. Maker routing where possible, the right venue, and a fee rebate on the taker legs you can't avoid.

The break-even spread: the only number that matters

Before any arb is profitable, the observed price gap has to clear your all-in cost. For a simple two-venue spot arb where both legs are takers at 0.05%:

Leg 1 (buy, taker) = 0.05%
Leg 2 (sell, taker) = 0.05%
Fee break-even = 0.10%, before slippage, withdrawal/network fees and the risk of the gap closing mid-execution.

So a 0.08% price discrepancy — which looks like a clean win — is actually a guaranteed loss of about 0.02% per turn once fees land. Repeat that thousands of times a month and you have a bot that trades constantly and bleeds steadily. The entire game is pushing that 0.10% break-even down so more of the gaps you observe are genuinely executable.

Three arb types, three fee profiles

Arb type	Legs & fee exposure	Maker possible?	Extra costs
Cross-exchange (spot)	2 taker legs (~0.10% total)	Rarely — speed-bound	Withdrawal/network fees, transfer latency
Triangular (one venue)	3 taker legs (~0.15% total)	Sometimes on the slow leg	Slippage on the thin pair
Funding-rate / cash-and-carry	2 legs (spot + perp), held	Often — post the non-urgent leg	Funding payments over the hold

Fee figures use a 0.05% taker reference at the regular tier; your real rate depends on the venue, product and VIP tier. Funding-rate arb is the most fee-friendly because at least one leg is usually patient enough to post as a maker order.

Worked example — a $10M/month cross-exchange spot bot

A cross-exchange spot arb bot turns over $10,000,000 of notional per month (counting both legs), almost entirely taker, at 0.05% per fill:

Total fees = $10M × 0.05% = $5,000 / month (about $60,000 a year)
If the average captured gross spread is 0.12%, gross revenue ≈ $10M × 0.12% = $12,000 — but you only count notional once for revenue, so on ~$5M of round trips that's ≈ $6,000.
Net before fees ≈ $6,000; after $5,000 in fees, ~$1,000/month survives. Fees are eating ~83% of the gross edge.

Now apply an up-to-40% rebate on those fees:

Rebate ≈ $5,000 × 40% = $2,000 / month back, settled weekly in USDT.
Net jumps from ~$1,000 to ~$3,000 / month — a 3× improvement in take-home P&L, with zero change to the strategy, the signal, or the infrastructure.

That is the whole point: on a thin-margin strategy, the rebate isn't a nice-to-have — it can be the majority of your net profit. See the mechanism on the OKX rebate and Binance rebate pages.

The three levers that keep arb alive

Maker routing wherever the trade tolerates it. Funding-rate and some triangular legs can post limit orders. Moving even one leg from taker (0.05%) to maker (0.02% or rebated negative) meaningfully drops break-even. This is the same maker-over-taker logic from our maker vs taker breakdown.
Venue choice. A 0.01% difference in taker fee is enormous when you're turning over $10M+/month. Compare real rates in the cheapest exchange ranking and the Binance vs OKX comparison.
Rebate the taker legs you can't avoid. Cross-exchange latency arb is all-taker by nature — you can't maker your way out of it. The rebate is the only lever left, and it gives back up to 40% of those unavoidable fees.

For a broader view of how fees vary by bot type, see our crypto trading bot fee comparison; the arb profile is simply the most fee-exposed corner of that map.

Don't forget the non-trading costs

Trading fees are the headline, but a complete arb model includes withdrawal and network fees on cross-exchange transfers, funding payments on held legs, and the opportunity cost of pre-funding both venues to avoid transfers. None of these are rebatable — which is exactly why squeezing the trading-fee line as hard as possible matters: it's the one big cost you actually control.

FAQ

Why do fees matter so much for arbitrage bots?+

Arbitrage captures tiny price discrepancies — often a few basis points — but it pays a fee on every leg of every round trip, and those legs are usually market (taker) orders because arb has to execute now or the gap closes. A two-leg arb at 0.05% taker per side costs 0.10% in fees, which can be larger than the spread it is trying to capture. Unlike a directional trade where fees are a small fraction of the move, in arb the fee is often the single biggest line in the P&L.

How do I calculate the break-even spread for an arbitrage trade?+

Add up the fees on every leg plus expected slippage, and that sum is your break-even. For a two-venue spot arb at 0.05% taker each side, break-even is at least 0.10% before slippage and transfer costs. For a cash-and-carry (spot plus perp) trade you also factor funding payments over the holding period. If the observed spread is below your all-in cost, the trade loses money even though it looks profitable on the screen.

Does a fee rebate help arbitrage bots?+

Yes — directly. A rebate gives back a percentage of the trading fees you pay, which lowers your all-in cost per leg and therefore lowers your break-even spread. A lower break-even means more observed gaps become profitable, so the same market produces more executable opportunities. Up to 40% of fees back, settled weekly in USDT, can be the difference between a marginal arb book and a profitable one — without touching the strategy.

Is maker or taker better for arbitrage?+

Maker is far cheaper and sometimes even pays you, but arbitrage usually demands speed, which pushes you toward taker orders that cross the spread immediately. Some strategies — like funding-rate (cash-and-carry) arb where one leg is held — can post the non-time-critical leg as a maker order to cut cost. Cross-exchange latency arb is typically all-taker. The practical answer: post maker wherever the trade allows it, route taker where it doesn't, and rebate the taker legs you can't avoid.

Do cross-exchange arbitrage transfers add hidden costs?+

Yes. Beyond trading fees, cross-exchange arb can incur withdrawal/network fees, on-chain confirmation delays, and the risk that the spread closes before assets settle. Many desks avoid physical transfers by pre-funding both venues and rebalancing periodically, which trades transfer cost for capital cost. Either way, model these alongside trading fees — they are part of the all-in number that defines your break-even.

Lower your break-even spread

Get up to 40% of your trading fees back on the taker legs you can't avoid — settled weekly in USDT, fully reconcilable, no strategy change. On a thin-margin arb book, that's often the majority of your net.

Binance rebate OKX rebate Get a tier assessment

Disclaimer: Fee schedules and rebate rates reflect published rates at the time of writing and may change without notice. Examples are illustrative; your real costs depend on venue, product, maker/taker split and VIP tier. This article is educational and not investment advice. JackTrader is an independent referral / sub-broker partner and is not Binance or OKX. Single-tier referrals only, no downline or multi-level structure.