We tracked 238 markets across 11 exchanges for 19 days. Of the 435 funding carries that stayed above 6% APR for a full day, zero repaid their own round-trip cost.
Funding arbitrage — short a perp on one venue, go long the other leg on another, collect the funding spread — is the most advertised "delta-neutral yield" in crypto. Every dashboard sells the same picture: 30–80% APR, basically every day. Free money, on screen.
We wanted to know how much of that survives contact with real execution.
So we recorded funding, order books and fees for 238 markets (13 assets across 11 venues: Binance, Bybit, OKX, Bitget, Gate, KuCoin, HTX, MEXC, Hyperliquid, dYdX, Paradex) every few seconds for 19 days. Then, instead of screenshotting the peak APR, we replayed what a disciplined trade would actually earn.
The rule. Enter only after a carry has held above 6% gross APR for 24 continuous hours — a deliberately low bar, meant to catch carries that last, not 40-minute spikes. Enter at the end of that window (no hindsight). Hold while it keeps paying at a real $10k fill (slippage included). Net realized = funding collected − round-trip cost.
1. Perp–perp, cross-venue. 435 carries cleared the 6%/24h bar, 413 (95%) executable at $10k. 0 of 435 repaid their round trip. The best one missed break-even by basis points. This isn't missing edge — it's fees: switch taker→maker and 3 barely clear. The spread is real; the round trip eats it.
2. Cash & carry (short perp + long spot, same venue). A weeks-long position, so we measured it honestly: hold every entry to today and ask "is cumulative funding ≥ round-trip cost right now?" 205 entries, all executable. 0 above break-even at taker fees (at maker, two clear — barely, and both are still open). Median cumulative funding: −0.4% of notional. Different strategy, same wall.
3. The weirdest thing we caught. HTX showed ZEC spot at −380 bps vs everywhere else — a monster arb on any raw table. ZEC withdrawals there were suspended; the "opportunity" was the market pricing the risk you can't get your money out. A scanner that doesn't check withdrawal status sends you straight in. 1 of 239 markets, filtered automatically.
Caveats, stated plainly. 19 days, one funding regime, $10k size. A sustained high-funding regime could change the perp–perp picture — which is exactly why we keep measuring. But on real order books, in this window, the loud APR did not survive its own execution.
This is not "funding arb is dead". The spreads are real and opportunities do show up — but there is no free lunch: after fees, slippage and rate decay, what's left is smaller and rarer than advertised, and never risk-free (the ZEC case above is what "risk-free yield" looks like up close). The gross number is the ad; the net number, at your size, after the round trip, is the trade.