Perp Market Week 35, 2026: Volume, Funding, Costs

What this report measures

PerpFinder collects four stored series of the perpetual market. The daily snapshot series stores the reported 24-hour volume of each venue once a day. The funding sweep stores the rate of each venue every 15 minutes. The execution-cost sweep walks the order book of each venue every 6 hours. This report reads those stored series for ISO week 35 of 2026, which runs from 2026-08-24 to 2026-08-30.

Each weekly volume figure is the sum of 7 stored daily snapshots. A snapshot covers the 24 hours before its own capture instant. The sum therefore measures 7 stored days, and it does not measure an exact calendar week. Each table on this page states its own measurement instant, because the collection clocks are different.

The generator reads only stored PerpFinder data. It fetches no venue endpoint at build time, and it writes no value that the stored series does not hold. The methodology of each series is on the research hub.

Volume in week 35 against week 34

The centralized venues stored $1,476.96 billion of perpetual volume across the 7 days of 2026-W35. The perp-DEX protocols stored $158.87 billion across the same 7 days. Both figures apply the spike rule. That rule holds a venue-day above 3 times the 7-day median of the same venue out of the total.

The stored total of 2026-W34 is $1,438.33 billion. That week does not sit on the same basis. The snapshot series reconstructed 6 of its 7 stored days after the fact, from the 00:00 UTC volume-quality sweep. Each reconstructed day carries fewer venues. A comparison of the two headline totals therefore compares two different venue sets.

The comparable set holds the 20 venues that both weeks store on every day. On that set the centralized total moved from $1,402.13 billion to $1,356.30 billion, a change of -3.3 percent. The set leaves out Backpack, BitMEX, WEEX, because 2026-W34 does not store those venues on every day.

The spike rule and the venue-days it held out

PerpFinder flags a venue-day when the venue reports more than 3 times the median of its own previous 7 stored days. A flagged venue stays visible on the site and leaves the headline total until the figure holds. The rule needs at least 3 stored days behind the median.

No venue-day of 2026-W35 passed the rule, so the weekly total of 2026-W35 holds every stored venue-day. The rule flagged 21 venue-days of 2026-W34, and the spike table names each one.

The flagged venue-days of 2026-W34 fall on the reconstructed days. A reconstructed day reads the 00:00 UTC sweep, and the captured days read a later instant, so the two measurements sit on different clocks. The multiple measures that basis change as much as it measures a market move.

Venue ranks and the moves inside them

The comparable set puts five venues at the top of 2026-W35. The list is Binance at $389.76 billion, OKX at $158.48 billion, Toobit at $130.53 billion, Bybit at $95.59 billion, MEXC at $88.03 billion.

The rank table counts the venue-days that entered each weekly figure. The spike rule took 21 venue-days out of 2026-W34, so a venue with a lower day count also carries a lower weekly figure. A move that pairs with a different day count is partly a coverage effect.

Among the venues with 7 counted days in both weeks, the largest move belongs to Bitunix, from rank 8 to rank 12.

Perp-DEX share of the measured volume

The perp-DEX protocols hold 9.71 percent of the measured perpetual volume of 2026-W35. The share divides the stored perp-DEX total by the sum of that total and the centralized total.

The daily share moved between 7.22 percent on 2026-08-24 and 17.90 percent on 2026-08-29. Both ends come from the daily volume table.

The snapshot series stores a perp-DEX total on 1 of the 7 days of 2026-W34. This report therefore states no perp-DEX share for that week. PerpFinder writes no backfill to close such a gap.

BTC funding: level and cross-venue spread

The open-interest-weighted BTC funding rate held a median of 0.000796 percent per hour across the 672 sweeps of 2026-W35. At a simple annual rate that median is 6.98 percent. The median stayed positive, so a long paid the short over the week on the weighted average of the venues.

The cross-venue spread is the highest venue rate minus the lowest venue rate in one sweep. Its median over the week is 0.00320 percentage points per hour. The widest single sweep of the week is 2026-08-25 06:30 UTC, at 0.01735 percentage points per hour. In that sweep 26 venues published a rate, between -0.00035 percent and 0.01700 percent per hour.

The week stored 672 of the 672 sweeps that a full week holds at this cadence. Venue coverage stayed between 25 and 27 venues per sweep.

Realized funding, venue by venue

The realized ledger stores each settlement that a venue published. For 2026-W35 it holds 420 settlements from 6 venues, and every one of those venues published a complete week.

A long position paid the most on Hyperliquid, at 0.2182 percent of the position over the week. A long paid the least on dYdX, at -0.0068 percent of the position over the week. A negative figure means a long received funding instead of paying it.

The settlement interval differs between venues, so the realized table states the interval of each venue in minutes. A venue with a shorter interval writes more rows into the ledger. The weekly sum stays comparable between venues, because each sum covers the same period.

Execution cost of a $1,000,000 BTC perpetual order

The cost engine walks the order book of each venue every 6 hours and records the total cost of a $1,000,000 BTC perpetual order. Total cost is the taker fee plus the half-spread plus the size-dependent impact beyond the best quote. Across the 28 stored sweeps of 2026-W35 the median of the 926 order-book observations is 6.464 basis points.

The set of 28 order-book venues filled the order in every stored sweep of the week. The three lowest weekly medians in that set are Lighter at 1.571 basis points, MEXC at 2.024 basis points, Extended at 3.173 basis points. Each figure is the median of the sweeps of that venue. The cost table states the lowest sweep and the highest sweep of each venue.

An oracle venue has no order book, so it is not in these numbers. The walk found 12 venues that could not fill the order in at least one sweep, and a sweep without a fill contributes no cost figure. The week uses cost methodology version 4, which is the version the store held for the whole week.

What this week rests on

The volume figures rest on 7 stored days, 23 centralized venues per day and between 86 and 89 perp-DEX protocols with a positive figure. The funding figures rest on 672 sweeps and 420 stored settlements. The execution-cost figures rest on 28 sweeps and 926 order-book observations. The volume-quality series adds 28 sweeps of 20 venues inside the same week.

The week carries the gaps named here.

The snapshot series reconstructed 6 of the 7 stored days of 2026-W34 from the 00:00 UTC volume-quality sweep. That week therefore carries fewer venues and a different read instant.

The snapshot series stores a perp-DEX total on 1 of the 7 days of 2026-W34, so no perp-DEX comparison between the two weeks exists.

The comparable venue set leaves out Backpack, BitMEX, WEEX, because 2026-W34 does not store those venues on every day.

The spike rule held 21 venue-days out of the 2026-W34 total.

The cost series carries fewer than 28 sweeps for 6 order-book venues, so the weekly median of those venues rests on a shorter series.

The active cost methodology is now version 5, and this week is stored under version 4. The two versions are never merged.

Limitations

Reported volume comes from the ticker endpoint of each venue. PerpFinder normalizes the unit and the market type, and does not audit the venue figure.

A daily snapshot is one reading per day. A venue that moved inside the day, or that published a different figure at another hour, produces a different weekly sum on another clock.

The funding sweep and the cost sweep each read one instant. Depth and rates change between readings, and a later reading returns different numbers.

The realized ledger covers the venues that publish a settlement history. It does not describe the venues that publish none.

How to reproduce these numbers

Every number comes from public endpoints that need no key. The funding figures come from https://perpfinder.com/api/data/funding-history?asset=BTC&type=aggregates&days=30 and from the same endpoint with type=realized. The execution-cost figures come from https://perpfinder.com/api/data/cost-history?market=perp&asset=BTC&size=1000000&days=all. The volume figures come from https://perpfinder.com/api/data/venue-history?venue=Binance&days=30 and the same endpoint for each venue.

Those endpoints serve the current state, so a later call returns later numbers. Each table on this page carries a CSV with the exact rows of 2026-W35, and the fact block carries a CSV of every headline figure.

Collection cadence and updates

The collection cadence is fixed. The funding sweep runs every 15 minutes, the execution-cost sweep and the volume-quality sweep run every 6 hours, and the venue snapshot runs each day.

PerpFinder keeps no fixed publication schedule for this series. A published report is not edited after publication, and a later week is published as a new page.