Bitcoin closed -3.2% at $83,679 after $546.89m of positions were flushed in a day. The interesting read is not the flush but what surrounded it: funding sat at 0.002%, the Coinbase premium stayed negative at -0.118%, and Hyperliquid's positions hit a record 435,564 on $4.54bn of daily volume.
This column reads the perpetual futures market through its two most important prices: the funding rate and open interest. Both tell you whether the marginal buyer is renting leverage or deploying capital. This week the answer inverted from September's: capital left at the spot layer while leverage stayed in the book, and the flush that followed was a function of geometry rather than of a repricing in the cost of carry.
Bitcoin ranged $80,315-$86,996, failed at the round number for a third straight week, and closed at $83,679. Through the whole move the funding clock barely stirred: Binance's BTCUSDT contract sat near 0.002% per eight hours, which annualises to a rounding error, and the dated curve held near 5% annualised, with the major December contracts at roughly 4.7-5.7%. A genuine leveraged breakdown would have driven funding sharply negative as shorts crowded in; that did not happen, because the crowd never flipped.
The flush hour itself is the cleanest funding signal. $412.62m liquidated in one hour, $400.24m of it longs (97.0%), against just $12.38m of shorts. When a single hour of forced selling is 97.0% one-sided and funding still prints flat, the forced sellers were late entrants with tight stops, not a concentrated leveraged position that the market had been paying to carry.
| Signal | Reading | Interpretation | Source |
|---|---|---|---|
| Perpetual funding, 8-hour clock | 0.002% | Binance's BTCUSDT funding sat near 0.002% per eight hours | Binance via FinanceFeeds |
| Coinbase premium | -0.118% | the U.S. spot discount stayed open all five sessions, widening to -0.118% on 11 October | TokenPost |
| Dated-futures basis | about 5% annualised | the dated curve held near 5% annualised, with the major December contracts at roughly 4.7-5.7% | Zerocap weekly wrap |
| Wednesday flush, peak hour | $412.62m | $400.24m of longs (97.0%); shorts only $12.38m | CoinGlass via TokenPost |
| ETF weekly flow | -$681.1m | the -$487.07m session mid-week | SoSoValue via Hodl.Press / TokenPost |
| Hyperliquid OI, tracker cohort | $17.29bn | $920m off the prior record | HyperTracker via Cryptopolitan |
Hyperliquid's book is the best single window into the on-chain perp complex, and this week it set a record: 435,564 open positions on 6 October, 115% above the 202,943 recorded in January, across 320,600 active wallets against 155,000. Open interest closed the week at $17.29bn, $920m below the prior record, with the venue's own two-sided mark at $12.77bn. Daily perp volume was $4.54bn, inside a $2-6bn band, down 65% from the $12bn-$16bn peaks in August.
The volume-to-open-interest ratio is the cleanest way to separate carry from churn, and it now reads close to four times. In September the desk read that ratio as maturity - capital staying in the book and turning over less often. The same ratio is the reason October behaved differently. A book that turns over three to nine days' worth of volume per full cycle cannot absorb a one-cohort exit without moving price, which is what a drop of less than two percent producing a $546.89m liquidation window looks like from the inside. Patience and fragility are the same property read from two directions.
The cohort arithmetic confirms the venue's position. Hyperliquid took 32.9% of the $616.2bn trailing-30-day perp DEX volume, printed $87.5m of fees over that window, and has settled $5.4tn of cumulative perp volume across 178 markets.
The US spot bitcoin ETF ledger is not a funding rate, but it functions as a spot-funding proxy: it measures whether new dollars are entering the bitcoin collateral pool without borrowing. This week it turned negative for the first time in a month. The five sessions ran -$89.9m, +$118.86m, -$487.07m, -$244.13m, +$21.13m - a net $681.1m - ending the three-week streak that had built $3.08bn. three straight positive weeks had lifted the year-to-date ledger before this week's reversal. The year-to-date ledger reads about $494m.
The mechanics matter for funding. When spot demand is strong, the cash-and-carry trade - long spot, short perp - is the natural arbitrage, and it compresses funding precisely because the demand is satisfied in the cash market. When spot demand turns, that compression mechanism weakens and funding becomes a purer read on the leveraged crowd. The fact that funding stayed at 0.002% while the ledger turned negative at $681.1m says the leveraged crowd neither grew nor shrank - it simply stayed, which is the configuration the flush then cleared.
| Session | Net flow | Note |
|---|---|---|
| 5 October | -89.90m | the week opened negative |
| 6 October | +118.86m | a brief rebound ahead of the Fed minutes |
| 7 October | -487.07m | the week's worst session of the five |
| 8 October | -244.13m | selling continued as oil and yields rose |
| 9 October | +21.13m | a tentative inflow to end the week |
| Five-day total | -681.11m | a net outflow that ended a three-week inflow streak |
Funding rates mean nothing without a measure of the collateral that can be posted against them, and this week the measure contracted. Total stablecoin supply fell to $304.2bn, supply contracted about $0.9bn on the week, with USDC leading the seven-day decline. That supply sits underneath both ETF settlement and perp margin, so a shrinking float tightens two systems at once. A falling collateral base with a flat funding clock is the mirror image of September's build: capacity to add is falling while the willingness to hold is not.
The perp DEX cohort's aggregate open interest was $14.80bn, against $616.2bn of trailing-30-day volume. Hyperliquid's share of that volume was 32.9%, and its own open interest $17.29bn on the tracker cohort. The ratio of 30-day volume to open interest is now lower than it was in August, which is the same carry-market read the prior issues established - and the same reason the week's flush was geometric.
The macro tape reinforced the read rather than contradicting it. The September minutes on 7 October showed that most officials judged a further increase appropriate before year-end (the 2026 median dot rose to 4.1%), leaving October odds near 17% against 75-85% odds of a hike by year-end. The 10-year closed at 5.24% and the 2-year at 4.80%, with the 30-year briefly touched 5.70% on 7 October, its highest since 2002. Gold added +1.2% to $4,220.30 and Brent rose +2.4% to $104.72 on Gulf shipping risk.
The cross-asset read is a repricing of the discount rate, not of growth. Equities took it in stride - the S&P 500 +1.2%, the Nasdaq +0.6% - while gold and crude rose with the inflation impulse. Crypto was the outlier that fell. For a carry trade, that is the hard combination: a higher risk-free rate raises the cost of financing a position while the asset's own demand is weakening, which is why the desk reads this week as a financing headwind rather than a growth scare.
A $546.89m window that clears $479.73m of longs inside a single hour is a violent event, and the first question a risk desk asks is why it did not become a broader break. The answer is where the leverage sat and how it was financed. The part of the perp market that publishes its book - Hyperliquid's $17.29bn tracker cohort, 435,564 open positions and $12.77bn on the venue's own two-sided mark - was observable right through the event, which is why this flush can be measured as it happened rather than inferred afterwards.
The second half is the financing. A stop-cluster becomes a break when the funding behind the book is withdrawn at the same moment as the price move, and this week the financing did not move: funding held at 0.002%, the Coinbase premium stayed at -0.118%, and open interest stayed at close to four times the day's turnover. The book rebuilt open interest inside the same session, which is the clearest evidence that the sellers were late entrants rather than the carry cohort the desk had tracked. That is a floor rather than a guarantee: it holds only while the two slow variables in front of it repair.
The five items share one theme: whether the two slow variables - the ETF base flow and the stablecoin collateral stock - repair before the next forced exit. The fast variable, funding, is already telling a calm story at 0.002%, and the desk treats a calm fast variable in front of two heavy slow ones as a setup to respect rather than an all-clear to buy.
That the flush was not a leverage event. Funding on Binance's BTCUSDT contract sat near 0.002% per eight hours through a $546.89m liquidation window, and the dated curve held near 5% annualised, with the major December contracts at roughly 4.7-5.7%. When 97.0% of the flush hour is one-sided and funding never moves, the forced sellers were late entrants with tight stops, not a crowded leveraged position the market was paying to carry.
Because it sets exit capacity. Positions hit a record 435,564 while daily volume sat at $4.54bn - a 65% decline from August peaks - leaving open interest at close to four times the day's turnover. A book that takes three to nine days to turn over once cannot absorb a coordinated exit without moving price.
ETF inflows add spot collateral to the bitcoin float without borrowing, which makes the cash-and-carry trade attractive and compresses funding. When the ledger turns negative - as it did this week at $681.1m, ending a three-week streak - that compression mechanism weakens and funding becomes a purer read on the leveraged crowd. Funding stayed at 0.002%: the crowd neither grew nor shrank, it simply stayed.
Three things: funding moving sharply away from 0.002%; a second consecutive week of ETF redemptions that takes the year-to-date total (about $494m) back toward the line; or a close below the week's $80,315 low on rising volume, which would convert the flush into a lower low rather than a shakeout.
The 14 October September CPI, with headline seen at 0.6% on the month and 3.6-3.7% on the year, followed by the 27-28 October FOMC decision. The market currently prices roughly 17% for October against 75-85% odds of a hike by year-end.