Bitcoin 2026-09-01 daily brief — AI stance: bearish

Funding is at or below neutral and gives no support to a crowded-long story. Open-interest-weighted across venues it annualizes to 8.98%, below the ~11% per year that the exchange-default 0.01% per eight hours implies as the resting baseline. The single major venue in the dataset reads cooler still at 5.74% annualized, a 3.24 percentage-point gap to the all-venue aggregate — meaning whatever mild long-carry demand exists sits elsewhere in the market, not on that venue. Neither figure is hot. This is the strongest single argument against the bearish call: there is no leverage excess here to unwind, so any decline has to be driven by spot supply and macro repricing rather than by a cascade.

De-risking without panic. Market-wide futures open interest is $53.51B, down 5.93% from $56.89B on Aug 21 while price fell only 1.54%, which is the signature of position closure rather than directional pressure. Options open interest at $38.78B is off the $45.09B peak of Aug 28, with volume at $3.91B well below the $6-12B days seen through the squeeze. Funding sits below neutral, forced liquidations have rotated to longs, and the aggregate picture is a market that has quietly taken chips off the table after the Aug 19-27 run rather than one that is positioned for either direction. That matters for the risk framing: a market with thin leverage falls on spot supply, which is slower and grindier than a cascade but also harder to reverse quickly.

The polarity has flipped. The Aug 19-27 advance from roughly 69,200 to 80,297 was paid for by short liquidations — $261.6M on Aug 20 and $470.1M on Aug 21, $731.7M in two sessions on days that printed +4.85% and +7.66%. Since Aug 28 the forced flow has come from the other side: $102.8M of longs on Aug 28's -3.24% day, and on Sept 1 $56.6M of longs against just $13.3M of shorts, a ratio of 0.23. Total liquidation activity sits in the 55th percentile of the past year, so the tape is not violent — it is simply that the seller of last resort has switched sides.

Neutral, but persistently and mildly offshore-led. The Coinbase premium is -2.36 bps, well inside the +/-10 bps that marks a practical extreme, and the trailing thirteen sessions average about -1.1 bps with 9 of 13 negative. The single most informative point in that series is Aug 28 at +3.21 bps — the only session with a clear US bid, and it was the -3.24% down day, i.e. US buyers stepping into weakness rather than leading strength. Conversely Aug 31 printed -1.27 bps on a session that took in $217M of ETF inflows. Read together, the US spot bid is present but not leading, which is a weaker demand backdrop than the headline ETF numbers on their own suggest.

CONFLICT. The macro tape pushes down hard — a Fed with 66% hike odds, the highest 10-year since January 2025, a dollar up 1% in twelve sessions, WTI up ~5% through $90 on the Strait of Hormuz strikes and Iranian attacks on the UAE and Jordan, and equities giving back. The local Bitcoin data does not push down at all: the cycle monitor says NEUTRAL with zero top indicators, valuation is at the 38th percentile, funding is below its neutral baseline, and price is 13.81% above the 50-day. Only one local variable agrees with the macro, and it is the flow variable — ETF absorption fading by 59% in five sessions. I take the macro side for this horizon, for a specific reason: the on-chain valuation read is a months-scale signal being asked a one-week question, whereas the dated catalysts and the flow deceleration both live inside the window.

Three threads, and two of them are negative. First, ETF demand is decelerating hard: the 7-day average flow has fallen from 4,853 BTC on Aug 27 to 1,992 BTC on Sept 1, a 59% collapse in five sessions, with Sept 1 itself at -457 BTC and a 30-day flow z-score of -0.63. The 30-day average is still positive at 1,380 BTC per day, so structural absorption has not reversed — it has thinned. The feed cross-checks cleanly against Farside, whose reported ~$217M inflow for Aug 31 matches the +2,758 BTC recorded that day. Second, miner economics are weak: Puell at 0.9271 is below 1, fees per transaction sit at 479.8 sats in the 11th percentile even as transaction count hits 627,345 in the 98th percentile — high blockspace usage with almost no fee bid — and the hash ribbon ratio is still 0.9923, below 1, though rising for eight consecutive sessions. Third and more subtly, long-term-holder SOPR is 0.9627 in the 24th percentile, meaning the coins the long-term cohort is actually moving are being realized at a loss even with price 11% above the 200-day average. Set against that, the corporate bid returned: Strategy bought 4,603 BTC for $369.7M in the week ending Aug 30 at an average of $80,318, ending a ten-week pause and funded by $602.8M of common share sales. That purchase is already 3.83% underwater, and their aggregate cost basis of $75,412 sits only 2.42% below spot.

Hawkish, and hawkish in the rare direction — the market is pricing a hike, not a cut. Effective fed funds is 3.63% inside a 3.50-3.75% range held 9-3 on July 29, and after Chair Warsh told Jackson Hole on Aug 28 that the Fed still has work to do if underlying inflation is not moving to 2%, CME FedWatch odds of a September 16 hike jumped to roughly 66% from about 35%. Sentiment has not caught up: the Fear and Greed index reads 69, squarely in Greed, while the policy path is repricing tighter. M2 growth at +5.41% year-over-year is the single loosening term in the picture and it is not enough to offset a hiking bias layered onto an oil shock.

The 10-year is 4.75%, up 10bp from 4.65% on Aug 20 and, per Bloomberg's Aug 31 report, the highest since January 2025 with an intraday 4.788%. The 30-year is 5.272% and the selloff is global: 10-year JGBs above 3% for the first time since 1996, Bunds at 3.364%, gilts at 5.254%. There is no credit-spread feed in this dataset, so I have no read on credit — that gap matters here, because a synchronized global term-premium repricing is exactly the setup where the equity tape and the credit tape can diverge, and I can only see one of them.

DXY at 99.742, up 1.02% from 98.735 on Aug 20 and grinding higher on eight of the last twelve sessions. A dollar rising on rate-differential expansion rather than on a growth scare is the version that historically works against a long-duration, dollar-denominated risk asset, and it removes the mechanical tailwind that carried the Aug 19-27 advance.

Deteriorating at the margin. The S&P 500 closed at 7,631.47 on Sept 1, down 1.29% from its 7,730.99 close on Aug 27, while VIX rose from 14.43 on Aug 28 to 16.34 — a 13% jump in three sessions. This is not risk-off yet, but it is the first sustained give-back in the window and it coincides with the oil move rather than preceding it.

The structure below is layered rather than a single number. First support is the 76,998-77,000 shelf that has contained the last eight sessions, essentially at spot. The meaningful level is 72,660, the Aug 20 close from which the squeeze launched — a full round-trip of the +7.66% Aug 21 day, 5.9% below spot. Below that there is little structure until the 50-day at 67,871 (-12.1%) and the 30-day low at 62,830 (-18.7%). Sizing those with the volatility actually observed: a seven-day standard deviation is roughly 4.5% using last week's 32.37% realized and roughly 6.1% using the 30-day 43.92%, so a tag of 72,660 inside the horizon is somewhere between a one-in-ten and a one-in-six event on a symmetric read, and higher than that given the event stack skews the distribution left. A touch of the 50-day within the week is a two-to-two-and-a-half sigma move — low single-digit percent, not the base case. The scenario I would actually underwrite is a grind to the low 70,000s on continued flow deceleration rather than a cascade, precisely because open-interest-weighted funding at 8.98% annualized means there is no leverage stack to accelerate it.

moderate — with an unusual implied/realized configuration that is the crux of the view. DVOL at 37.93 sits in only the 24th percentile of the past year and 30-day ATM implied is 35.71% after dropping 5.85 vol points in five sessions, so options are priced near the calm end of the year. Realized volatility is split: 43.92% over 30 days but 32.37% over 7. Implied is therefore 8.2 points below trailing 30-day realized, which sounds like cheap options — but the honest caveat is that trailing 30-day realized is inflated by the Aug 19-21 squeeze days (+6.92%, +4.85%, +7.66%), and against last week's 32.37% the implied level is a normal small premium. So the 'options are cheap' claim rests entirely on whether the calm holds through the payroll print and the pre-positioning into mid-September. The event calendar says it should not, which is precisely why the recent 5.85-point volatility sale reads as complacency rather than as information.