Bitcoin 2026-09-02 daily brief — AI stance: bearish

Weighted by open interest across venues, funding runs at about +6.97% annualized. The exchange-default 0.01% per eight hours is roughly 11% a year and is the neutral baseline, so the market as a whole is paying less than baseline to be long — uncrowded, with nothing stretched to punish. The single major venue in the reference feed tells a different story: it has flipped to -0.6% annualized, the first negative print in the last fourteen sessions, where every prior day ran positive. That leaves it about 7.6 points below the all-venue aggregate. Read that gap as a location, not a market view: the short-side bid is concentrated on that one book rather than distributed across the market. The honest summary is that perpetual positioning is not the story here — an uncrowded book means a decline would have to come from spot selling, which is precisely what the flow data is showing.

Market-wide futures open interest is $53.10B, down from $56.89B on August 21 — $3.8B, or 6.7%, of leverage withdrawn while spot gave up only 1.8%. Options open interest is $38.89B against $4.37B of daily volume, with the term structure in ordinary contango at both the 7-to-30 day (+0.034) and 30-to-90 day (+0.039) segments and 30-day skew neutral at +0.005 in the 40th percentile. Put it together — falling futures open interest, below-baseline open-interest-weighted funding at +6.97% a year, quiet liquidations in the 31.8th percentile, and implied volatility in the 22nd percentile of 90 days — and this is a market that has de-risked into the event calendar rather than one positioned for a move. Thin books cut both ways, so the tail in either direction is fatter than the current volatility pricing implies. With the flow impulse negative and no crowded longs to squeeze, the path of least resistance into payrolls is lower.

Small and long-tilted. $23.9M of longs against $18.8M of shorts in the session, a 0.79 ratio, with total liquidations at only the 31.8th percentile of the past year. The contrast with August 21 is the point: $470.1M of shorts were force-covered that day on a +7.66% move. The violent positioning has been cleaned out in both directions over the past fortnight, and what remains is a thin book with little stored energy in either direction.

The Coinbase premium is -5.21 bps, US spot trading below offshore. That is inside the roughly ±10 bps practical extreme, so the level alone is not a signal. The trajectory is: -1.27, -2.36 and -5.21 bps over the last three sessions, the most negative reading of the 14-day window, against +3.21 bps on August 28. US-side demand led the August rally and has now been absent for three consecutive sessions. This is not independent confirmation of the ETF story — it is the same institutional bid measured a second way — but it does mean the flow reversal is visible on the spot tape and not only in the fund-flow print.

ALIGNED, bearish. The macro tape (10-year at 4.79% and rising, DXY at 99.56 and firming, Brent above $93, a hike-priced September FOMC) and the crypto-native tape (excess absorption at -613.8 BTC, the Coinbase premium at a 14-day low of -5.21 bps, futures open interest down 6.7% since August 21, short-term-holder NUPL compressed 34% in six sessions) push the same direction. The single genuine disagreement is price itself: 13.05% above the 50-day average and only 4.1% off the 90-day high, with 0 of 8 top-cycle indicators firing. I take the flow-and-macro side for a 5-to-7-day horizon and the trend side for anything longer, which is why this is a medium-confidence call and not a high-confidence one.

The demand impulse is rolling over faster than price is. The seven-day average of ETF flow has fallen from 4,853.7 BTC per day on August 27 to 980.9 BTC per day on September 2 — an 80% collapse in four sessions — with three net-negative days in the last six (-2,597 BTC on 08-28, -3,062 BTC on 09-01, -186 BTC on 09-02). The September 1 print reconciles exactly with the reported $236.5M net outflow, the largest single day since July 31, of which BlackRock's IBIT was $201.2M — and it follows $3.52B of August inflows, so this is a sharp reversal of the flow that produced the best August since 2017. Excess absorption is now -613.8 BTC with the flow-to-issuance ratio at -0.43: the funds are no longer soaking up miner issuance, they are adding to it. Miner economics are soft but not distressed — Puell at 0.894 in the 36th percentile, Hash Ribbons at 0.993 in the 12th percentile, the recovering-from-capitulation shape that is one of the three firing bottom triggers. The binary regulatory catalyst is the Senate cloture vote on the motion to proceed to the CLARITY Act at 2:15pm ET on September 15, which needs 60 votes against 53 Republican seats, landing on the same day the FOMC convenes.

Hawkish, and unusually so relative to what risk assets are pricing. Fed funds sit at 3.63% with the 10-year at 4.79% and M2 growing 5.41% year over year, and fed funds futures have moved to majority odds of a 25bp HIKE — reported in the mid-50s to roughly 60% — for the September 15-16 meeting, after Chair Warsh's August 31 Jackson Hole remarks were read as hawkish. The FOMC blackout runs September 5-17, so there is no speaker to walk that back before the decision. Against this, the Fear and Greed index reads 63, in Greed. Sentiment is positioned for calm into a central bank that is priced to tighten, which is the specific mismatch that defines this week.

The 10-year is at 4.79% and has risen in every session since August 31 (4.73, then 4.75, then 4.79), up 10bp from 4.69% on August 21. That move is happening alongside Brent above $93 following the September 2 US strikes on Iranian targets, so it is a term-premium and inflation-expectations move rather than a growth-optimism move — the worst version for a long-duration, zero-cashflow asset. There is no credit-spread feed in this stack, so I cannot say whether credit is confirming or diverging; in a week where an oil shock meets a hawkish Fed, that is a genuine blind spot rather than a formality, because credit is where a stagflationary impulse would show up first.

DXY at 99.56, up from 98.84 on August 21 — a 0.73% grind higher over twelve sessions, rising in step with the 10-year. The level is not extreme and by itself would not move BTC, but a persistently firming dollar into a possible hike removes the tailwind that carried the August rally, and it has been directionally consistent for two straight weeks rather than a single-day move.

Risk-on but stalled. The S&P sits at 7,666.60, 0.83% below its August 27 high of 7,730.99, with VIX at 15.2. Equities have gone nowhere for a week while yields rose. The tell is gold at $4,366.3 — a safe-haven bid coexisting with equity indices near highs says this is being traded as a rates and inflation problem, not a growth problem, and BTC's beta in that configuration is to real yields rather than to the equity tape.

The immediate shelf is 76,998 to 77,005 — the August 22 close and today's price sit $7 apart, which is a twice-tested floor and the lowest close since August 22. Below that the structure is unusually thin, and this is the most important line in the risk picture: the August 19, 20 and 21 sessions ran +6.92%, +4.85% and +7.66% back to back, a compounded 20.7% vertical move that left almost no traded volume between roughly 77,000 and the low-to-mid 60,000s. The 30-day low is 62,830 and the 50-day average is 68,114, so a decisive break of 77,000 has no meaningful structural support until 11.5% lower. That air pocket, combined with futures open interest already down 6.7% and liquidity thinned by de-risking, is why a modest catalyst could produce a disproportionate move. The base case is not a crash — it is a grind of 3 to 6% toward the low 70,000s over the week as the flow vacuum persists, with the fat left tail reserved for a hot payrolls print or a hawkish surprise that puts 77,000 through on volume. On the upside, a soft payrolls number that cuts hike odds is a fast 3 to 4% squeeze back toward 80,297.

low. DVOL at 37.19 sits in the 17th percentile of the past year and 30-day at-the-money implied volatility is 33.86%, the 22nd percentile of 90 days, after shedding 5.57 vol points in five sessions. Realized volatility agrees at the front end — 32.36% over 7 days against 44.03% over 30 and 38.31% over 90 — so the compression is real rather than an options-market opinion. The classification is low, but it is low in the wrong place: this is the cheapest volatility of the year priced into the week that contains payrolls, with CPI and a possibly-hiking FOMC directly behind it.