Bitcoin 2026-08-14 daily brief — AI stance: bearish

Open-interest-weighted funding across venues is 0.0089 per 8 hours, which annualizes to 9.76%. The exchange-default 0.01% per 8 hours annualizes to roughly 11%, so the market as a whole is paying slightly less than the neutral carry to hold long exposure. This is not a crowded book and nothing about it is stretched. Nor has it built: the same cross-venue measure printed 9.37% annualized on Aug 2, so across the full two weeks it is flat. What happened in between was a round trip — down to a 2.56% annualized trough on Aug 4 as the early-August rally ran on spot rather than leverage, then back up to 10.46% by Aug 13 as leverage returned on the way down. Leverage rebuilding into a falling price is a mildly negative sequence, but the level it rebuilt to is unremarkable. Separately, the single major venue broken out on its own prints only 4.08% annualized, 5.7 percentage points cooler than the cross-venue aggregate; that locates the leverage away from that venue rather than in it, and it is a venue-level observation, not a spread between two market-wide measures. Net read: no fuel for a long squeeze in either direction, and no leverage-driven support underneath either.

Market-wide futures open interest is $47.94B. Read across the full window it is flat to slightly lower — $48.19B on Aug 2, a peak of $49.50B on Aug 5, and 3.2% below that peak now — so the two-week story is de-risking into the mid-month break, not a leverage build. The sharper detail is the most recent session: open interest rose 2.5% from $46.78B to $47.94B while price fell 1.02%, so yesterday specifically saw new positioning added on the wrong side of a declining tape. Options open interest of $27.08B is within 0.4% of the two-week high of $27.20B set Aug 7, on $2.05B of daily volume, and the 30-day skew sits at the 23rd percentile of the past month — the options market is not paying up for downside protection, which is consistent with a 30-day implied index at the 3rd percentile of the year. The one venue whose open interest is broken out separately runs $1.46B, roughly 3% of the aggregate and at a two-week high; that is a venue-level datapoint, not the market's positioning.

Today saw $21.09M of longs liquidated against $3.66M of shorts, a 5.8-to-1 skew, on a day when total liquidations sat at only the 16th percentile of the past year. The regime flipped cleanly on Aug 10, the session after the Aug 9 high of 65,030: every session before it was short-dominant, including Aug 5 at $6.6M long against $37.9M short, and every session since has been long-dominant. The character of that flip matters more than its direction. There has been no single flush — the largest long print in the window is $35.1M on Aug 10, which is small in absolute terms — so the long inventory that got trapped between 64,000 and 65,000 during the Aug 4-9 rally has never been forcibly cleared. It is still there, and it is still underwater. That is unresolved supply rather than a completed washout.

The US-versus-offshore spot spread is -10.12 bps, meaning US venues are trading a tenth of a percent below offshore. Against an empirical distribution whose 99th percentile is roughly 15 bps, that is bottom-decile. The level matters less than the persistence: this spread has been negative in all 13 sessions with data from Aug 2 through Aug 14, spanning -4.87 bps on Aug 8 to -11.09 bps on Aug 3, and it has re-widened over the last three sessions from -8.30 bps on Aug 9 into the -10.1 to -10.8 bps band. Offshore is the marginal buyer and has been for two straight weeks. This is the most consistent bearish signal in the window, and it now corroborates rather than merely anticipates the fund flow data: a US institutional bid that is absent in the spot spread is the same bid that just turned the 7-day flow average negative.

CONFLICT

The dominant narrative is that the US-listed fund bid has switched sides. Four straight positive sessions Aug 3 through Aug 7 (+2,677, +3,293, +3,784, +2,140 and +1,568 BTC) drove the 7-day average to +1,834 BTC; four of the last five sessions have been negative and the 7-day average crossed to -210 BTC on this print. Excess absorption is now -477 BTC, meaning the funds are not taking down even the miner issuance. Miner economics are the quiet second story: transaction count at 694,502 is the 98.9th percentile of all history while fees per transaction at 454 sats sit in the 10.8th percentile — record throughput generating near-record-low revenue per transaction, which leaves post-halving miners fully issuance-dependent and structurally forced sellers. Puell at 0.771 is the 26th percentile and the hash-ribbon ratio at 0.989 is the 10th, though it has ticked up from its Aug 9 low of 0.98615. On regulation, the near-term catalyst was removed rather than delivered: the SEC canceled its Aug 14 open meeting at which commissioners were to vote on proposing Regulation Crypto Assets, and separately delayed the innovation exemption for tokenized securities. The Senate CLARITY Act cloture motion filed Aug 8 cannot get a vote until after the chamber returns Sept 14. The one crypto-relevant positive on the calendar — Jackson Hole running Aug 27-29 on the theme of financial innovation and payments, with Chair Warsh delivering his first keynote as Chair on Aug 28 — falls outside this horizon.

Hawkish in effect, whatever the on-hold label says. Policy sits at 3.63% after the July 28-29 meeting held the range at 3.50-3.75%, and the reaction function is the tell: July retail sales fell 0.6% month-over-month against a +0.1% consensus, the largest drop in nearly a year, and the University of Michigan preliminary August sentiment print collapsed almost 8% to 51.0 from 55.2 — and market-implied odds of a September rate hike fell only from 35% to 29%. A central bank that responds to a visibly cracking consumer by shaving four points off hike odds while pricing no cut at all is tightening in real terms. M2 is growing 5.53% year over year, so nominal liquidity is expanding, but not fast enough to offset a 3.63% policy rate. The live sentiment gauge reads 25, or Extreme Fear, but it is stamped 2026-07-18 and is therefore four weeks stale; the 30-day implied-volatility index at 34.74 and the 3rd percentile of the past year says the options market is registering no fear whatsoever, and I weight the current reading over the stale one. The Aug 19 minutes cover a meeting that changed nothing; the real event risk is Aug 26 core PCE and the Aug 27-29 Jackson Hole symposium, both outside this brief's horizon.

The 10-year at 4.63% has fallen 12bp from 4.75% on Aug 3, a modest bull-flattening on weak consumer data. But it still sits a full 100bp above the 3.63% policy rate — a steeply positive slope that, arriving alongside collapsing retail sales and sentiment, reads as term premium and inflation concern rather than growth optimism. Gold at $4,432 is consistent with that interpretation. There is no credit-spread feed in this dataset, so I have no read on corporate credit and will not invent one.

Neutral and unusually quiet. DXY at 99.64 has traded a 99.60-to-100.02 band across all 13 sessions with data, a range of under half a percent, and is net down 0.08% over two weeks. This matters mainly for what it removes: there is no strong-dollar squeeze to blame BTC's 19.25% drawdown from the 90-day high of 77,809 on. The weakness is idiosyncratic to the asset.

Risk-on and diverging hard from crypto. The S&P at 7,785.76 is within 0.2% of its two-week high of 7,798.99 set Aug 13, and up 2.4% from 7,600.50 on Aug 3. VIX at 14.25 is down from 16.50 on Aug 4. Over the same window BTC fell from an Aug 9 high of 65,030 to 62,830. Equities making highs on a 14-handle VIX while BTC grinds to a 30-day low is the single most important cross-asset fact in this brief.

Immediate downside is thin rather than deep. Spot at 62,830 is 0.08% above the 30-day low of 62,780, and there is no tested level between that shelf and the 60-day low at 58,519 — a 7.4% gap with no structure inside it. Scaling by realized volatility gives a wide range of plausible weeks: at the 7-day reading of 13.23% a one-sigma week is plus or minus 1.8%, which would not even reach the gap; at the 30-day reading of 22.15% it is plus or minus 3.1%; at the 90-day reading of 34.45% it is plus or minus 4.8%. Since volatility this compressed mean-reverts upward, the honest framing is bimodal rather than a single number: the modal outcome is a 1-to-3% grind that holds the 62,780 shelf and keeps the range intact, but the tail toward 58,500-60,000 is materially fatter than a normal week would imply, because the 30-day skew sitting at the 23rd percentile of the month says very little downside protection has been bought. Context for scale: the market is already 19.25% below its 90-day high of 77,809, so this would be the continuation of an existing drawdown rather than the start of one, and the 200-day at 69,515 is 9.62% overhead as the level any recovery has to eventually reclaim.

low