Bitcoin 2026-08-04 daily brief — AI stance: neutral

Funding is cold, not crowded. Weighted by open interest across venues — the right measure for a market-wide claim — perpetual funding annualises to 2.56%, roughly a quarter of the ~11% that the 0.01% per-8h exchange default implies as neutral. Traders are paying almost nothing to hold leveraged longs. That is a real change from two sessions ago, when the same OI-weighted measure ran near 9.4% annualised; leverage demand has drained out fast. One major venue is running hotter at 4.56% annualised, about 2 percentage points above the all-venue aggregate, so what crowding exists is concentrated on that single book rather than being a market-wide condition — and even there the level is under half the neutral baseline. The practical read: there is no long-side fuel for a liquidation cascade, which removes the most common mechanism for a sharp downside break in the next five sessions, and equally there is no momentum bid to power a breakout.

Flat and clean. Market-wide futures open interest is $48.60B and has not moved out of a $47.22-49.00B band in 14 sessions — no build in either direction alongside a 1.43% price decline, which means the churn has been position rotation rather than accumulation. Options open interest halved across the July monthly expiry, from $35.99B on July 31 to $25.26B on August 1, and has only crept back to $25.93B on $2.93B of daily volume, so dealer gamma has reset and there is less structural pinning holding price inside the band than a week ago. Combining cold funding, mid-range OI, 23rd-percentile liquidations and 0.3rd-percentile implied vol: nobody is positioned, which is exactly why the next macro print rather than the next order flow decides direction.

Two-way churn at low intensity. Total liquidations sit at the 23.29th percentile of the past year. The most recent session flipped the pattern: $29.0M of shorts stopped out against $6.1M of longs, a 4.79:1 ratio, on a 1.05% up move — shorts getting picked off near the bottom of the range. That follows two weeks in which longs took the heavier damage, with $71.7M stopped on July 27 and $71.3M on July 31, both on down days. The sequence describes a market that punishes whoever presses the edge of the band and rewards no one for holding a view.

US spot is the persistent laggard. The Coinbase premium is -8.42 bps today, and it has been negative in every one of the 13 sessions with data over the past two weeks, averaging about -9.6 bps with two readings past the ±10 bps extreme threshold (-13.78 bps on July 30, -12.33 bps on July 26). Offshore has led every session for a fortnight. Today's -8.42 is the least-negative print in a week and marginally improves the trend, but the two-week persistence is the signal, not the one-day improvement. This directly undercuts the bullish read of two consecutive positive ETF days: if US institutional spot were genuinely bidding, the premium would not be sitting at the bottom decile for ten straight sessions.

CONFLICT

The marginal-buyer story has inverted. Strategy disclosed selling 1,638 BTC between July 27 and August 2 at an average $63,957 — below its $75,419 cost basis — the second consecutive week of sales, with proceeds routed to STRC preferred repurchases and dividends, no purchases since June 22, and holdings down to 842,138 BTC. The largest corporate holder is now a price-insensitive seller at a loss. On the ETF side the two-day green streak is real but thin: +2,676.6 BTC on August 3 (about $172M, the '$170.09M' Farside print) and +641.6 BTC on August 4 (about $41M), lifting the 7-day average to +319.3 BTC — which is still only the 38.66th percentile — while the 30-day average sits at -1,028.95 BTC/day, the 9.97th percentile. Against that backdrop the first-ever US spot bitcoin ETF closure was announced on August 4, with the reporting explicitly tying dwindling inflows to investors rotating toward AI. Miner economics are deteriorating in parallel: Puell 0.7336 at the 22.99th percentile, hash ribbons 0.9877 at the 9.67th percentile and falling daily, and fees at 528.5 sats/tx at the 11.31st percentile against transaction count at the 97.80th percentile — very high throughput generating almost no economic value per transaction. MARA's Q2 call on August 6 is the live read on that stress. On the catalyst side, the CLARITY Act still has no scheduled floor vote with the Senate days from recess and not back until September 14, so the one dated regulatory upside is effectively off the calendar. Finally, an active Coldcard exploit traced to a March 2021 firmware PRNG error has moved roughly 1,816 BTC (~$116M) across more than 5,200 addresses, with Coinkite telling all affected users to migrate funds — too small to move price, but a live dent in self-custody confidence.

Hawkish. The effective rate is 3.63% inside a 3.50-3.75% target held 9-3 on July 29, with Hammack, Kashkari and Logan all dissenting in favour of a 25bp HIKE — dissents pointing up, not down. M2 is growing 5.53% year over year, positive but not a flood against a 4.70% 10-year. September 15-16 is the next live meeting and the July minutes land August 19. Live sentiment is Extreme Fear at 25 on the alternative.me gauge, but that print is stamped 2026-07-18 and is 17 days stale — it predates the entire equity rally and should be discounted, not traded.

The 10-year is 4.70%, up just 3bp over 14 sessions from 4.67% and in the upper half of a 4.61-4.75% range. The notable thing is what did NOT happen: VIX collapsed from 18.70 to 16.50, oil fell roughly 10% on the war-premium unwind — an outright disinflationary shock — and the 10-year still refused to rally. Risk-on with no duration bid. The supply explanation is dated and specific: Treasury expects $739 billion of privately-held net marketable borrowing in July-September, $68 billion above the May estimate, with refunding financing details due 8:30am ET on August 5. Yields pinned near 4.70% by issuance while the inflation impulse falls is a structural headwind for long-duration, non-yielding assets. There is no credit-spread feed in this dataset, so I make no claim about credit conditions.

DXY 99.835, down 1.51% over 14 sessions from 101.369 and holding below 100. Two things matter here. First, the timing: the entire decline occurred July 28-31 (101.380 to 99.803); the joint US-Japan yen-buying operation was reported August 3, and DXY actually ROSE that day to 100.010. The softness predates the intervention and cannot be attributed to it. Second, the character: the US joining the first coordinated yen operation since 1998 — driven by concern that a forced Japanese seller would blow out Treasury yields — is a plumbing-stress signal, not a liquidity-easing signal. A sub-100 dollar is normally a BTC tailwind; this particular sub-100 dollar arrived with BTC down 1.43% over the same window, so the reflex does not apply.

Risk-on, and specifically a supply-shock-relief rally. The S&P is at 7,736.52, up 4.43% over 14 sessions from 7,408.30, with VIX down from 18.70 to 16.50. The proximate driver is Trump confirming resumed Iran peace talks and calling off planned strikes, taking crude down more than 5% in a session and stripping out a war premium analysts had put near $30 a barrel. This is the cleanest risk-on impulse in months — and BTC fell 1.43% through it.

Base case over 5-7 days is a ±4.3% envelope from 64,218 (30-day realized vol 30.97% annualised), or roughly 61,500 to 67,000, which is effectively the existing 30-day range. The first real damage threshold is the 61,849 low, 3.7% below spot; below it there is very little structure until the 60-day low at 58,519, 8.9% below, and the 90-day range shows the market covering 82,243 down to 58,519 in a quarter, so that traverse is not slow when it happens. Two of the last 30 sessions produced 3%-plus moves, so a single adverse 3-5% day is inside the recent envelope and would on its own be enough to break support and flip the trend gate. The asymmetry worth naming: 30-day implied vol at the 0.3rd percentile means options are priced for the calm continuing across a refunding announcement, a payrolls print and a CPI print. If any of those three surprises, realised outcomes will be materially wider than the current pricing, and the cheap-vol regime that makes the range look safe is exactly what makes the break violent. Upside is more constrained than downside in magnitude — 66,257 is 3.2% up and the 200-day at 70,827 caps at 10.3% — so the distribution is mildly left-skewed even from a neutral stance.

low

Deribit 30-day implied vol is 34.04, the 0.3rd percentile of the past year — effectively the lowest implied reading of the year. Realised vol is compressing in the same direction: 7-day 28.29%, 30-day 30.97%, 90-day 35.59%. Skew at 0.0453 sits at the 43.33rd percentile (balanced, no crash bid) and the 30-90 term structure is in contango at +0.0588. Low, and compressed enough that the regime label itself is the risk.