Bitcoin just booked its best month in a year — and celebrated by bleeding out on the final day. BTC ended July up roughly 7.5% per CoinDesk, surviving a hawkish Fed hold with three dissents, an AI-sector selloff, oil's biggest month since March, and a $38 million wallet exploit. Then, on Friday, July 31, it fell 2.9% to about $62,929 while the Nasdaq rose 1% to 25,373.85 and Amazon ripped 15.6% higher on its AWS blowout. As of Saturday, August 1, BTC trades near $63,000.
That one-day split is the tell. Two days after Bitcoin decoupled from a falling Dow on Fed day (BTC green, Dow −840), it decoupled again — in the opposite direction. When the equity tape is risk-off, Bitcoin has traded like a rate asset; when the equity tape is an AI melt-up, money leaves Bitcoin to chase it. This is what a marginal-buyer market looks like: flows, not conviction, set the price.
The flows said it first
US spot Bitcoin ETF flows told the rotation story in two acts. On Thursday, July 30, the complex pulled in +$233 million — its streak-breaking recovery day — with BlackRock's IBIT taking $183 million, or 78.7% of the total, per FinanceFeeds. CryptoSlate called the recovery 'an illusion propped up by BlackRock' — and Friday agreed: on July 31 the complex gave back $265.4 million, with IBIT shedding $122.7 million and Fidelity's FBTC $54.8 million, per Crypto Briefing (Friday's cell remains preliminary under our one-day settlement rule).
Where did the money go? The same session, Amazon gained 15.63%, Alphabet 7.12% and Nvidia 3.46%; the S&P 500 closed at 7,489.72 (+0.7%) and the Dow at 52,485.03 (+0.53%). Coinbase — the crypto-native equity — fell as much as 14% intraday on its Q2 miss before closing down roughly 10%, per The Motley Fool. Risk appetite wasn't missing on Friday. It was simply pointed at AI infrastructure, not at crypto.
The mechanics of the rotation deserve spelling out, because 'money rotated' is usually hand-waving. Three concrete channels were visible this week. First, the month-end rebalance: institutions running target allocations trim whatever outperformed intramonth — and through July 30, that included Bitcoin — while adding to whatever they're benchmarked against, which after Amazon's guidance is increasingly AI-weighted equity indices. Second, the derivatives channel: Friday's July monthly options expiry forced market makers to unwind delta hedges built around the $64,000–66,000 strikes where open interest clustered, removing a standing bid precisely as spot weakened. Third, the narrative channel: with three Fed dissents pushing September hike odds to ~81% and the CLARITY Act's summer window closing, per analysts cited in Friday's wraps, the two catalysts crypto bulls spent July waiting on — legislative clarity and a dovish pivot — both slipped to autumn. Equities offered a catalyst that had already arrived: $220 billion of confirmed Amazon capex. Markets pay for certainty first.
Friday's Coldcard exploit added a fourth, subtler channel. CoinDesk's hypothesis — that a $38M self-custody failure pushes marginal holders toward ETF custody — would be flow-positive for IBIT over weeks. But on day one it read as a pure risk headline, and Forbes reported outright 'crash fears.' Watch whether the scare converts into ETF inflows (constructive) or simply into smaller aggregate exposure (not).
Grading the markers
We grade every marker against its printed definition in the article where it was set. Scores below reflect Friday's close and the latest settled flow data.
| Marker | Printed definition | Verdict (Aug 1) |
|---|---|---|
| J1 (set Jul 30) | September hike odds ≥75% (hawkish branch) or <65% (soft branch) by Friday close | HAWKISH BRANCH FIRED. CME FedWatch printed ~81% on July 31 (versus 72.3% on July 29); other trackers showed ~82% by mid-September. No print near the 65% soft threshold. |
| J2 (set Jul 30) | July 29 + July 30 ETF cells sum negative (final Aug 3) | FAILED — settled early by arithmetic. Jul 29 closed +$32.1M and Jul 30 +$233M; the sum cannot go negative. |
| J3 (set Jul 30) | Strategy announces a new capital instrument within 24h of earnings | FAILED at Friday close — no new instrument beyond the previously announced framework. The separate H1 "another color" window stays open to Aug 10. |
| K1 (set Jul 31) | September odds ≥75% at Friday Aug 7 close | OPEN — tracking at ~81%, above threshold with a week to run. |
| K2 (set Jul 31) | MSTR closes every session inside $85–110 through Aug 7 | DAY 1 HOLD. Friday prints diverge across data providers ($92.85 vs $93.97 after an earnings-driven ~4% intraday drop) — both inside the band. |
| K3 (set Jul 31) | July 31 + Aug 3 ETF cells sum positive | OPEN but STEEP — Friday's preliminary −$265.4M means Monday needs a >+$265M print, which July produced exactly zero times. |
Also graded: Thursday's committed watch on Amazon's after-hours pop. It didn't fade — it doubled, closing +15.63% with AWS growth of 37% and a ~$220 billion capex guide behind it. That confirms the AI-infrastructure trade's leadership and is precisely the bid Bitcoin competed against on Friday.
Marker-book scorecard, one month in
With the J series now closed, the running tally since we began printing falsifiable markers in early July: flow-based markers (the F/G/I/J-flow family) have skewed heavily toward outflow or fail verdicts — the marginal-buyer thesis keeps grading well — while rate-odds markers have fired hawkish in every window since the July 23 oil spike. The clean lesson from J2's early arithmetic failure: two-cell sum markers resolve faster than their formal deadlines, and we'll grade them the moment they become arithmetically settled rather than waiting for the calendar. K3 may meet the same fate Monday.
The month in one paragraph
July's ledger: +7.5% price gain; a Fed hold with the most hawkish dissent cluster since 2019 and September hike odds ending near 81%; core PCE cooling to 3.3% y/y while oil logged its biggest month since March; Strategy confirming its largest-ever BTC sale and an $8.22B quarterly loss; ETF flows netting roughly +$438M for July 1–30 per Farside data — barely positive across 22 sessions; and, on the final day, the Coldcard exploit draining $38M from self-custody wallets. That Bitcoin gained 7.5% through all of it is the bull case. That every rally kept dying at the marginal ETF bid is the bear case. Both were true all month.
New markers — the L series (grade by Aug 8)
- L1 — Stolen-coin watch: the consolidated 562 BTC Coldcard-theft address moves any coins by Friday, Aug 8. Movement toward exchanges would pressure sentiment; continued dormancy defuses the story.
- L2 — Reclaim test: BTC posts any daily close above $65,000 by Thursday, Aug 7. Fires = rotation was month-end noise. Fails = the AI trade is structurally out-competing the ETF bid.
- L3 — IBIT pulse: IBIT prints at least one +$100M inflow day during the Aug 3–7 week. IBIT was the entire recovery on Jul 30; if the lone marginal buyer doesn't return, K3 and L2 likely fail together.
What to watch next week
- Monday Aug 3: J2 formal settlement, K3 first cell, and the F3/G3 ETH out-draw checks from earlier series.
- Through Aug 7: K1 and K2 finals; miner Q2 reports begin (AI/HPC segmentation watch).
- Aug 10: H1 "another color" window closes for Strategy.
- ~Aug 11: difficulty retarget — tests July's hashrate-floor thesis.
- Ongoing: Coldcard fallout — further Coinkite advisories, the L1 address, and whether the self-custody scare shows up as ETF inflows, as CoinDesk hypothesized.
Disclaimer: This article is for informational and educational purposes only and does not constitute investment, financial, legal, or security advice. Cryptocurrency investments are volatile and carry a high risk of loss. Always do your own research and consult a qualified professional before making investment or custody decisions.