On the final trading session of October, US spot Bitcoin ETFs printed a net outflow of $643.2 million. Three funds absorbed $564.2 million of it โ 87.72% of the day's total. BlackRock's IBIT alone accounted for $233.7 million, or 36.3% of the redemptions.
I ran the identities before I read the analysis. Daily components sum to the stated total to the decimal place. The top-three share recomputes to 87.72% against a reported 87.7%. The weekly net of โ$202.4 million is exactly the four-day inflow aggregate of $440.8 million minus the Friday outflow of $643.2 million. IBIT's weekly balance of +$176.9 million reconciles against $410.6 million in and $233.7 million out. The calendar rows resolve in order โ 10/26 through 10/30 land on Monday through Friday, with no orphaned sessions.
Every check passes. That is precisely the problem. Ledger lines bleed, but the arithmetic never lies โ and here the arithmetic is telling a story the coverage around it did not tell. The $643.2 million print was accurate, internally consistent, and structurally incomplete. What follows is a decomposition, not a correction.
The mechanics first, because they determine what the number even means. A spot ETF flow figure is not a measure of buying pressure. It is the net of creation and redemption activity in the primary market, executed by authorized participants against baskets of the underlying. The disclosure lands once per day, after the close, with settlement trailing behind it. Market makers manage inventory in the secondary market, and that inventory work generates daily swings of hundreds of millions of dollars with no directional content whatsoever. An AP that built a position on Wednesday will redeem on Thursday because its book is full, not because its view changed.
Against a $57.05 billion cumulative net inflow anchor, a single-day print of $643.2 million is 1.13%. That is inside ordinary inventory noise for an asset with this volatility profile. The figure is large in absolute terms and small in structural terms, and the coverage treated it as a regime change.
The data itself comes through a narrow pipe. Farside Investors is the de facto standard aggregator for these tables, assembled by hand from issuer disclosures published once daily. Every headline figure in circulation โ the $643.2 million, the $233.7 million IBIT line, the $53.3 million GBTC line, the weekly and monthly aggregates โ traces back to that single table. No issuer's official flow statement was cited. No cross-verification against a second aggregator appears anywhere in the record.
That is not a reporting failure; it is the industry's actual infrastructure. When I built our real-time ingestion framework in 2024, wiring Glassnode and CryptoQuant feeds into the desk models and compressing latency from hours to seconds, the one table I could never automate was the flow table. It has no API. It has no canonical schema. It is a human transcription of issuer PDFs, and it gets revised.
Provenance is the only proof of value. Every downstream conclusion inherits the reliability of that single hand-assembled table, and every reader should price that in before quoting it.
There is a second-order defect: rounding. The monthly figure of โ$515.9 million is built by summing daily values that have each already been rounded. Across roughly twenty-one trading sessions, cumulative error reaches the ยฑ$10 million range โ call it 2%. The source disclosed this. That is honest handling, the kind of discipline I built into my audit checklists in 2017 and still apply. It also means the monthly flip from positive to negative is not statistically significant at the scale the coverage implied.
Now the decomposition. Three series matter.
Series one โ inflow concentration. Across the first four sessions of the week, the complex absorbed $440.8 million. IBIT took $410.6 million of it. That is 93.2%. Every other fund in the complex, combined, took $30.2 million.
Series two โ outflow share. On Friday, IBIT redeemed $233.7 million against a $643.2 million market total. That is 36.3% โ well below its inflow share.
Series three โ the offset ratio. Weekly, IBIT's +$176.9 million offset 46.6% of the $379.3 million the rest of the complex bled. Monthly, IBIT's +$932.1 million offset 64.3% of the $1.45 billion drained from everything else.
Put the three together and the picture inverts. October was not a withdrawal from Bitcoin ETFs. October was a consolidation into IBIT, interrupted by a redemption wave that exceeded IBIT's absorption capacity. The offset ratio decayed from 64.3% monthly to 46.6% weekly. That decay, not the $643.2 million headline, is the thing worth tracking.
I want to be explicit about a number nobody computed: 93.2% is not "demand remains concentrated." 93.2% is a monopole. In a six-issuer complex with meaningful fee dispersion, one product capturing more than nine of every ten dollars of gross weekly inflow is a structural fact about distribution, not a passing trend. It deserved its own paragraph in every article written that week. It got a clause.
Then the timeline, which is where the published narrative and the ledger diverge.
The frame was: October was positive, one day erased it, October turned negative. Technically true. Also selective.
Reconstruct. The month closed at โ$515.9 million. The final week contributed โ$202.4 million. Therefore 10/1 through 10/23 cumulated to โ$313.5 million. Within that window, 10/22 alone printed โ$635.6 million. So the residual โ 10/1 through 10/21 plus 10/23 โ sums to roughly +$322 million.
For the first three weeks of October, the complex was already in net outflow of $313.5 million. The four-day rebound of $440.8 million briefly pushed the monthly total positive by $127.3 million โ a margin under 0.3% against the $57.05 billion cumulative anchor. That positive status survived roughly two trading sessions before Friday's print erased it and added $202.4 million of fresh selling on top.
The published framing โ "still positive before the last session" โ is accurate and steers perception in a specific direction. It implies a stable month punctured by a single event. The ledger shows an unstable month briefly rescued by four days of buying, then re-punctured. The "reversal" is an artifact of where you draw the window.
One more scale correction. Four days of inflow totaled $440.8 million; the Friday outflow was $643.2 million. The outflow did not erase the week's gains โ it erased them and added $202.4 million of net selling. The word choice was conservative in the wrong direction.
Quantify the severity while we are here. October's $515.9 million of Bitcoin ETF outflow equals 0.90% of the $57.05 billion accumulated since launch. The coverage said the month "did not erase cumulative capital." True, and understated. At 0.90%, the month is a rounding event against a three-year accumulation curve, not a test of it.
Now the blind spot.
The Ethereum complex printed roughly โ$1.61 billion for the month. Bitcoin printed โ$515.9 million. That is 3.1ร. On the weekly frame: Ethereum โ$431.4 million against Bitcoin's โ$202.4 million, or 2.1ร. And Ethereum's five sessions were negative on all five days, with Friday contributing โ$155.7 million. Not one green print all week.
October's most extreme ETF flow signal was Ethereum, and it appeared in two paragraphs near the end of a twenty-five-paragraph piece whose headline named BlackRock. The reason is structural: a headline needs conflict, and "BlackRock bucks the trend" supplies it. "Ethereum bleeds five for five" does not. Conflict outranked materiality in the edit.
That ranking has a cost. A reader who takes away "the October ETF problem was a month-end Bitcoin redemption" has the direction wrong. The October problem was a sustained, uninterrupted exit from Ethereum products that the Bitcoin complex partially masked through issuance concentration.
The necessary caveat, also omitted: Ethereum ETF assets under management are far smaller than Bitcoin's. A $1.61 billion outflow may be a larger relative shock or a much smaller one. Without per-product AUM, the comparison cannot be normalized. That is the one field that would have made the Ethereum story actionable, and it is missing.
Third item: GBTC.
Grayscale's trust carries a fee near 1.50%. IBIT charges 0.25%. Six times. Every day that spread exists, a rational long-horizon holder migrates. GBTC's $53.3 million redemption on Friday is not sentiment โ it is a fee arbitrage executing on schedule. The monthly GBTC decomposition was never published, and that single line item is what separates structural migration from demand contraction. Lumping GBTC into a redemption table beside FBTC and ARKB invites readers to read a fee schedule as a mood.
Yields are illusions until the vault is open โ and here the vault is a 125-basis-point gap doing exactly what fee gaps do.
Here is where I diverge from both the coverage and the reflexive pushback against it.
The published article was honest about its limits: flow data cannot say why investors redeemed, and cannot say whether capital leaving one product entered another. That boundary statement is correct, and rare enough to respect.
But two competing hypotheses explain the same 93.2%/36.3% asymmetry, and only one was presented.
Hypothesis A โ IBIT demand is structurally stronger. Evidence: 93.2% of inflow share, 64.3% monthly offset.
Hypothesis B โ liquidity-first liquidation. When an institution de-risks, it sells the deepest, tightest-spread instrument first. IBIT is the largest and most liquid Bitcoin ETF. Its outsized creation share and its outsized redemption share are one fact seen from two sides: IBIT is where institutions transact. Under this reading, 36.3% is not "bucking the trend" โ it is the mechanical consequence of being the market's exit door.
Both fit the data. The coverage ran Hypothesis A because it produced the cleaner story. Correlation is not causation, and in flow data the causal direction is usually the one the narrators declined to pick.
The missing variable that costs nothing: price. The piece never stated what Bitcoin did on 10/30, never mentioned a macro calendar event, never cited ETF-related news. Flow without price is uninterpretable โ profit-taking, risk-off, and rebalancing are indistinguishable in a table that only counts shares created and destroyed. Price data is free, same-day, and requires no investigation. Its absence is the most consequential gap in the piece, not because it is hard, but because it is easy.
The evidence chain compounds the problem. Issuer primary data: zero citations. Aggregator: one source, four data points, no cross-check. Third-party analysis: a same-source restatement, not an independent voice. Author's own arithmetic: roughly twenty derived figures, and to the writer's credit, all of them check out. The computation is trustworthy; the inputs are unverified. That distinction never made it into the copy.
Four fields would have made the piece reusable, ranked by cost. Price on the day and for the week โ free, and it closes the largest explanatory hole. Per-fund AUM โ one lookup each, and it puts $643.2 million in correct proportion. GBTC's monthly standalone line โ the difference between migration and decay. And Ethereum's per-fund breakdown, currently a blur, despite being October's single largest flow event. The first two are the ones that should have been there without anyone asking.
The observation calendar for next week is short. Watch the offset ratio first: if IBIT's weekly absorption drops below 40%, the consolidation thesis breaks and the redemption wave becomes the real signal. Watch GBTC's monthly line, because a stable migration rate means the remainder of the outflow is genuine. Watch Ethereum's streak, because a sixth consecutive negative session on a smaller AUM base is a different animal.
The $643.2 million will be forgotten by Tuesday. Structure dictates survival in the digital wild โ and the structure here says the complex is narrowing, not emptying. The question worth asking is not how much left, but how few doors remain.