ETF Flow Focus · Published

Bitcoin ETF Inflows Stall as IBIT Drives 72% of Selling

US spot bitcoin ETF inflows open this week with last Monday's open question answered, and answered badly. This column asked seven days ago whether BlackRock's IBIT — the fund that has supplied the category's entire net demand — was still the engine, and noted that its previous week had cancelled almost exactly to zero. It has now picked a direction: IBIT took -$165.4 million out of the category over the five sessions through October 16, which is 72% of the week's total -$228.4 million. The hurdle for keeping the twenty-session figure positive has risen from +$437.8 million over ten sessions to +$666.2 million over the five that remain.

All figures below come from bryptoflows' own API, sourced from Farside Investors, current through the October 16 close with zero missing sessions — a change from last Monday, when the feed was flagged delayed and three days stale.

Why bitcoin ETF inflows now need $666 million in five sessions

Last Monday's column set out the arithmetic: the twenty-session window was about to shed the week of September 21 to 25, worth +$2,385.8 million, so the ten sessions then ahead had to net better than +$437.8 million or the window would turn negative for the first time since July 28.

Five of those ten sessions are now on the books, and they came in at -$228.4 million. The remaining hurdle is therefore exactly +$666.2 million across this week — an average of +$133.2 million a session. For scale, one of the last ten sessions cleared that daily figure: October 13, at +$175.7 million.

The twenty-session net currently reads +$1,719.6 million. If this week nets exactly zero, it prints -$666.2 million. If it repeats last week, roughly -$894.6 million. That outcome needs no new selling at all, which is the single most important thing to understand about the number that will be quoted everywhere by Friday.

BlackRock stopped being the engine and became the brake

Last week IBIT's five sessions ran -$65.4 million, +$65.9 million, -$110.9 million, +$16.8 million, -$71.8 million: three negative, two positive, -$165.4 million net, or -$33.1 million a session.

The comparison that matters is against the pace this column cited last Monday. Over the then-trailing twenty sessions IBIT had averaged +$86.5 million a session, a rate that would have cleared the whole ten-session hurdle on its own inside six days. The actual week came in roughly $120 million a session below that pace. The swing, not the absolute size, is the finding.

Because by IBIT's own standards the week was unremarkable. Of 706 five-session windows on record, 181 (25.6%) have been negative and 129 (18.3%) were more negative than this one — the most recent being the window that ended on October 13. A $165 million outflow week is normal for a fund this size. What is not normal is the category having nothing else to lean on while it happens.

That dependence is unchanged. IBIT's lifetime net of +$65,567.8 million is 115.1% of the category's +$56,949.3 million, because Grayscale's two vehicles have returned $25,005.1 million between them. The live split sits on the ETF flow dashboard.

Fidelity was the one thing that went right

The counterweight, and it is a real one: Fidelity's FBTC decelerated hard. Over October 5 to 9 it ran -$380.3 million, or -$76.1 million a session. Over October 12 to 16 it ran -$60.2 million, or -$12.0 million a session — roughly 84% less drag, with one positive session (+$103.1 million on October 13) inside it.

That was the second of the three conditions this column named last week, and it is the only one that materialised. It was not enough, because the arithmetic last Monday predicted for FBTC still lands this week: its twenty-session figure of +$93.2 million exists only because it took in +$701.6 million during the September week. When that rolls off, FBTC's twenty-session contribution becomes -$608.4 million on arithmetic alone.

The stretch still ranks badly in Fidelity's own history. FBTC's ten-session net of -$440.5 million sits in the worst 7.4% of all 701 ten-session windows it has recorded, against a record of -$1,044.4 million set on January 29 this year. Its 2026 total is -$1,659.6 million — the category's second-largest source of redemptions this year, behind only GBTC — against a lifetime net of +$10,464.6 million built almost entirely before this year.

The handover, stated plainly

Put the two windows side by side and the rotation is exact. The ten-session net of -$907.3 million breaks down as:

IBIT's ten-session loss of $164.3 million is slightly smaller than its five-session loss of $165.4 million — meaning BlackRock was net +$1.1 million across October 5 to 9 and did all of its damage in the week just ended. Fidelity ran the mirror image.

So early October was a Fidelity story and last week was a BlackRock story. The category's aggregate barely registered the difference, which is precisely why the aggregate is the wrong number to read. This is a handover of the selling, not an intensification of it — and it means the two largest funds have now each had their turn, with nothing left in reserve behind them.

The active session this column asked for has not arrived

The third condition named last week was not a number but a shape: a session in which eight or ten funds report non-zero flows and most of them are positive. The most active session last week had three. The last session in which eight or more funds reported any movement was October 1 — eleven sessions ago.

Counting funds that reported a non-zero flow, the per-session averages:

Six of twelve funds did not move a dollar last week: ARKB, EZBC, BRRR, HODL, BTCW and GBTC all reported exactly $0.0 million across five sessions. On October 12 a single fund reported anything at all. Valkyrie's BRRR has not reported a non-zero session in sixty; WisdomTree's BTCW has reported one in twenty. Total gross activity across all twelve funds for the entire week was $617.8 million.

Read the zeros with the care last week's piece urged — a reported 0.0 is ambiguous between no activity and no report. But the trend across five different window lengths is consistent, and the institutional signal model's breadth component now scores -33.3 on its normalised scale off exactly these sessions. The composite itself reads 19.2 this morning with agreement at 64.4, down from the 28.9 Friday's signal piece logged on October 15 data, after the October 16 session was added.

Scoring last Monday's three conditions

One of three, and the one that landed was the one with the least capacity to move the total. That is the honest scorecard.

Where -$907 million sits in the category's own history

Less dramatically than it sounds, which is worth stating plainly:

The genuinely interesting structural fact: a negative ten-session net sitting underneath a positive twenty-session net has occurred on only 12.6% of sessions in the category's history. It is the shape a trend takes while it is turning, in either direction.

What a flat week does to each issuer

Running the same zero-flow projection fund by fund shows where the category's apparent strength actually lives. If this week nets zero across the board, the twenty-session figures become:

The headline number is one fund's trailing total minus everybody else's redemptions. MSBT remains the quiet exception on the other side — still essentially without a redemption history, as last week's column noted — but its October tickets have been +$7.0 million, +$2.4 million, +$7.8 million and +$6.7 million, against +$61.7 million and +$99.0 million on two days in late September. The fund that never sells is also barely buying.

What would change the read this week

The bull case is specific and testable. +$666.2 million in five sessions is not a rare week by the category's own standards: 35.0% of all 706 five-session windows have cleared it, including 50 of this year's 199 sessions, most recently the window ending September 29. It requires IBIT back near the +$150 million-a-session cadence it ran in late September, and it requires at least one of the six dormant funds to rejoin.

The bear case requires nothing. A flat week delivers a negative twenty-session print on arithmetic alone.

The two things that move before the aggregate does: the sign on IBIT's daily print, and the non-zero fund count climbing back above four. Both are visible on the overview dashboard a week before the trailing windows show them.

What this flow data cannot tell you