Bitcoin ETF Inflows Faded 87% in Five Straight Sessions
Bitcoin ETF inflows are still positive, and smaller every session
US spot bitcoin ETF inflows open this week on a seven-session positive streak that has been shrinking for five straight days. The twelve US spot funds took in +$134.5M on Friday, September 25 — the seventh consecutive positive session, and also the fifth consecutive session smaller than the one before it, down from +$999.0M on Monday, September 21. In the 695 reported trading sessions bryptoflows has recorded since January 2024, that five-session fade is the longest unbroken decline inside a positive streak on record.
Both halves of that sentence are true at once, and they point in opposite directions. The streak itself is real money: +$2,385.8M arrived across the week of September 21 through 25, enough to push 2026 year-to-date flow positive for the first time. But every session of that week was worth less than the one before it, and by Friday the category was running at roughly one-seventh of Monday's pace. Nothing in the data says which of those facts the coming week will honour.
A five-session fade has never happened before
The shape of last week is unusual in a way the weekly total conceals. Session by session: +$999.0M, +$714.7M, +$346.9M, +$190.7M, +$134.5M. Each print retained 71.5%, 48.5%, 55.0% and 70.5% of the one before it. Friday's net flow was 13.5% of Monday's — a decline of 86.5% across four steps, with no reversal at any point.
Runs like this are rare. Across all 695 sessions, bryptoflows counts only eight stretches of four or more consecutive declining positive sessions, and last week is the only one that ever reached five. The seven prior four-session runs were February 2024, June 2024, July 2024, March 2025, April 2025, October 2025 and May 2026 — roughly one every four months, and none of them made it to a fifth.
What the fade is not is weakness in absolute terms. Friday's +$134.5M carried a z-score of +0.18 and sat in the 62nd percentile of all 695 sessions on record. The category's weakest day in a record-setting week was still an above-median day by its own history. The trailing averages say the same thing from a different angle: three sessions at +$224.0M a day, five at +$477.2M, ten at +$239.2M, twenty at +$135.7M and sixty at +$111.8M. The near-term numbers sit above the long-term ones, which is what a decelerating rally looks like rather than a reversal.
One number worth holding onto: the ten-session window totals +$2,391.9M against the five-session window's +$2,385.8M. The five sessions before last week — September 14 through 18, the Fed-decision week — contributed +$6.1M between them. Essentially all of the category's recent demand arrived inside one week.
Does a fading streak mean bitcoin ETF outflows are next?
This is the question the pattern invites, and the honest answer from bryptoflows' own record is that it does not forecast. Of the seven prior four-session declining runs, the session that followed was positive four times and negative three times: +$80.0M, -$64.9M, +$310.1M, +$89.6M, +$591.2M, -$4.5M and -$268.5M. That is a coin flip with a sample of seven. A decelerating streak is a description of what already happened, not a signal about what comes next, and anyone presenting it as one is reading more into it than the data carries.
A separate pattern in 2026 is somewhat sharper, and worth stating with its limits attached. This is the fifth streak of seven or more consecutive positive sessions the category has produced this year. The four that already ended — March 9 to 17, April 14 to 24, July 14 to 22, and August 17 to 27 — each ended not with a drift to zero but with a triple-digit outflow day: -$163.5M, -$263.2M, -$225.1M and -$201.9M respectively. In every 2026 case so far, the break was abrupt. That is four observations, all from one year, and it says nothing about when. It does suggest that if this streak ends, the first negative session is more likely to be a real one than a rounding error.
Two funds did all of Friday's work
Breadth narrowed sharply as the week wore on, and that is the part of the fade that shows up in the issuer data rather than the headline. On Monday, September 21, seven of twelve products posted a positive flow and none posted an outflow. By Friday, only two were positive — BlackRock's IBIT at +$97.0M and Fidelity's FBTC at +$49.3M — with Bitwise's BITB at -$11.8M and the remaining nine products flat at zero. That is a positive-breadth reading of 16.7%, on full 100% reporting coverage across all twelve known products.
Friday's gross figures make the same point: $146.3M of gross inflow against $11.8M of gross outflow. Almost nobody sold, and almost nobody bought. The day's product concentration measured 0.479 on bryptoflows' 0-to-1 scale, against the 0.083 an even twelve-way split would produce — the demand that remained was pooled in two funds.
A flat reading here means a reported zero, not a missing one. bryptoflows never substitutes zero for an unreported value, and Friday's session had every product reporting. Independent coverage of the same session put the category at $134.46M with IBIT at $96.99M, FBTC at $49.32M and BITB at -$11.85M, matching bryptoflows' own figures to the cent.
BlackRock's bitcoin ETF is carrying the whole year
Across the full week, IBIT took +$1,157.6M, or 48.5% of the category's net. FBTC added +$701.6M (29.4%), ARK 21Shares' ARKB +$294.7M (12.4%) and Morgan Stanley's MSBT +$203.3M (8.5%). Over the trailing twenty sessions the concentration is heavier still: BlackRock at +$1,883.8M of the category's +$2,714.0M, or 69.4%, with Fidelity at +$825.6M and Morgan Stanley at +$264.1M the only other issuers above $10M.
Stretch the window to the full year and the dependence is close to total. The category's 2026 year-to-date net flow is +$1,021.3M. IBIT's alone is +$3,189.7M. The other eleven products, taken together, are -$2,168.4M on the year. The BlackRock bitcoin ETF is not leading the category in 2026 so much as constituting it.
Fidelity is the sharpest illustration. FBTC just posted its strongest week of the year at +$701.6M, ran a six-session positive streak into Friday, and is still -$1,051.2M year-to-date. One exceptional week did not undo what the first half of 2026 did to it. The same arithmetic scales up: the category sits +$57,615.5M cumulatively since January 2024 but $5,120.4M below its own peak, and quarter by quarter, 2026 reads -$495.8M, -$4,894.0M, then +$6,411.1M so far in Q3. The ETF flow dashboard carries the full per-product history behind these figures.
GBTC has not posted an outflow since September 16
Grayscale's GBTC — the product whose redemptions defined the category's first two years, at -$27,841.4M lifetime across 356 negative sessions out of 679 reported — has now gone seven sessions without a single outflow, its last being -$18.2M on September 16. Twelve of its last twenty sessions printed exactly zero.
Two caveats before anyone reads a turning point into that. The prior twenty sessions contained thirteen zero days, so dormancy is GBTC's recent normal rather than something new. And the fund's longest run without an outflow remains ten sessions, ending April 24, 2025 — the current run is notable but not a record. GBTC is still -$2,586.9M year-to-date.
What has changed is mechanical, and it matters for reading the sessions ahead. For most of the category's history, a negative day could arrive purely from Grayscale redemptions with nobody selling anywhere else. With GBTC inert, a negative print now essentially requires IBIT or FBTC to actually turn. That raises the bar for breaking the streak — and it also means that when the streak does break, the cause will be more informative than it used to be.
Morgan Stanley's MSBT has one negative day in 119
The most under-covered product in the category is MSBT, which first reported on April 8, 2026. In 119 reported sessions it has posted 70 positive days and exactly one negative day — -$5.3M on May 29. Its lifetime net flow is +$775.8M, all of it accumulated this year, making Morgan Stanley the second-largest positive contributor to 2026 year-to-date flow after BlackRock, ahead of every other issuer including Fidelity.
MSBT contributed +$203.3M last week, concentrated in Monday and Tuesday, then tapered to +$10.2M Thursday and zero Friday. Whether the near-total absence of outflows reflects a genuinely different holder base or simply a young fund that has not yet met a bad week, this data cannot say — 119 sessions is not a long record, and it does not include a drawdown of the kind Q2 delivered to everyone else.
What would have to happen for the streak to break
The bar is low in dollar terms and the calendar is heavy. Friday's +$134.5M is small enough that a single ordinary IBIT outflow — the fund has 165 negative sessions in its history — would end the run on its own, with the other eleven products behaving exactly as they did Friday.
The scheduled US data this week runs JOLTS and consumer confidence on Tuesday, September 29, the PCE price index and the third estimate of Q2 GDP on Wednesday, September 30, ISM manufacturing on Thursday, October 1, and the September employment report on Friday, October 2. Flow data cannot establish that any of these moved the category, and this article does not claim they will — but ETF creations are reported on a one-session lag, so a reaction to Wednesday's PCE print would land in Thursday's numbers rather than Wednesday's.
bryptoflows' own institutional support signal has already eased with the fade, reading 45.0 on the September 25 close against 49.7 a session earlier, with its regime still classified as moderate accumulation. The flow-breadth component is the piece that deteriorated: two positive products out of twelve, against seven on Monday.
What this flow data cannot tell you
Net flow is the dollar value of creations minus redemptions. It is not a measure of conviction, and it cannot distinguish between fewer buyers and the same buyers writing smaller tickets — which is precisely the ambiguity at the centre of a five-session fade. Both produce an identical decay curve.
A zero is a reported zero, but a fund reporting a flat net day may still have had offsetting creations and redemptions inside it. The data shows the net, not the churn. It also never identifies who is on the other side of a flow: an authorised participant's creation says nothing about whether the end holder is a pension allocator, a basis trader unwinding, or a retail flow aggregated through a platform.
Finally, flows are not price. Bitcoin closed at $84,066 on September 25, and the overview dashboard tracks the two series together, but nothing in the flow record establishes causation in either direction. For the completed week in isolation, the weekly recap covers what arrived and where it went; last Monday's flow focus examined the single-fund concentration that started this run, and Friday's institutional read covers the composite in full. The question this week answers is a narrower one: whether five straight declines were the end of a rally or a pause inside it. The record does not know yet, and neither does anyone quoting it.