ETF Flow Focus · Published

Bitcoin ETF Inflows Hinge on a $2.39B Week That Expires

US spot bitcoin ETF inflows open this week negative on every trailing window shorter than twenty sessions, and the twenty-session window itself is held up by a single week that is about to expire. Across the trailing twenty reported sessions through Friday, October 9, the twelve US spot funds took in a net +$1,954.1M — but +$2,385.8M of that arrived in one five-session stretch in late September, meaning that week alone accounts for 122.1% of the window's total. Strip it out and the other fifteen sessions are a combined -$431.7M.

That is the state of spot bitcoin ETF flows as this week opens, a different question from how last week finished. The completed week of October 5 through 9 landed at -$678.9M, covered session by session in Saturday's recap. What follows is the arithmetic that moves the trailing numbers over the next ten sessions whether or not a new dollar arrives, and which issuer has to show up for it to go the other way.

Where do bitcoin ETF inflows actually stand as this week opens?

Taking bryptoflows' rolling windows in order of length, from the live ETF flow dashboard:

The shape matters more than any one figure. The sixty- and twenty-session averages are nearly identical, at +$98.4M and +$97.7M a day, so the category's medium-run pace has been steady. The ten-session average has turned negative, and the five- and three-session averages are negative and getting worse. This is a demand curve that has rolled over at the short end while the long end has not yet noticed, and the only thing between the two is one week of September.

Cumulative net flow since launch now stands at $57,177.7M, some -$5,558.2M below the all-time peak of $62,735.9M. Year-to-date is +$583.5M, still positive for 2026 but thinner than at any point since it turned positive in late September.

The $2.39 billion week holding up the whole number

The week of September 21 through 25 delivered +$2,385.8M across five sessions, and it did so front-loaded: +$999.0M on the Monday, +$714.7M Tuesday, +$346.9M Wednesday, +$190.7M Thursday, +$134.5M Friday. That is a decay pattern, not a plateau — the week's internal trend was already fading by its final sessions, which is worth remembering before reading it as evidence of durable demand.

How unusual is it for one five-session block to carry more than the whole of a twenty-session window? Across the 475 positive trailing-twenty windows bryptoflows has on record since January 2024, 79 of them — 16.6% — were at least this dependent on their single best five-session block. So this degree of concentration happens, roughly one positive window in six. It is not an anomaly. What it does mean is that the window's positive sign carries very little information about current demand, because it is a statement about late September rather than about October.

The roll-off arithmetic for the next ten sessions

Here is the part that is pure mechanics and does not depend on forecasting anything.

This week, the trailing-twenty window sheds the week of September 14 through 18, which netted just +$6.1M. That week was close to flat on its own, two heavy outflow sessions offset by three inflow days, so its departure neither helps nor hurts. After five sessions the trailing-twenty will read roughly +$1,948.0M plus whatever this week delivers — short of a week worse than any in the category's history, it stays positive through Friday.

Next week, it sheds the September 21 through 25 week, all +$2,385.8M of it. That is the cliff. Expressed as a single threshold: the next ten reported sessions must net more than +$437.8M in total, or the trailing-twenty window turns negative. The arithmetic is exact: the trailing-twenty window two weeks from now consists of the ten sessions already on the books from September 28 onward, totalling -$437.8M, plus the ten yet to come. A negative trailing-twenty would be the category's first since the window ending July 28, 2026, at -$49.6M.

The ten-session window faces a smaller version of the same problem this week. It sheds September 28 through October 2, which netted +$241.1M. So this week has to beat +$241.1M just to stop the ten-session figure from deepening — and last week came in at -$678.9M. The five-session window, by contrast, gets a clean slate: it sheds the entire -$678.9M and will read exactly whatever this week prints.

Is BlackRock's bitcoin ETF still the engine?

This is the week's real question, and the honest answer is that nobody can tell yet.

BlackRock's IBIT supplied +$1,729.5M of the trailing-twenty window's +$1,954.1M, or 88.5% of the category's net over that stretch. Over the category's full life its dominance is starker still: cumulative net flow of +$65,733.2M is 114.96% of the category's whole, because Grayscale's legacy trust has bled so heavily on the other side. IBIT alone is 58.1% of all gross flow activity ever recorded.

Against that, last week IBIT inflows netted +$1.1M. Not negative — flat. The fund went -$207.7M on Wednesday, -$5.5M Thursday, and +$22.4M Friday, and the five sessions cancelled almost exactly to zero. Checking that against history: of 701 trailing five-session IBIT windows on record, only four have landed closer to zero, and all four were stretches in 2024 when IBIT reported no activity at all. Among weeks in which BlackRock's bitcoin ETF actually traded, this was the flattest five-session stretch it has ever had.

So the fund responsible for essentially all of the category's net demand spent last week doing nothing in either direction. Over the trailing twenty sessions IBIT averaged +$86.5M a session; at that pace, ten sessions would deliver roughly +$865M and clear the +$437.8M hurdle comfortably. At last week's pace it delivers about +$2M. The gap between those two outcomes is the entire story of the next two weeks, and there is no way to resolve it from flow data that ends on Friday.

Which funds are actually driving bitcoin ETF outflows

The selling is narrower than the aggregate suggests, and concentrated in two funds.

Breadth is narrowing, but read the zeros carefully

Friday's session showed 2 products positive, 1 negative and 9 flat, a positive share of 16.7%. That looks alarming and mostly is not: it is a reporting artifact as much as a demand signal.

Total gross activity across all twelve funds on Friday was $28.3M: $24.7M of gross buying against $3.6M of gross selling, spread across just three products (IBIT +$22.4M, HODL +$2.3M, FBTC -$3.6M). Of the 689 sessions on record in which any fund reported any activity, only six were thinner — Friday ranks in the thinnest 0.9%, and the last session thinner than it was September 11. Friday's product concentration came in at an HHI of 0.649 against a life-to-date 0.409, but that gap reflects how few products traded at all, not demand consolidating into fewer hands. Life-to-date concentration is unchanged, at a product HHI of 0.409 and issuer HHI of 0.441.

What the thinness does say is that nobody was transacting much in either direction on Friday — a weak foundation for any read on direction, bullish or bearish. Four of the last ten sessions were net negative.

How does this run compare with the category's own history?

Less dramatically than the week's headlines imply.

The trailing-ten figure of -$437.8M sits at the 24th percentile of all 696 ten-session windows on record — poor, but inside the normal range, and negative ten-session windows are routine: 35.2% of every such window since January 2024 has been negative. The trailing-twenty at +$1,954.1M sits at the 54.5th percentile, meaning the window everyone would call fragile is still slightly above the category's historical median; 30.8% of trailing-twenty windows have been negative. Friday's session carried a z-score of -0.20 and landed at the 45.8th percentile of 705 sessions — almost exactly median.

What would have to happen for this pattern to break

In rough order of how much each would move the trailing numbers:

One cross-reading: bryptoflows' composite institutional demand score ticked back up to 24.9 at Friday's close from the 17.4 Friday's signal piece logged a day earlier, returning the model to its moderate accumulation band. That recovery is thinner than it looks. The composite's heaviest input, at 24% weight, is ETF flow momentum, computed on exactly the twenty-session window above — +$1,954.1M, normalising to 47.7 on the model's scale. When the September week rolls off, that component falls whether or not institutional demand itself changes. Its price confirmation component puts bitcoin's trailing twenty-session return at -3.21%. Detail sits on the institutional signal dashboard.

Strategy's most recent 8-K, filed October 5, had direction acquired: 334 BTC for $28.7M at an average $85,928.14, taking holdings to 848,000 BTC. Thursday's column owns that subject, and last week's treasury piece has the funding detail.

What this flow data cannot tell you

These are reported daily net creations and redemptions sourced from Farside Investors. They measure share creation and destruction at the fund level, not investor intent: a redemption can be a market maker unwinding a basis trade as easily as a long-term holder selling, and the data cannot distinguish them, nor identify the end buyer.

Two reporting caveats. bryptoflows never substitutes zero for an unreported value — nulls stay null — but a reported 0.0 is ambiguous between no activity and no report, and on thin sessions like Friday's that carries real weight. And the dataset ends at the October 9 close, three calendar days stale as this publishes, with the feed flagged as delayed rather than current. The flow overview shows the live state.

Last Monday this column found three issuers supplying 114% of the category's net. A week later the concentration has not eased but narrowed, to one issuer supplying 88.5% of a window that a single week of September is holding above zero. The hurdle for the next ten sessions is +$437.8M — a low bar by the category's own standards, and one last week would have missed by a mile.