Institutional Signal · Published

Institutional Bitcoin Demand Score Falls 26 Points to 28.9

Are institutions buying bitcoin right now?

Yes, but through a channel that has narrowed sharply in five sessions. bryptoflows' composite institutional bitcoin demand score reads +28.9 on a -100 to +100 scale as of the September 10 close, down from +55.2 a week ago. All eight components were computable on both runs — the missing-components list is empty and weight coverage is 1.0 — so the two readings are directly comparable, which is the first thing worth establishing before anyone reads meaning into a 26-point fall.

The fall is real. What it measures is not what most people would guess. The size of institutional bitcoin buying barely changed this week. What changed is how many funds were doing it, and for how many days in a row.

What took the institutional bitcoin demand score down 26 points

The composite fuses eight components, each normalised to -100..100 and weighted. Here is the full contribution breakdown as of Thursday's close, in points of the final 28.9, with last Friday's contribution alongside:

Four of the eight components moved by less than half a point. The 26-point decline comes almost entirely from three: breadth (about -14), persistence (about -11) and price confirmation (about -5), offset by roughly +4 from treasury accumulation. The model's own agreement figure — which measures how much the eight components corroborate each other, not a confidence interval around the score — fell from 61.0 to 34.3. That is the honest summary of this week: the components have stopped agreeing.

The regime label still reads moderate accumulation, unchanged from last week. A label that survives a halving of the underlying score is a reminder that the band is wide, and that the number is more informative than the name attached to it. The live breakdown sits on the institutional signal dashboard.

Only one of twelve funds bought on Thursday

Thursday, September 10 was the category's weakest breadth reading of the period. Net flows were -$282.7M, a z-score of -1.24, in the 10.5th percentile of all 684 sessions on record. Gross inflows across the entire twelve-product shelf totalled $4.0M against $286.7M of gross outflows.

That $4.0M was a single fund. Morgan Stanley's MSBT was the only product with a positive flow. The rest of the session, by bryptoflows' data: ARKB -$164.3M, GBTC -$36.4M, FBTC -$33.6M, IBIT -$24.5M, HODL -$15.3M, BITB -$12.6M. The three-session total from September 8 through 10 is -$449.5M, figures that match independent reporting on the same sessions.

One detail deserves stating because the component does not state it. Breadth counts a product as non-positive if it reports zero, and five of the twelve reported exactly zero on Thursday — so the day was one positive, six negative and five flat, not one against eleven sellers. The component reads -83.3 because it divides positives by all reported products, which pushes a day with many flat funds close to the floor. That construction is defensible, but a reader comparing -83.3 to a day of genuine broad selling would be reading it as worse than it was.

The more consequential fact is which fund sold. IBIT has supplied essentially all of the category's net demand — Monday's flow piece put the eleven non-BlackRock products at a barely-positive +$47.1M over 60 sessions while the category ran +$2,076.9M. On Thursday BlackRock's fund was itself a net redeemer. That is a different configuration from the one this site has described for most of 2026.

The two components carrying the score did not move at all

Here is the tension at the centre of this week's reading, and the reason the score is not lower.

Supply absorption and ETF flow momentum carry 20% and 24% weight respectively — 44% of the composite between them — and together contribute +36.2 of the 28.9. Both are computed on a 20-session window. Both are almost exactly where they were last Friday. Meanwhile every component that reads the last few days went sharply negative.

That is not a contradiction in the data. It is a statement about window length. The 20-session window still runs from August 13 and still contains the strong second half of August and September 3's +$730.8M session. The shorter windows tell the opposite story: trailing three sessions -$449.5M, trailing five +$455.9M, trailing ten +$577.6M, trailing twenty +$3,192.2M. Each longer window is larger than the one inside it, which means demand has been decaying steadily from the front. Year-to-date net flow remains negative at -$1,357.4M even after a strong quarter of +$4,032.4M.

So 44% of this composite is currently reporting a four-week average that the most recent three weeks no longer support. That is not a flaw to be corrected — a trailing indicator is supposed to trail — but it is the single most important thing to understand about a 28.9 reading.

The 20-session window is holding up a score the last three weeks no longer support

This part is arithmetic rather than forecast, and it is fully determined by data already recorded. Tuesday's supply piece made the same argument in the opposite direction — that roll-off alone would lift the three-month absorption ratio. The 20-session window is now set up to do the reverse.

Take the hypothetical of exactly zero net flow from here — no creations, no redemptions — and hold the model's reference deviation fixed:

The non-linearity matters more than either number. The first five sessions of roll-off are nearly harmless; sessions six through ten are not, because August 20-26 alone carried 23,445 BTC of absorption and $1,797.9M of net flow. A fortnight of genuinely flat flows would not gently erode this composite — it would take it through zero, on data that already exists.

This is a sensitivity, not a prediction. Flows will not be exactly zero, and a single strong session would reset much of it. But it establishes which way the mechanical pressure runs, and it is the reason a 28.9 built this way carries more downside than upside from its own construction.

The treasury component caught up, partly for the wrong reason

Last Friday this column flagged that the treasury component was running on Strategy's August 24 filing and understated corporate demand by roughly three points. That gap has closed: the composite now reads the September 8 8-K, holdings of 845,050 BTC, and the component moved from -1.8 to +2.5.

The reason is worth naming, because it is not new buying. The component sums bitcoin acquired across the four most recent filings. Last week that window held Strategy's two August sales, -1,638 BTC and -1,690 BTC. Those have now aged out, and the August 31 purchase of 4,603 BTC aged in — while the most recent filing itself discloses no bitcoin bought and none sold, with $176.3M going to a preferred-stock buyback instead. Thursday's treasury piece owns that subject in full; the point here is only that a +4-point improvement in an institutional bitcoin demand component came from two sales leaving a four-filing window, not from fresh accumulation. Primary filing history is on the MSTR dashboard.

What this composite cannot see

Four limits bound what this number can honestly claim.

It sees US spot ETFs and one company's treasury. Direct institutional purchases, offshore vehicles, futures basis positioning and every other corporate treasury are outside the dataset. Broad claims about bitcoin institutional adoption are not what this score measures.

Price confirmation runs on a different clock. Its window counts price observations, which include weekends, so it reaches back about three calendar weeks — to an implied reference near $76,300 — while the flow components reach back four trading weeks to August 13. The +0.30% reading means bitcoin is flat against late August, not that it went nowhere over the same span the flow components cover. Two components labelled "20-window" are not measuring the same period, and it carries only 6% weight partly for that reason.

Absorption is an accounting identity. The 4.70x figure converts net ETF dollar flows to BTC at each day's close and compares that to the 450 BTC mined daily. It says nothing about coins moving between existing holders.

A complete composite is not an accurate one. All eight components computed this run. That is necessary for a trustworthy reading, not sufficient — as last week's stale treasury component demonstrated while the missing-components list sat empty.

What would move the institutional bitcoin demand score next week

The composite is not a forecast and there is no mechanical link between it and future price. But its structure makes the sensitivities explicit:

Watch breadth and persistence before the headline number — they moved first this week and they carry the fastest response. Track the live composite on the institutional signal dashboard, the flow picture on the overview and ETF flow pages.

The institutional support score is a directional proxy for institutional bid strength. It is not a price forecast, there is no mechanical link between the score and future returns, and nothing here is investment advice. All composite, flow and absorption figures are from bryptoflows' own API, computed September 11, 2026 and current through the September 10 close; last Friday's comparison figures are as published in that week's signal piece.