Institutional Signal · Published

Institutional Bitcoin Demand at 55: Deep, Not Broad

Are institutions buying bitcoin right now?

Yes, but through a narrower channel than the headline number suggests. bryptoflows' composite institutional bitcoin demand score reads +55.2 on a -100 to +100 scale as of the September 3 close, a reading the model classifies as moderate accumulation. All eight components were computable this run — nothing was dropped and nothing was renormalised away — so this is a complete composite, directly comparable to a full-strength reading. The important detail is underneath: two of the eight components supply roughly two-thirds of the entire score, and the two components built to measure how widely that demand is spread are the weakest in the set.

That is the difference between deep demand and broad demand. Right now the data says deep.

What the institutional bitcoin demand score actually measures

The composite fuses eight components, each normalised to a -100 to +100 scale and weighted, into one number. The weights this run were:

One design decision matters more than any other for reading this number honestly. When a component cannot be computed, the model omits it and renormalises the remaining weights — it never substitutes a zero. That keeps a missing dataset from being silently read as a neutral signal, but it also means a score built on five components is not the same object as a score built on eight. This run reported an empty missing-components list and full weight coverage, so no caveat is needed on that front today. It is the first thing to check on any run, and the live figures sit on the institutional signal dashboard.

The model also publishes a confidence figure, which this run put at 61.0. That is not a confidence interval around the score — it measures how much the eight components agree with each other. A composite where every component points the same way scores high; one assembled from components pulling in opposite directions scores lower, even if the weighted average lands somewhere respectable. At 61.0, the components are pointing broadly the same way, with real dispersion.

Two components are carrying two-thirds of the score

Here is the full contribution breakdown, in points of the final 55.2:

Supply absorption and ETF flow momentum together contribute 36.8 of the 55.2 points. Strip them out and the remaining six components produce a score in the high teens — technically positive, but nothing anyone would call institutional conviction.

There is also a time-window quirk worth naming. The 20-session net inflow is $3.37 billion, while the 60-session net inflow is only $1.69 billion. Both are positive, but the shorter window is larger than the longer one, which means sessions 21 through 60 were net negative by roughly $1.68 billion. The strength this composite is reading is genuinely recent. Monday's flow piece covered the 60-session line crossing above zero in detail; the composite is picking up the same recovery, and the same fragility.

The weak link: only 7 of 12 funds bought on a 96th-percentile day

September 3 was a large day by any measure. Net flows were +$730.8 million, a z-score of 2.03, landing in the 95.9th percentile of all sessions in the 680-day history. Gross inflows were $755.6 million against just $24.8 million of gross outflows.

And still only seven of twelve products finished positive.

That combination is what pulls flow breadth down to a normalised 16.7 — barely above neutral, on a 14% weight, on one of the biggest flow days in the sample. A day where money arrives in size but through five or six vehicles is a different market structure from one where the whole category bids, even when the dollar totals look identical. The concentration component says the same thing from another angle: a product HHI of 0.384 against an even-split value of 0.083 means the flow is heavily bunched, and ETF flow attribution shows why — IBIT alone accounts for 57.7% of gross flow across the full history and 115% of cumulative net flow, a figure above 100% because GBTC's -$27.7 billion drags the denominator down.

Persistence adds a third caution. The positive streak is two sessions. The model's persistence component saturates around a four-session run, so two days registers as 46.2 — a real signal, not yet a durable one.

Momentum is also cooling underneath the composite. The 5-day average net flow is $122.0 million per day, below both the 10-day average of $204.4 million and the 20-session average of $168.5 million. The composite reads the 20-session window, so a slowing five-day trend does not show up in the score yet. It would show up first here.

The treasury component is reading a filing that is out of date

This is the caveat that matters most this week, and it is not the one the model's own integrity check catches.

The missing-components list was empty — so by the model's own accounting, nothing is wrong. But the treasury accumulation component was computed from a materialized treasury view still showing Strategy's August 24 filing: holdings of 840,447 BTC, direction "none". Its four-filing window therefore sums to -3,328 BTC, which is why the component reads -13.1 and is the composite's only negative contributor.

Strategy's actual August 31 8-K, covered in Thursday's treasury piece and read directly from the primary filing, discloses a purchase of 4,603 BTC for $369.7 million, taking holdings to 845,050 BTC. The composite has not seen it.

Recomputing that one component from the primary filing, using the model's own published normalisation, moves it from -13.1 to roughly +11.4 — a swing from a -1.8 drag to a +1.6 contribution, which would lift the composite from 55.2 to approximately 58.7. The treasury capacity component would shift slightly too, since it values the position off the same stale holdings figure.

So the honest statement is this: the published score understates corporate treasury demand by roughly three points as of this writing, because the pipeline behind one 14%-weight component is running about nine days behind the primary source. That is a smaller distortion than a dropped component would be, and it is in a known direction. But it is exactly the kind of quiet staleness that a clean the missing-components list list will not flag, which is why it is worth stating rather than burying.

What the composite cannot see

Four blind spots are worth being explicit about, because they bound what this number can honestly claim.

It only sees US spot ETFs and one company's treasury. Direct institutional purchases, offshore vehicles, structured products, futures basis positioning and every other corporate treasury are outside the dataset entirely. When Wednesday's macro brief discussed rate expectations reshaping allocation decisions, none of that transmits into this score except through the flow and price channels it can actually measure.

Absorption is an accounting identity, not a claim about scarcity. The 4.93x reading converts net ETF dollar flows into BTC at each day's close and compares that to issuance. It says nothing about coins moving between existing holders, and it eased from the 5.15x that Tuesday's supply piece reported earlier in the week — the same ratio, three sessions later.

Price confirmation is partly circular. A 28.6% price gain contributes positively to a score meant to gauge demand, but price and flows influence each other in both directions. It carries only a 6% weight for exactly this reason, and it is near its saturation point, so further price strength would add almost nothing to the score.

A complete composite is not an accurate one. Today's run had all eight components, which is the necessary condition for a trustworthy reading — not a sufficient one, as the treasury staleness above demonstrates.

What would move this score next week

The composite is not a forecast, and nothing here predicts price. But the model's structure makes it clear which specific developments would move the number, and roughly by how much:

Note the asymmetry. Two of the three biggest positive contributors — supply absorption at 99.3 and price confirmation at 92.0 — are effectively pinned at their ceilings and can only fall. The components with room to rise are the ones currently weakest: breadth, persistence and treasury accumulation. A score of 55.2 built this way has more mechanical room below it than above it, which is a statement about the indicator's construction, not a prediction about bitcoin.

Track the live composite and its component breakdown on the institutional signal dashboard, the underlying flow picture on the overview and ETF flow pages, and Strategy's filing history on the MSTR dashboard.

The institutional support score is a directional proxy for institutional bid strength, built from the flow, supply, treasury and price data described above. It is not a price forecast, there is no mechanical link between the score and future returns, and nothing here is investment advice. Composite, flow and absorption figures are from bryptoflows' own API, computed September 4, 2026 and current through the September 3 close. Strategy's August 31 purchase figures are read from SEC Form 8-K accession 0001193125-26-375463 rather than from bryptoflows' treasury view, for the reason given above.