Supply and Scarcity · Published

Bitcoin Supply Shock: Three-Month Ratio Hits 0.77x

How much bitcoin did ETFs absorb against new supply this quarter?

Over the sixty reporting sessions from June 11 through September 4, US spot bitcoin ETFs absorbed 20,729 BTC against the 27,000 BTC the network minted across the same sessions — a ratio of 0.77x. That is the first positive three-month reading since May 28, and it is the single most consequential number in this dataset right now, because a week ago the same window read -231 BTC. The bitcoin supply shock that has dominated the one-month data since mid-August has finally worked its way into the quarter. It has not yet reached parity: over three months the ETF complex still consumed roughly three-quarters of new supply, not more than the miners created.

The arithmetic is unchanged and worth restating for anyone arriving cold. Each session's net ETF flow is converted to bitcoin at that day's close, then divided by the 450 BTC the network mints daily. Above 1.0 the funds took more coins off the market than miners produced; below zero, redemptions handed coins back. bryptoflows builds the series from Farside Investors flow data and Hyperliquid closes — the rolling chart lives on the institutional signal dashboard, the underlying flows on the ETF flow dashboard. Every figure below is computed from that dataset, current through the September 4 session, with the US holiday on September 7 leaving Friday as the most recent reporting day.

The three-month window cleared zero for the first time since May

The crossover happened on September 1 and then accelerated. Reading the trailing sixty sessions in coins, day by day:

Last week's article named this reading as the thing to watch — "the number to watch is not whether absorption stays positive; it is whether the sixty-session figure can climb into positive territory as June's redemptions age out of it." It did, in four sessions, and the mechanism was exactly the one described: new demand arriving at the front of the window while June's redemptions rolled off the back.

The shorter windows are stronger and have been for three weeks. As of September 4 the trailing five sessions ran 5.42x, ten sessions 5.30x, twenty sessions 5.00x, and forty sessions 3.26x. Ninety sessions remains deeply negative at -1.11x, carrying the spring redemptions that have not yet aged out. The curve across horizons still inverts — intense at the short end, negative at the long end — but the inflection point has moved out from about two months to about three, and that is what a genuine change in bitcoin supply absorption looks like when it is real rather than a single good week.

The 20-session reading itself sits at 5.00x, above 96.5% of the 170 rolling readings recorded in 2026 and above 78.4% of all 662 readings since January 2024. It has been essentially flat for a week — 5.15x on August 31, dipping to 4.17x on September 2, back to 5.00x on Friday. The one-month story is no longer accelerating. The three-month story is.

Why the bitcoin supply shock strengthens without a single new dollar

This is the part worth stating in advance, because it is fully determined by data already recorded and will look like news when it prints.

Both trailing windows currently have negative sessions sitting at their back end, waiting to roll off. If net flow from here were exactly zero — no creations, no redemptions, nothing — the ratios would still rise:

That is a real and underappreciated property of a trailing ratio: for the next two weeks, the three-month bitcoin supply absorption figure improves whether or not anyone buys anything. Reaching 1.0x on the sixty-session window from today's flow requires only 6,271 BTC, about $499M of net inflow at Friday's $79,623 close — roughly half of last week's total. Between the roll-off and any ordinary week of demand, a three-month reading above 1:1 is close to the base case.

The honest flip side is that this cuts both ways later. A trailing window that rises mechanically today will fall mechanically in October, when September's strong sessions age out and have to be replaced. Anyone quoting a 1.7x three-month figure in two weeks should be asked how much of it was flow and how much was the calendar. The answer, on today's data, is that most of it is the calendar.

One session took twenty days of issuance off the market

The September 3 session did most of the work in the crossover and deserves its own accounting in coins rather than dollars. The category's $730.8M net inflow that day converted to 8,995 BTC at the $81,244 close — 19.99x that day's 450 BTC of issuance, and the largest single-session coin absorption of 2026, ahead of January 14 (8,673 BTC) and April 17 (8,620 BTC). Across the full 681-session history it ranks 27th.

That one day is 43% of the entire sixty-session net absorption. It is also the reason to be careful: a three-month ratio that rests this heavily on one session is not the same thing as three months of steady demand. Monday's flow analysis makes the complementary point from the dollar side — that the ex-BlackRock cohort has only just stopped subtracting from the category, at +$47.1M over the same sixty sessions. Deep demand through a narrow channel absorbs supply exactly as effectively as broad demand does, and it reverses considerably faster.

Bitcoin daily issuance is 450 BTC until April 2028

The denominator is the only fixed part of this calculation. Bitcoin daily issuance is 450 BTC — a 3.125 BTC block subsidy across roughly 144 blocks a day — and it does not respond to price, to hashrate migrating toward AI data-center contracts, or to difficulty retargets, which exist precisely to hold issuance at target regardless of how much computing power is attached to the network. At the next halving, estimated for around April 2028 at block 1,050,000, the subsidy drops to 1.5625 BTC and daily issuance to roughly 225.

That step matters for the arithmetic in a specific way: it doubles every ratio on this page for identical demand. Today's 20,729 BTC over sixty sessions reads 0.77x against 27,000 mined; the same coins against a post-halving 13,500 would read 1.54x. Nothing about demand would have changed. It is worth internalising now, because a great deal of 2028 commentary will present that doubling as a demand story when it is a denominator story.

Cumulative absorption is now four years of issuance

The rolling windows swing hard. The cumulative figure does not, which makes it the more durable expression of bitcoin scarcity from this channel.

Since January 2024, net ETF creations converted at each day's close total 663,017 BTC. At 450 a day that is 1,473 days — just over four years of production held by a product category thirty-two months old, and about 3.16% of the 21 million coins that will ever exist. Across the full 681-session record, absorbed against issued, the life-to-date ratio is 1.96x: the ETF complex has taken just under two coins off the market for every one the network created since it opened.

The annual breakdown shows how uneven that has been:

2026 remains a net negative year for ETF supply absorption even after the last month. September's four sessions have added 9,441 BTC at 5.24x, and it would take roughly 22,400 more to bring the year back to flat. That is the context that keeps the current reading in proportion: the channel spent two and a half years absorbing supply, spent this year returning it, and has spent one month reversing part of that.

What the absorption ratio does not count

The metric is narrow by construction, and three limits matter more than usual this week.

Execution price is the fourth limit and it is still working against this window. Across the sixty sessions, 34 inflow days put $7,102M to work and bought 100,781 BTC, an average of $70,472 per coin, while 26 outflow days withdrew $5,025M and released 80,052 BTC at an average of $62,776. Redemptions cleared roughly 11% cheaper than creations. Had every one of those dollars executed at Friday's close instead of the prices actually available, the window would read 26,084 BTC rather than 20,729. Bitcoin rose 25.2% across the window, and a rising price mechanically lowers this ratio — each subsequent dollar removes fewer coins. That is an argument against reading the absorption ratio as a price forecast. It is a demand proxy measured in coins but funded in dollars, and its own arithmetic works against extrapolating it into a price path.

Corporate treasuries are pulling the other way

The second channel that removes coins faster than miners create them moved in the opposite direction over the same window. On bryptoflows' Strategy dataset — current through the August 24 filing, with holdings of 840,447 BTC — the eleven weekly filings since June 11 net to -4,809 BTC: 2,107 acquired against 6,916 sold. Over the same sixty sessions in which ETFs absorbed 20,729 coins, the largest corporate holder was a net seller.

That gap is the argument for treating the two channels separately rather than summing them into one institutional bid. Thursday's treasury piece covers Strategy's more recent filing and owns that subject in detail; the point here is only that the quarter's absorption was an ETF story, not a broad one.

The clean summary is this. Three months of ETF demand has, for the first time since May, removed more bitcoin than it handed back — a net 20,729 coins, or 0.77 of every coin the network mined over the same span. It is a real crossover and it will likely strengthen through parity in the next fortnight, mostly because June's redemptions are aging out rather than because anything new arrives. A supply shock that reaches 1:1 by subtraction is worth measuring accurately and worth describing precisely, which is not the same as worth getting excited about.