I have the spreadsheet open in front of me right now. Three tabs. The first is daily issuance — block reward times 144 blocks per day, with the April 2028 halving stepped in as a single cell change. The second is net ETF inflows pulled from the daily flow reports the issuers publish. The third is the subtraction. That third tab is the only one anyone in this conversation actually argues about, and almost nobody who argues about it has built it.
I built it because I got tired of reading the same thread.
The thread goes like this. The 2028 halving cuts daily new BTC supply roughly in half. ETFs are absorbing more than the daily supply already. Therefore the supply squeeze is structural, the price floor is mechanical, and the only question is how far up. I have read variations of that argument from people who I respect. I have also read it from people who definitely have not opened a spreadsheet. The numbers used are almost always wrong, and the framing is almost always missing the part that decides whether the thesis survives contact with a real liquidation cascade.
So I want to walk through what the model actually says. Not what the headline says. What the cells say once you make them honest.
The Front-Run Thesis Is Real. The Mechanism Most People Describe Is Not.
Let me concede the strongest version of the argument up front, because it is the part the skeptics get wrong.
ETF demand absorbing more than daily issuance is a real thing. It happened in 2024. It happened again in 2025 during specific windows. On those days, every freshly mined coin had a buyer before it left the miner's wallet, and the marginal buyer for actual price-discovery flow was a sell-side desk handing inventory to BlackRock. That is structural and it is new. The pre-ETF cycle did not have this dynamic. Anyone telling you the ETF flows do not change the market microstructure has not been reading the desk reports.
Concede. That part is real.
But then the thread skips a step. It treats "ETFs absorb more than daily issuance" as if it were the same statement as "circulating supply is shrinking on a net basis." It is not. Daily issuance is roughly 450 BTC right now and falls to 225 after April 2028. The total addressable supply that can hit a bid on any given day is not 450 BTC. It is whatever fraction of the 19.8 million existing coins decides to move that day, plus the 450, minus whatever gets locked into long-term storage. The issuance number is the smallest input in the model.
I keep seeing the spreadsheet built with issuance as the denominator. That is the wrong denominator. The right denominator is liquid float — coins that have moved in the last 90 days, sitting on exchange order books or in active wallets, ready to be sold into a bid. By the on-chain estimates I trust, liquid float is somewhere around 3 million BTC, give or take half a million depending on whose heuristic you use. Daily issuance is 450. The ETFs are absorbing, in good months, between 2,000 and 8,000 BTC per day net.
So when the ETF absorbs 5,000 BTC in a day, it is consuming roughly 0.17% of liquid float. Per day. That is meaningful. It is also nowhere near the dramatic "issuance gets eaten in three hours" framing that the threads use. The squeeze is real but the squeeze is gradual, and gradual squeezes get interrupted by sell-side events more often than they get rewarded with vertical price action.
The April 2028 halving cuts the new-issuance input from 450 to 225. That changes the daily net-absorption math by 225 BTC. At a $100,000 BTC price, that is $22.5 million of daily flow shifted. The aggregate ETF complex traded $3.5 billion of net flow on its biggest weeks of 2025. The halving impulse is, in flow terms, less than 1% of what the ETFs themselves do on a busy week. The structural shift is being priced in over years, not unlocked at a single block height.
That is the first thing the model tells you that the threads will not. The halving matters. The halving is also smaller, in flow-units, than three good ETF days in a row.
The Numerator Lies More Than the Denominator Does
Now the harder problem. Net ETF inflows are the line everyone watches. Net ETF inflows are also, in the way most people use the number, misleading during exactly the conditions where the thesis is supposed to pay off.
Here is what I mean. The published net flow is creations minus redemptions for the spot ETF complex. When a bull thesis is consensus, creations dominate, the number is positive, and the chart looks like a structural bid. When sentiment cracks, redemptions spike, and the number turns negative for a stretch — sometimes for weeks. The framing that breaks is the assumption that the absorption rate is roughly constant and the halving will tip a static system into structural deficit. It will not. The absorption rate is the most volatile input in the entire model. ETF flows in March 2024 looked nothing like ETF flows in August 2024. The standard deviation of weekly net flow is larger than the mean of weekly net flow. That alone should disqualify any clean "issuance minus inflows" subtraction as a forecast tool.
The model I built does not forecast. It backtests. I ran it against the 2024 inflow data and asked: in a counterfactual where the halving had already happened a year early, would the supply-demand picture have looked structurally different? The answer was yes during inflow weeks and no during outflow weeks. The halving does not change the sign of the absorption equation. It changes the magnitude during periods when the sign is already positive. That is a less exciting answer and it is the answer the data gives.
There is another dishonesty in the standard framing — and I will name it because nobody else in this conversation does. The ETF flow number is directional, not custody-final. A creation today can be redeemed in six months. The coins do not sit in BlackRock's cold storage as a permanent sink — they sit there as long as the AP arbitrage and end-investor allocation say they should. The squeeze framing assumes ETF custody is functionally equivalent to coins-burned-forever. It is not. The Coinbase Custody and Fidelity Digital Assets cold-storage holdings are a queue, not a vault. Coins flow in. Coins flow out. The net is what it is, but the cumulative is not a one-way ratchet.
If you want to model this honestly, you need to weight inflows by the probability they get held through a 30% drawdown. That probability is not 100%. Based on the 2024 redemption pattern, it is closer to 60-70% during a normal correction and considerably lower during a panic. The model becomes much less dramatic when you discount inflows by their retention probability. The thesis survives. The vertical chart that the thesis is selling does not.
What the Model Actually Says About 2028
The honest output of the spreadsheet — once issuance is correctly compared to liquid float rather than absorbed by net flow, and once net flow is discounted by retention probability — is that 2028 produces a measurable supply-demand asymmetry under one specific condition. The condition is that net ETF flow remains positive on a six-month rolling basis through the halving date.
Under that condition, the math gets interesting. Daily issuance halves. Even moderate flow continues to consume more than the new supply. The marginal price-discovery flow tightens. The liquid float contracts on the margin. Not dramatically. Measurably. Over months, not hours.
Under the other condition — where flows turn neutral or negative in the 12 months leading into the halving — the halving is a non-event in microstructure terms. The new supply is already a tiny fraction of float, and a 50% cut of a tiny fraction is a smaller tiny fraction. The price reacts to the narrative, briefly, and then resumes whatever the macro tape is doing. This is what happened in 2016 and what mostly happened in 2020, the latter rescued by an unrelated monetary impulse.
The thesis is conditional. The threads sell it as structural. Those are different claims.
For the reader trying to update their view, I will not give you a prediction. I will give you signals. Watch four things. One: the six-month rolling net flow into the spot ETF complex — track it weekly, and if it crosses into sustained negative territory before April 2028, the squeeze thesis weakens materially. Two: the ratio of liquid float to total supply as estimated by the on-chain analytics providers — if liquid float drops below 14% of supply heading into the halving, the absorption math gets meaningfully tighter. Three: the retention behavior of inflows during the next correction of 25% or more — if the retention rate stays above 70% in the drawdown, the institutional bid is sticky and the squeeze model strengthens; if it falls below 50%, the bid is tactical, not structural. Four: the qualified custodian holdings disclosed by Coinbase Custody, Fidelity Digital Assets, and Anchorage Digital — these are the cold storage queues that matter, and the rate of change of the aggregate is the cleanest single signal you can watch.
Those are the four cells in the spreadsheet I check every Monday morning.
This article started as a model build because I wanted to refute a specific tweet I had been carrying around for a week. It turned into something more annoying — a piece that says the thesis is right and the mechanism is wrong, which is the most thankless position to defend in a market that wants binary answers. The spreadsheet does not give a binary answer. The spreadsheet says: conditional. The condition is the ETF flow regime through 2027. If you cannot tell me what that regime looks like, you cannot tell me what the halving does. And nobody, including me, can tell you what that regime looks like with any confidence.
FAQ
Does the 2028 halving guarantee a supply squeeze given current ETF demand?
No. The halving cuts daily new issuance from 450 BTC to 225 BTC, but daily issuance is the smallest input in the supply-demand equation — liquid float and net ETF flow are both an order of magnitude larger. The squeeze materializes only if six-month rolling ETF net flow stays positive through the halving date. If flows turn neutral or negative in the run-up, the halving becomes a microstructure non-event regardless of the narrative around it.
Why is net ETF inflow the wrong number to watch in isolation?
Because net inflow is directional, not custody-final. A creation today can be redeemed in six months, and the 2024 data shows redemption rates spike during corrections — meaning ETF custody is a queue, not a permanent sink. Modeling honestly requires discounting inflows by the probability they get held through a 25-30% drawdown. Historical retention sits around 60-70% in normal corrections and lower in panics, which softens the structural-bid framing considerably.
What is liquid float and why does it matter more than issuance?
Liquid float is the subset of total BTC supply that has moved in the last 90 days and sits on exchange order books or active wallets, ready to hit a bid. Current estimates put it around 3 million BTC. Daily issuance of 450 BTC is roughly 0.015% of liquid float — almost nothing. The right denominator for any squeeze model is liquid float, not issuance. Most threads use issuance because the math looks more dramatic that way.
Which custodians actually hold the ETF cold storage and why does it matter?
The bulk sits with Coinbase Custody (NY DFS Trust Company) and Fidelity Digital Assets (NY DFS Trust), with Anchorage Digital (OCC Federal Trust Charter) holding a smaller institutional slice. It matters because the rate of change of their aggregate holdings is the cleanest signal of whether ETF coins are accumulating or rotating out. Watching headline net flow alone misses the qualified-custodian queue dynamics that determine whether absorbed supply is structurally locked or just temporarily parked.
How sensitive is the model to assumptions about retention probability?
Very. Shifting the retention assumption from 100% (the implicit assumption in most squeeze threads) to 65% (the empirical 2024 figure during corrections) cuts the effective absorption rate by roughly a third. That single adjustment changes the model output from "structural deficit by Q3 2028" to "conditional deficit only during sustained inflow regimes." Retention probability is the variable that quietly does most of the work in any honest version of this model.
Does self-custody behavior affect the squeeze thesis?
Indirectly, yes. Coins held in cold storage on hardware wallets — Ledger, Trezor, GridPlus Lattice1 — are functionally removed from liquid float and behave more like a permanent sink than ETF custody does. The growth rate of long-term-holder supply, which the on-chain analytics providers track, is a more reliable indicator of structural supply removal than ETF flows. The self-custody cohort does not redeem during corrections the way ETF allocators do, which is why their behavior is the closest thing the market has to a true issuance offset.
What signals would falsify the front-run thesis before April 2028?
Three would do it. Sustained negative six-month rolling ETF net flow at any point in the 12 months before the halving. A drop in qualified-custodian aggregate holdings of more than 15% from the prevailing peak. Retention rate below 50% during the next 25%+ drawdown. Any one of those would weaken the structural framing; two together would invalidate it. Watch them weekly. Update your view on the data, not on the narrative.
Why not just buy BTC and stop modeling?
Fair question. The model is not an investment thesis — it is a framework for not being wrong about the mechanism if you do buy. Plenty of people will be right about the direction in 2028 and wrong about why. Being right about the why matters when the price action diverges from the thesis for six months and you need to decide whether the original argument is broken or just early. The spreadsheet tells you which one it is. The tweet does not.