Bitcoin outlook, 2026-09-15

The argument

There is a case for self-custody that does not depend on anyone actually practising it, and it is more interesting than the usual one. Treat the ability to withdraw as a deterrent.

A state with a nuclear arsenal does not need to use it to change how others behave. The weapon works before it is fired, because rivals adjust their calculations knowing it exists and could be used. The claim is that bitcoin custody has some of the same structure, and that the button is the withdrawal.

As long as clients can actually move their coins to a wallet whose keys they hold, the custodian is not negotiating from strength. It knows it is replaceable. And the sharper half of the point is that clients do not even need to find a replacement: they can leave for no custodian at all. That option does not exist for a bank deposit, a brokerage account or a gold vault, and it is the thing that makes the threat unusual.

If that is right, then custodial discipline does not rest mainly on the virtue of custodians. It rests on what they stand to lose if their clients can no longer leave.

What it costs a custodian to close the door

The obvious objection is that an exchange can simply refuse. That is true, it has happened, and it will happen again.

But look at what the decision costs. Suspending withdrawals is a single-shot weapon, and the thing it is most likely to destroy is the business firing it. It is immediately visible: on a public ledger with public complaints, the whole market knows within hours. That visibility does not prove the reserves are gone, because there are real outages, real incidents, real legal constraints. It does mean the clock starts, and the longer it runs the more the explanation has to carry.

Meanwhile the competitor that suspended nothing gains depositors, and therefore revenue, for no effort at all. It only had to not do the thing.

The record supports the mechanism. Mt. Gox halted bitcoin withdrawals on February 7, 2014, suspended trading on February 24 and filed for bankruptcy four days later. Celsius suspended withdrawals in June 2022 and never resumed them. FTX halted withdrawals on November 8, 2022, with bitcoin falling from about $20,500 to about $16,900 inside a day, and filed for bankruptcy that week. In each case the door closed once and the institution did not survive it.

The evidence that the deterrent works, and the evidence that it is blunt

What happened after FTX is the closest thing to a natural experiment.

Two responses followed, both of the kind the argument predicts. Coin balances moved off exchanges into self-custody, a shift that has never fully reversed. And proof of reserves went from a niche demand to an industry expectation within weeks, after Binance's founder publicly called on exchanges to publish wallet evidence and the major platforms shipped dashboards and Merkle-tree tools. Nobody legislated that. Competitors did it because not doing it had become expensive.

So the mechanism is real. It is also blunter than the analogy suggests, in two ways that matter.

First, what the pressure actually bought was thin. Four years on, exchanges still largely prove assets without proving liabilities, which is a different and much weaker claim: an attestation that coins exist somewhere is not an attestation that they exceed what is owed. The market extracted a gesture in the shape of the thing it wanted. Deterrence produced a response, not necessarily the right one.

Second, the flight to self-custody is not a one-way ratchet. Exchange balances have been climbing again through 2026, up roughly 45,000 BTC since May, and Binance alone held about 693,000 BTC in early September, a two-year high and around 30 percent of all bitcoin held on exchanges. That puts the exchange float near 2.3 million BTC and concentrates a large share of it in one place. Concentration does not by itself break the argument, but it narrows the number of doors that have to stay open for the threat to mean anything.

Where the analogy strains

Deterrence theory needs a threat that is credible and exercisable. Two cases test that here.

The first is a simultaneous block imposed from above, where the well-behaved competitor is not allowed to be well behaved. Within one jurisdiction that removes the competitive punishment entirely. What remains is geographic competition, the same game played between jurisdictions rather than between firms, which is slower and much less reliable.

The second is subtler and gets less attention. A right that is never exercised can quietly stop working. Deterrence assumes the arsenal is operational and known to be operational. A withdrawal path that almost nobody uses is a path that can degrade, through fees, friction, delays, compliance gates or plain neglect, without anyone noticing until it is tested. The analogy's own logic requires that the capability be maintained, not merely declared.

ETFs are the umbrella, not the button

A spot ETF holder is in a different position and it is worth being precise about it, because this is where the argument is most often stated loosely.

Someone holding shares of a spot bitcoin ETF cannot withdraw bitcoin. Their only exit is selling the shares for cash. Creation and redemption happen upstream, in large baskets, between the fund and authorized participants. Since July 29, 2025, when the SEC permitted in-kind creations and redemptions for crypto exchange-traded products, those participants can take delivery in bitcoin itself; before that date the same trades were cash-only.

The pressure still transmits, though. If selling pushes the share price below net asset value, authorized participants can buy shares, assemble baskets and present them for redemption to capture the spread. Bitcoin leaves the fund either way: in-kind it goes to the participant, in cash the fund sells it. Assets under management fall, and so do the manager's fees.

So the discipline exists, but it is held by someone else. Holding an ETF is not holding the button. It is standing under an umbrella that other institutions are carrying, and it works only as long as their incentive to arbitrage the discount is intact.

Treasury companies have less of a door than ETFs

This is the case Signal21 covers most closely, and it is further along the same spectrum. We track eleven listed companies whose business is holding bitcoin, and Strategy alone reported 845,050 BTC as of mid-September, a little over 4 percent of circulating supply.

A shareholder in one of those companies has no redemption mechanism at all. There is no basket, no authorized participant, no path by which selling the share returns bitcoin to anyone. The only exit is selling to another buyer at whatever the market pays, which is why these vehicles can and do trade at large premiums and discounts to the bitcoin they hold. Whatever discipline applies to their management comes from equity markets and disclosure, not from the threat of withdrawal.

That makes a treasury company, on this specific measure, a more closed structure than an ETF. It is not a judgment on the companies. It is a description of where the button is.

How much bitcoin now sits where the door was never built

This is the part of the argument that can be counted rather than asserted, and it is the part that should worry anyone who finds the deterrence framing persuasive.

US spot bitcoin ETFs held about 1.66 million BTC in early September, roughly 8 percent of circulating supply. Strategy holds another 845,050. Those two categories alone come to about 2.5 million BTC, over 12 percent of the roughly 20.08 million in circulation, in structures where no end holder can withdraw a coin. Add the rest of the listed treasury vehicles and the share rises further.

Set that against the roughly 2.3 million BTC on exchanges, where the door does exist and can be walked through.

The risk the argument points at is therefore not mainly that some exchange closes its doors one morning. It is that a growing share of bitcoin ends up in buildings that were never given a door, so the arsenal shrinks relative to the stock while the deterrent is still being described as though it were intact. On that reading, the holders who keep their own keys are not only protecting their own coins. They are maintaining the credibility that the other arrangements quietly depend on.

Choosing not to walk through a door and not having one are different situations, even when the picture looks identical from inside the room.

What we are not asserting

Two limits on the above. We are presenting this as an argument to be weighed, not as a finding: the causal claim that the withdrawal right is what disciplines custodians is consistent with the post-FTX record but not proved by it, because competitive pressure, litigation, insurance and regulation all moved at the same time. And we make no recommendation about how any reader should hold bitcoin. Custody involves operational risks that run in the other direction, and they are real.

What it means for our view

Today was a hard session, and for a reason that is not about custody. The Senate's cloture vote on the CLARITY Act failed on Tuesday afternoon, short of the 60 votes needed, with more than 40 senators against, which ends market-structure legislation in this Congress for the year. Bitcoin had run to $79,530 overnight and fell through the afternoon to a low of $75,595, trading near $76,100 as we publish, down about 3.6 percent on the day.

That low deserves a note, because it is our own line. We have been writing for three weeks that the bears' burden is a daily close below $75,600, and today the price traded a few dollars through it and came back. The level was touched, not broken: there is no daily close below it, so our short-term view stays neutral and the $70,000 region stays a conditional target rather than an active one. If tonight's close lands underneath, that changes tomorrow, and we will say so.

Our three horizon views are unchanged and refreshed today with the same views on all three. The Federal Reserve announces tomorrow, September 16, with futures markets near 90 percent odds of a rate increase after August inflation at 3.4 percent. The custody question in this piece is slower than any of that, and it will still be there when the week's headlines are gone.

This is general market commentary, not investment advice or a recommendation to buy or sell any asset.