- $298.5Bthe whole float, 1 November
- -$2.1Bten months to 28 October
- -$3.1Bthe four days that followed
The float is the inventory this market runs on, and the ten boring months are the ones worth reading. Four days took $3.1 billion out, and almost nobody asked which issuers absorbed the difference. The total is the least interesting number in the series.
I have a dashboard I open before I open anything else, and it does not show a price. It shows the stablecoin float. Not because the float predicts anything, but because it is the one line on the screen that answers a question price cannot: how much cash is actually sitting inside this market, ready to be spent?
On 1 November 2026 that line read $298.5 billion. It was the lowest reading since August, down $5.2 billion from the first of January, and down from a 2026 peak of $315.5 billion on 17 May, which means the year has given back $17.0 billion, or about 5.4 percent of the high.
Here is the part that made me stop. Split that $5.2 billion by when it left:
- From 1 January to 28 October, the float fell $2.1 billion. Ten months. On a base near $300 billion, that is 0.7 percent. Nothing happened.
- From 28 October to 1 November, the float fell $3.1 billion. Four days.
So the “2026 stablecoin contraction” is not a trend. It is one week wearing a trend’s clothes. And if you have been reading the float as a demand signal, that week just taught you something about what the number is and what it is not.

Take the machine apart first
A stablecoin float looks like money, but mechanically it is a stock, and the stock is maintained by three valves and a padlock. If you can name all four, the data stops being mysterious.
Valve one: mint. A customer sends a dollar to the issuer, the issuer sends back a token and buys a Treasury bill with the dollar. The float goes up. This valve is not open to the public. You need an account, and an account requires a compliance review, which means mint velocity is a function of who has been onboarded, not of who wants in.
Valve two: redeem. A token holder sends the token back, the issuer burns it and wires the dollar. The float goes down. Same gatekeeping, and a different latency: a mint can be instantaneous because the issuer already has the counterparty’s money, while a redemption has to survive banking hours, a bank that may not want the wire, and whatever the issuer’s terms say about large tickets.
Valve three: nothing. Most of the float does not move. It sits on exchanges as collateral, in lending pools as the base asset, in market makers’ inventories. It is not being spent; it is being held so that it can be spent. That is what inventory means, and it is why a small change in the float can mean a large change in behaviour.
The padlock: freeze. Both major issuers can block an address. When they do, the balance stops being transferable. Here is the part people miss: a freeze does not change the total supply. The token is not burned. It stays on the contract, held by an address that can no longer send it. So every headline float figure includes coins that nobody is allowed to move.
That last point matters more than the other three, and I will come back to it.
float = minted - burned
usable = float - frozen$298.5B
total stablecoin float, 1 November 2026The arithmetic is the story, not the total
Every write-up I read this week quoted the same shape: stablecoin supply is shrinking, risk appetite is fading, here is the price chart to prove it. The supply figure is real. The reading is lazy, because the total is a sum of three things that moved in opposite directions.
Over the same ten months to 28 October:
| line item | change | window |
|---|---|---|
| USDT supply | -$5.3B | 1 Jan to 28 Oct |
| USDC supply | -$4.0B | 1 Jan to 28 Oct |
| every other issuer, combined | +$7.2B | 1 Jan to 28 Oct |
| net, ten months | -$2.1B | 1 Jan to 28 Oct |
| net, final week | -$3.1B | 28 Oct to 1 Nov |
| net, year to date | -$5.2B | 1 Jan to 1 Nov |
Check the column. The two largest issuers lost $9.3 billion between them. Every other issuer on earth gained $7.2 billion. The net is minus $2.1 billion, which on this base is noise. The composition is a migration.
That is not a liquidity signal. That is a custody signal that happens to be denominated in dollars.
To see why, go back to the valves. A token from either of the two largest issuers redeems through a compliance queue thousands of institutions have already cleared, against a reserve book that is published and attested, and in one case about to be supervised by a banking authority. A token from a three-month-old issuer with $400 million outstanding redeems through whatever its terms say this month, against reserves you cannot independently see. Both count as one dollar in the total. They are not the same dollar.
The float is a count of claims. It is not a measure of how good the claims are. And a market that is quietly exchanging good claims for worse ones will show up as a flat line.

Then the week arrived
Between 28 October and 1 November, $3.1 billion left the float. You already know what else happened in that window, so I will not pretend the two are unrelated, and I will not pretend I can prove which caused what.
What I can do is count. In the Binance incident, roughly $2.7 billion of the stolen assets were frozen or otherwise immobilised, mostly in the two dollar tokens whose issuers can block an address. Tether and Circle acted within about 40 minutes of the Ethereum leg. Circle alone reports blocking more than $800 million.
Now apply the padlock rule. Those tokens were not burned. They are still in the supply, sitting in addresses that cannot spend them. So the figures everyone is quoting for “how much stablecoin exists” include a few billion dollars of stablecoin that no longer functions as money.
If you want a usable number instead of a reported one, you have to subtract. And when you subtract, the picture sharpens: the float did not just shrink because holders got nervous. Part of it stopped being float.
A frozen balance is still a liability on the issuer’s book and still a line in the supply data. The only thing that changed is that it stopped being spendable, and the supply series has no column for that.
There is a second-order effect I keep thinking about, which is what freezing does to the people who were not frozen. If you hold a dollar token because you believe you can move it in thirty seconds, and you watch a few billion dollars of the same token stop moving in forty minutes, your belief about your own balance adjusts even though your balance did not. That is a run that never shows up in the outflow data, because nobody sold anything. They just moved to a place they trusted more, or converted to an asset with no issuer at all. A float that shrinks by $3.1 billion in a week when bitcoin is already down 7 percent on the year is consistent with all of these at once.

What the float cannot tell you
I like this metric and I want to be honest about its limits, because the temptation with a clean series is to over-read it.
- It cannot separate selling from fleeing. A dollar leaving the float because someone bought bitcoin and a dollar leaving because someone withdrew to a bank look identical in the total. The first is bullish for risk assets. The second is not. The series has one number and two stories.
- It cannot see the tail’s reserves. The $7.2 billion that moved to smaller issuers is counted at face value. What backs it is not in the data, and for most of those issuers it is not public in any form a skeptic would accept. You are trusting a mint.
- It cannot see frozen balances. As above. Reported float is an upper bound on spendable float, and in a week like this one the gap is measured in billions.
- It cannot see velocity. $298.5 billion sitting still and $298.5 billion changing hands ten times a day are the same number and completely different markets. Supply is a stock. What you actually want is turnover, and nobody publishes it cleanly.
The honest summary is that the float tells you the size of the inventory, and the size of the inventory tells you how much force a given flow can exert. When inventory is $300 billion, a $3 billion move is noise. When inventory is $30 billion, the same $3 billion is a crisis. That ratio, not the level, is the thing worth watching.
What I actually watch now
I did not change my mind about stablecoins this week. I changed one habit, which is that I stopped reading the total as a single line.
- Watch the three lines separately. Total, largest two, everyone else. When the total is flat and the split is moving, the market is not neutral; it is reallocating trust, and trust reallocations are cheaper to see before they become price.
- Subtract the frozen. Any week with a large freeze event, the reported float is overstated by roughly the frozen amount. Adjust before you conclude that demand fell.
- Check the week against the year. A twelve-month change that is dominated by four days is not a trend. Compute the same delta excluding the largest week and see whether the story survives. This week it does not.
- Ask who can redeem. For any issuer whose share is growing, the only question that matters in a stress event is how fast a large holder gets a dollar back. If the answer is “per the terms”, the share is not growing on quality.
The float fell 1.7 percent this year and the pool changed shape underneath it. Price told you a story about demand. Inventory told you a story about custody. Only one of those is new.
(The End)


