Coinbase and The Block say 73% of stablecoin volume moves outside banking hours. We checked the clock, then followed the money. The numbers tell a more complicated story.
Stablecoins have a compelling new number to advertise.
According to research published by Coinbase and The Block on October 20, roughly 73% of screened stablecoin transfer volume occurs outside US banking hours. For an industry that has spent years arguing that money should move around the clock, the finding appears to validate that argument.
The networks never close. People keep moving dollars. Traditional banking schedules increasingly lag behind how people use money.
Except there is something strange about that 73%.
A week contains 168 hours. If conventional banks operate from 8 a.m. to 5 p.m., Monday through Friday, they are open for just 45 of those hours.
That leaves 123 hours outside the banking window.
Or 73.2% of the week.
The stablecoin figure? 73%.
The similarity does not prove that the report is wrong. Its actual banking-hour definition may differ from our assumption, and its underlying transaction data were not publicly reconstructed for this analysis.
But it does expose a question that the headline alone cannot answer.
If roughly three-quarters of the week already falls outside banking hours, how much does observing three-quarters of transfer volume during that period really tell us about demand?
And there is a second problem.
The research describes enormous stablecoin transfer activity, including more than $400 million an hour in adjusted volume during the third quarter. Yet a transfer is not necessarily a payment. Moving $10 million between two treasury wallets is fundamentally different from paying $10 million to suppliers, even though both appear on a blockchain as value-changing addresses.
We examined the report’s underlying measurements, compared them with payment-classification data published by Allium, and tested the banking-hour claim against several operating schedules.
The result is not evidence that stablecoin adoption is imaginary.
It is evidence that some of the industry’s most impressive numbers still measure something other than what the payment narrative implies.
The Calendar Makes the 73% Claim Less Extraordinary
Start with the statistic that gives the report much of its appeal.
In The Always-On Dollar, commissioned by Coinbase and produced by The Block Pro Research, the argument is straightforward: global commerce does not stop when American banks close, so money should not have to wait for them.
There is a genuine economic case behind that argument. A company receiving stablecoins on Saturday does not need to wait until Monday for the blockchain to process its transfer. Someone sending funds across time zones can initiate a transaction regardless of local banking schedules.
The report’s 73% figure appears to demonstrate that people are already taking advantage of this capability.
But measuring activity outside banking hours requires knowing how much time belongs to that category.
Our first test used a simple 8 a.m.–5 p.m. weekday schedule. Under that assumption, banking hours represent 26.8% of the week, while non-banking hours represent 73.2%.
Now change the assumption.
If banks operate from 9 a.m. to 5 p.m., their weekly operating window shrinks to 40 hours. The proportion of time outside banking hours rises to 76.2%.
Extend operations to ten hours per weekday, and the proportion falls to 70.2%.
These are not alternative measurements of stablecoin activity. They are calculations of the calendar.
And that is precisely why they matter.
The report’s 73% volume figure could indicate unusually strong demand outside banking hours, activity broadly proportional to available time, or even lower average hourly activity outside the banking window, depending on the definition used.
To establish which interpretation is correct, we would need the exact operating calendar and a comparison of average transfer volume per banking hour versus average volume per non-banking hour.
The published percentage alone cannot settle the question.
There is another limitation buried in the methodology.
The full 24-page report identifies the source of its hourly analysis as Dune Analytics. The 73% calculation covers the 12 months ending July 31, 2026, and excludes transfers below $1, self-transfers, minting, and burning.
Unlike the report’s adjusted-volume series, however, this dataset still includes automated transactions.
The accompanying chart even identifies a pronounced spike associated with a default global stablecoin settlement time around 00:00 UTC.
That raises an important distinction.
A transfer occurring at midnight does not necessarily represent someone choosing stablecoins because a bank was closed. It could reflect an automated treasury process, scheduled settlement, or another operational movement that would have happened at the same time regardless of banking availability.
Such activity can still be economically meaningful.
But it is not necessarily evidence of an individual or business facing a closed-bank problem.
The 73% figure therefore establishes something useful but narrower than the surrounding narrative might suggest: stablecoin transfers continue throughout periods that fall outside conventional banking schedules.
What it does not independently establish is how much of that activity depends on banks being closed or whether transfer intensity is unusually high during those periods.
The networks are always on.
The reported distribution does not, by itself, tell us why the money is moving.
The $400 Million Hour Has Another Problem
The banking-hours finding is only one part of the story.
The same research reports $891 billion in adjusted stablecoin transfers during Q3 2026, up from $216 billion in Q1 2023.
That represents more than fourfold growth in just three and a half years.
It is substantial, even after the reported filtering of exchange-related activity, DeFi interactions, bridges, automated transactions, and stablecoin issuance or redemption.
But the hourly presentation is worth examining.
The third quarter contains 92 days, equivalent to 2,208 hours.
Spread $891 billion across 2,208 hours, and that works out to $403.5 million per hour.
That closely matches the report’s claim that stablecoins move more than $400 million an hour in organic transfer activity.
So the number is arithmetically consistent with the reported quarterly total.
It is also an average.
It does not mean $403.5 million moved during every individual hour, and it tells us nothing by itself about whether those transfers settled purchases, investment transactions, or other financial obligations.
There is a further ambiguity in how the study has been presented.
The research landing page published by The Block describes stablecoins moving nearly $3 billion an hour, while the full report highlights the smaller adjusted figure.
The difference is large enough to warrant attention.
A broader transfer measure could explain why one number is substantially higher than the other, but the publicly accessible summaries do not provide a complete reconciliation establishing that relationship.
We therefore cannot treat them as directly comparable measures, much less assume that the difference represents fake transactions.
What can be established is that the choice of measurement dramatically changes the headline.
More importantly, the adjusted figure still does not answer the question that matters for payment adoption.
How much of that $891 billion actually paid someone?
This is where the distinction between useful blockchain activity and real-world payments becomes essential.
Consider three transfers.
A merchant receives $1,000 in USDC from a customer. A crypto investor moves $1,000 between personal wallets. A company transfers $1,000 from one treasury address to another.
All three move exactly the same amount.
Only the first is clearly a commercial purchase. The second may represent storage or personal asset management, while the third could be an internal accounting operation.
An adjusted-volume filter can improve the quality of the aggregate number without necessarily determining the purpose of every remaining transfer.
The blockchain records that the dollars moved.
It does not automatically record why.
One Day of Allium Data Reveals a Different Picture
Fortunately, we do not have to rely entirely on hypothetical examples.
Allium, the data provider behind the report’s adjusted-volume series, has published a separate breakdown of stablecoin transactions by economic activity.
Its July 17, 2026, classification snapshot separates transfers into investment and trading, store of value, real-world payments, and unclassified activity.
The numbers are revealing.
Investment and trading led the way at roughly $9.25 billion, followed by $3.25 billion in store-of-value activity and $3.04 billion in real-world payments. Unclassified transfers accounted for another $1.51 billion.
Combined, the four categories add up to $17.05 billion.
We calculated the share of payment-classified transfers from those published figures.
It comes to roughly 17.8%.
That means less than one-fifth of the dollar value in this particular daily snapshot was classified as real-world payment activity.
But now look at the transaction count.
Of the roughly 5.08 million transfers across those four categories, Allium classified 4.37 million as payments.
That is about 86%.
The contrast is enormous.
Payment-classified activity accounted for 86% of all transfers, yet just 17.8% of their total dollar value.
Both numbers describe the same day.
Both are derived from Allium’s published data.
Yet someone examining transaction counts could conclude that payments overwhelmingly dominate stablecoin usage, while someone examining dollar volume would see investment-related activity absorbing most of the value.
Neither interpretation is automatically wrong.
They answer different questions.
And that is precisely the problem with presenting a large transfer-volume figure as evidence of payment adoption.
The snapshot also reveals how misleading transaction size can become.
Allium counted about 612,000 investment and trading transfers worth $9.25 billion, compared with 4.37 million payment-classified transfers totaling $3.04 billion.
Fewer investment-related transfers moved roughly three times as much money.
There are important boundaries around this finding.
It describes one day, not the entire third quarter. The figures were published by Allium rather than independently extracted from blockchain records by Ravel. The payment classifications are also based on wallet identification and behavioral analysis, not verification of every underlying invoice or merchant relationship.
Indeed, Allium’s payment taxonomy includes deposits to institutional wallets, a category that can encompass custody, settlement, and treasury activity rather than straightforward purchases.
The 17.8% figure is therefore not a definitive measurement of commercial payments, and it cannot be applied to Coinbase’s $891 billion quarterly total.
But the snapshot demonstrates why the missing breakdown matters.
Without knowing which activities dominate the adjusted quarterly figure, the industry cannot reliably translate that headline into a percentage of actual payments.
That is not a small statistical detail.
It changes what the growth number can legitimately be used to prove.
Even the Cleaned-Up Numbers Keep Changing
There is a deeper complication.
Adjusted transfer volume may sound like an objective measurement, but its accuracy depends on how effectively data providers identify the wallets and activities behind transactions.
Those classifications are not fixed.
On September 18, 2026, Visa’s Onchain Analytics dashboard documented a major methodology refresh.
The underlying identity coverage expanded from roughly 15 million to 600 million labeled addresses. With substantially more wallets identified as exchanges, contracts, automated systems, and other infrastructure, additional transactions were excluded from adjusted volume.
The result was lower adjusted dollar volume.
Not necessarily because fewer people used stablecoins.
But this was because the measurement system became better at identifying activity that should not have been included.
Visa also described an automated program on Solana that repeatedly cycled stablecoins through thousands of temporary wallets. Improved classification enabled that activity to be removed from adjusted measurements.
That example illustrates how a large transfer total can be inflated without corresponding growth in independent economic demand.
It also complicates historical comparisons.
When a dashboard updates its classification system, a change in the reported number may reflect economic activity, improved detection, or some combination of both.
This does not make adjusted metrics useless. Quite the opposite. Better filtering is necessary if stablecoin activity is to be measured seriously.
But it means that the word ‘adjusted’ should not be confused with verified payments.
There is already evidence of how large the distance between those concepts can become.
A January 2026 study by Boston Consulting Group and Allium examined more than $62 trillion in annual stablecoin transfers for 2025.
After filtering out trading activity, protocol mechanics, and other non-economic movements, the researchers identified an estimated $4.2 trillion in economically meaningful flows.
That was roughly 7% of the original total.
However, even the $4.2 trillion figure was not entirely made up of payment activity.
The researchers estimated that observable bilateral payments for goods and services ranged from $350 billion to $550 billion in 2025.
Those figures are not interchangeable with the Coinbase report’s Q3 2026 numbers. The studies cover different periods and use distinct classifications.
They nevertheless expose the scale of the measurement problem.
A trillion-dollar transfer total can shrink dramatically as researchers move from raw blockchain activity toward identifying actual economic payments.
Some legitimate payments may also be missed, especially when transactions occur within custodial platforms or through card systems without a separate visible blockchain transfer for each merchant purchase.
So even the smaller payment estimates should not be mistaken for perfect coverage.
The challenge is not merely removing noise.
It is identifying the economic purpose of the activity that remains.
The Stablecoin Payment Thesis Is Real. Its Headline Metric Is Not Enough
None of these findings establish that stablecoins are failing as payment infrastructure.
There is credible evidence pointing in the opposite direction.
In a separate February 2026 analysis, McKinsey and Artemis Analytics estimated that actual stablecoin payments were running at an annualized rate of $390 billion, based on December 2025 activity.
The researchers identified roughly $226 billion in annualized business-to-business payments and another $90 billion associated with global payroll and remittances.
Those are meaningful amounts of economic activity, even if stablecoins still represent a small fraction of the much larger global payments market.
And there are reasons that demand could grow.
Businesses operating across borders can benefit from faster settlement and reduced delays. Consumers sending remittances may face fewer intermediaries and lower transfer costs. Digital platforms can integrate programmable payment infrastructure into processes that traditional banking systems do not always serve efficiently.
The Coinbase–The Block report is right to examine those advantages.
The problem begins when several different kinds of evidence are compressed into one adoption narrative.
The report’s $891 billion quarterly figure demonstrates significant adjusted transfer activity. Its $400 million hourly figure expresses that activity as an average. Its 73% off-hours figure shows the timing of a separate, more broadly screened set of transfers.
None of those measurements directly establishes the dollar value of completed commercial payments.
And combining them does not solve the problem.
To test the payment thesis more convincingly, the next step would be to classify the Q3 2026 adjusted volume by economic purpose, disclose how much remains unidentified, and compare transfer intensity across banking and non-banking hours using consistent transaction filters.
That would allow researchers to distinguish three propositions that are currently too easy to blur.
Stablecoins can settle transactions around the clock.
Stablecoins are being used for genuine economic payments.
And stablecoins are attracting extraordinary new payment demand because they operate beyond traditional banking hours.
The first proposition is a technical reality. The second has support from independent research, although precise estimates vary. The third requires more specific evidence than a headline percentage.
Ravel’s calculations do not establish the true payment share of Coinbase’s Q3 volume. We reconstructed its hourly average, tested the 73% figure against illustrative banking schedules, and calculated transaction-count and dollar-volume shares from Allium’s published daily snapshot. We did not reproduce the underlying Dune or Allium transaction-level datasets. That limits the conclusions, but it also makes clear which claims remain unproven.
The stablecoin industry does not need every dollar transferred to represent a merchant payment. Treasury management, investment settlement, and other financial activities can create value of their own.
But those uses should be measured for what they are.
A network moving billions is not automatically a payment network processing billions in purchases. A majority of volume occurring outside banking hours does not automatically mean an unusual concentration of demand during those hours.
And a rapidly growing adjusted transfer total is not the same thing as a rapidly growing commercial payment market.
The money is moving. The infrastructure is working. Some of the payment demand is already real.
What the biggest numbers still do not tell us is how much of that money is actually paying someone.
