Brazil's Stablecoin Dashboards Are Lying to You
If I'm not the first, I'm one of the first to compile a BRL stablecoin dataset. And I'm guilty of this too. Eight months of data later, I found R$ 18.4 billion in settlement activity that doesn't show up in transfer data, most "users" carrying almost none of the value, and wallets that aren't really users at all.
I've been looking at Brazilian stablecoins for a while, and I've slowly realized that some of my first analyses and dashboards were incomplete. I knew there were parts of the story they weren't capturing, but I hadn't really had the time, or the data, to understand what was missing. I finally got some time to go back to the raw flows and try to unravel that story. What I found was quite a bit more interesting than I expected.
Two choices worth explaining first
Why two kinds of activity, not one
I split the addresses into two groups that behave nothing alike.
The first group is what everyone measures. One address sends tokens to another address. I call these peer transfers, and they are what a public dashboard reports as volume, including ours.
The second group is usually missed entirely. Tokens are minted into an address, and a short time later they are burned out of it again. The median gap between the mint and the burn is five seconds. No transfer between two addresses ever happens, so a dataset built to count transfers records nothing at all for this activity. I call this settlement activity.
How long a settlement address holds its tokens
One observation per rail address: the gap between its first mint and its first burn.
Note and method
Every Brazilian rail address, April 2025 when the rail opens to August 2026; the population is in practice the celo BRLA settlement contract, which carries 99.9 % of settlement rows. Median gap 5 seconds; of the 65,849 addresses that burn, 96.6 % burn within a minute and 98.9 % within an hour. Addresses that never burn are not on this chart. This is the pattern the settlement population is defined by, and it is why a mint and a burn on the same address are read as two legs of one movement rather than as two unrelated events.
What does that pattern represent? There are two plausible readings and the data cannot separate them. It could be a payment: money arrives by PIX, tokens are minted, tokens are burned, money leaves by PIX. Or it could be a currency operation: money arrives by PIX, tokens are minted, they are converted to dollars, and the reais are burned. Either way it is money moving, and either way a transfer-counting dataset shows zero.
Two ways a payment can cross the chain
In the first, the tokens move to another address and that move is the row a dashboard counts. In the second, the tokens are minted into the customer's address and burned from the same address, and no transfer between two addresses ever happens.
In the first flow the money crosses the chain as a transfer between two addresses, and that transfer is the row every published series is built from. In the second flow there is no transfer between addresses at all. There is still a real settlement leg happening on chain: the mint and the burn on one address are the on-chain record of reais entering the token and leaving it again. A dataset that counts transfers records neither leg, so the settlement is not undercounted, it is absent.
This second kind of activity is relevant in size, and it has grown over time, from a handful of rows a month to a population that now runs beside the peer transfers it is invisible to.
Peer transfers and settlement activity, by count
Monthly transfers and mints, log scale.
Note and method
All Brazilian tokens, from April 2023, where the settlement series actually begins: before that it is a handful of stray rows a month. Log scale, because the settlement population starts four orders of magnitude below the peer one and ends beside it. The settlement population is essentially one contract: BRLA on celo is 99.9 % of its rows and 99.2 % of its value over this span, and it opens in April 2025.
Peer transfers and settlement activity, by value
Monthly value in reais, log scale.
Note and method
The same two populations by value moved, in native reais, on the same log scale and the same start date.
Why only 2026
Every number here comes from a single eight-month window, 1 January to 31 August 2026. The data goes back to May 2021, so that choice throws away more than four years. It is worth saying why.
The market changed underneath the measurement, and the two halves of it changed in opposite directions.
These are Teiten's own figures, computed from the on-chain transfer records of Brazilian stablecoins on nine EVM chains and the XRP Ledger; the monthly series behind them is published with the data.
Take transfers of R$ 10,000 or more. In 2022, the first year with real volume, they carried 73.3 % of all Brazilian peer value. In 2026 they carry 94.9 %, and transfers of R$ 100,000 or more alone carry 86.0 %. Meanwhile those same large transfers are still only 3.8 % of the transfer count, and nearly half of all 2026 transfers are under R$ 10.
So the value concentrated into a small number of large transactions while the count stayed small-ticket, and the gap between the two widened. Averaging 2022 and 2026 describes neither.
It is also not a clean trend, and that matters more than the direction. The large-transfer value share peaks in 2023, falls back hard through 2024, and climbs past it again in 2025 and 2026. A five-year average would smooth over a market that was not doing one thing. One window inside it is a cleaner object to describe.
| year | volume, ≥ R$ 10,000 | volume, ≥ R$ 100,000 | count, ≥ R$ 10,000 | count, under R$ 10 | transfers |
|---|---|---|---|---|---|
| 2021 | 34.0 % | 6.4 % | 1.3 % | 8.4 % | 5,875 |
| 2022 | 73.3 % | 40.7 % | 0.9 % | 42.9 % | 468,276 |
| 2023 | 89.0 % | 72.2 % | 1.1 % | 56.9 % | 356,175 |
| 2024 | 74.7 % | 34.8 % | 2.2 % | 26.0 % | 3,368,764 |
| 2025 | 91.3 % | 77.8 % | 3.2 % | 24.0 % | 4,180,914 |
| 2026 | 94.9 % | 86.0 % | 3.8 % | 48.1 % | 3,969,116 |
The share of each year's peer volume, and of its transfer count, falling in the named size band; sizes are native reais, and the population is every Brazilian token, the XRP Ledger included, standard exclusions applied. The last column is the year's transfer count. 2021 holds 5,875 rows and is not a base for comparison; 2026 covers January to August.
The split
Among the Brazilian token-and-chain combinations with enough activity to analyse, eleven move R$ 36.6 billion in peer-to-peer transfers during the period. But that value is extremely concentrated: two of those combinations account for R$ 27.7 billion, roughly three quarters of all transfer value, through just 177 sending addresses. Meanwhile the two broadest combinations hold most of the wallets and little of the value: 86.2 % of the senders and 13.9 % of the value.
By number of transfers
Share of transfers per month, by transfer size in reais.
Note and method
Every Brazilian peer transfer, monthly to August 2026, native reais, nine size bands, each month normalised to 100 %. Population is peer transfers only, standard exclusions applied; the settlement population is not on this chart. This panel splits each month by how many transfers fell in each band.
By value moved
Share of value per month, same months, same bands.
Note and method
The same months and the same bands, split by how much value fell in each instead of by how many transfers did. The two panels tell opposite stories.
The gap is even clearer within BRLA on polygon: 81.8 % of senders transact fewer than three times and collectively account for only 2.7 % of value, while five wallets carry half of that series' value.
The implication matters. There is no single group of wallets that can meaningfully be called "Brazilian stablecoin users". The wallets that make up the broad user base are not the same wallets that generate most of the economic volume, and treating them as one group hides that difference.
The volume that is not published
The largest Brazilian stablecoin flow in this dataset contains no transfers.
BRLA on celo has a settlement mechanism that works differently from an ordinary wallet-to-wallet payment. Tokens are minted into an address and rapidly burned from the same address: the median address burns five seconds after its first mint, and 98.9 % burn within an hour. This only becomes visible when the market is split into peer transfers and settlement. Without that split, the flow disappears from a transfer-based view.
Because the tokens never move between addresses, a conventional transfer dataset records zero volume. But economically, value entered the token system and left again. This analysis therefore treats qualifying mint-and-burn activity as settlement, while separating mints whose tokens are subsequently held.
The two celo BRLA contracts' ledgers track each other closely, 0.04 % apart, with monthly net differences below R$ 0.3 million since April 2025, while one records almost twice the gross movement of the other. That supports reading the settlement contract as an internal-balance mirror rather than new issuance: its gross mints and burns are payment flow, not supply creation. Treating each of those mints as a settlement is the platform rule's premise, not a finding of this work.
The scale is substantial. From January to August 2026 the rail accounts for R$ 18.4 billion across 64,884 addresses and 3.59 million mints. In the first quarter alone it settled R$ 2.8 billion, against just R$ 0.5 million of corresponding peer transfers. Two narrower definitions of the rail give R$ 14.1 billion and R$ 10.2 billion, and none of the three is a lower limit.
Supply data misses it too, because a mint and its burn fall between two readings of a series whose finest grain is a day. The rail minted R$ 18.45 billion and burned R$ 18.35 billion, leaving only R$ 104.3 million of net supply change. R$ 18.4 billion moved while R$ 104 million appeared in the supply balance, a 177-fold difference. A supply series would show the latter and miss the former.
This is not another way of counting the same transfers. The two never count the same transaction: of 3,588,628 transactions carrying a mint on the settlement contract and 349,324 carrying a peer row, none carries both.
The settlement flow is also overwhelmingly payment-shaped. 82.1 % of its value sits in pass-through addresses, which burn approximately what they were minted. Another 9.6 % sits in payout addresses, which burn 4.03 million times against 244,480 mints. The remaining 8.3 % is held by 266 accumulators, which take 45.0 % of all rail mints and return only 8,830 burns against 1,613,221. That share is holding rather than paying, and it is the part to exclude when reading the rail strictly as payment activity.
Peer transfers, by hour and weekday
Hours in America/Sao_Paulo, weighted by value.
Note and method
Window 2026-01-01 to 2026-08-31, all Brazilian tokens, standard exclusions applied. Each grid is normalised to its own total, so the two are comparable in shape and not in size. Both carry a working-week shape: Monday to Friday takes 96.2 % of peer value and 94.2 % of settlement, and the twelve hours from 08h to 20h take 93.5 % and 90.0 %.
Settlement, by hour and weekday
The same hours and the same weighting.
Note and method
They do not keep the same hours, though. By value the peer population peaks at 09h and the settlement population at 11h; by number of transfers, the other reading the toggle offers, the peaks move to 12h and 19h. A clock alone cannot tell you which of them is a payment, which is why the case rests on the mint-and-burn pattern rather than on the hours.
The difference becomes clearer against another token with a similar economic process and a different contract design. On BRZ on gnosis a customer can spend through an intermediate hop into a burn contract. That intermediate transaction creates a wallet-to-wallet transfer, so a transfer-based dataset counts it as volume. On BRLA on celo the burn happens directly from the customer's own address, so there is no transfer between addresses and the same economic movement is invisible. The underlying economic activity can therefore look different, or disappear entirely, depending on how a token's contracts implement settlement.
The result is a measurement gap. A dashboard built only on wallet-to-wallet transfers can show Brazilian stablecoin activity without showing a large part of the value actually being settled.
| month, 2026 | peer volume, R$ millions | issuance-settled at token grain, R$ millions |
|---|---|---|
| January | 0.4 | 876.2 |
| February | 0.0 | 986.8 |
| March | 0.1 | 956.6 |
| April | 249.8 | 864.4 |
| May | 1,429.0 | 1,471.1 |
| June | 3,947.5 | 4,396.2 |
| July | 3,689.0 | 4,011.6 |
| August | 4,576.1 | 4,887.0 |
| January to March | 0.5 | 2,819.6 |
| January to August | 13,891.9 | 18,449.9 |
Figures are millions of reais. The peer column is all the published dataset shows. For January to March, peer volume is 0.02 % of peer plus settled value under all three grains; for January to August it is 42.95 % at the rule's own grain, with the two narrower definitions at 49.60 % and 57.62 %.
The rail is young, and most of what it has ever settled it settled this year. It settled R$ 2.4 billion across the nine months of 2025 in which it operated, and R$ 18.4 billion across the first eight months of 2026, close to eight times as much in fewer months.
That is what the next chart adds to the whole Brazilian market rather than to one token. Across the eight months of the window the published transfer series carries R$ 36.8 billion and the rail settles R$ 18.5 billion more, so a third of everything that moved — 33.4 % — is in the band the dataset cannot see. Over the rail's whole life, from April 2025, it is R$ 57.1 billion published against R$ 20.9 billion unpublished, or 26.9 %, and the share climbs month by month as the rail grows.
Before and after the settlement layer
Every Brazilian token on nine EVM chains and the XRP Ledger, monthly, from April 2025 when the rail opens. The striped band is the addition, and over this window 99.3 % of it is a single contract: BRLA on celo.
Note and method
The solid band is what the published transfer series measures; the striped band is the settlement rail, which it does not. Across the frozen window the two are R$ 36.8 billion and R$ 18.5 billion, so 33.4 % of the total is unpublished; since the rail opened in April 2025 it is R$ 57.1 billion against R$ 20.9 billion, or 26.9 %. Both columns are native reais and both are this whole market, so unlike the celo close-ups above they are directly comparable to the BRL series on /data/stablecoins once that one is read in reais rather than dollars. The peer column here is R$ 36.8 billion against the R$ 36.6 billion this post opens with: the opening figure counts the eleven populations with enough activity to analyse, and this counts every Brazilian token. The two are disjoint at transaction level, so the bands stack rather than overlap. The settlement band is almost entirely one contract — BRLA on celo is 99.3 % of its value across this chart's window, and 99.2 % over the earlier chart's span back to April 2023 — so what is being added to the whole market here is, in practice, what one token settles.
What a dashboard cannot show you
The public dashboard is not wrong. It is measuring one layer of the market.
A transfer dashboard tells you what moved between addresses. It does not necessarily tell you what economic activity that movement represents, who is behind it, or whether the addresses involved are users at all.
That distinction matters. In this dataset, R$ 18.4 billion of settlement activity never appears as a transfer, because the tokens are minted into an address and burned from that same address without moving between wallets. At the same time, a large share of the wallets that appear in transfer data contribute very little of the economic value, while a small number of addresses account for a disproportionate share of it. And some of the activity comes from addresses whose behaviour suggests infrastructure, contracts, or other automated processes rather than individual users.
A wallet-transfer view is only a partial measurement of the market. It can tell you how many transfers occurred and how much value crossed from one address to another. It cannot, by itself, tell you how much economic activity took place, how that activity was settled, or how many actual users were behind the addresses.
That is the question the next post takes on: what are these wallets actually doing?
Methodology. Skip it unless you want to check the work.
The window is 2026-01-01 to 2026-08-31 UTC, hours read in Brazil's own time zone. Four classes of row are excluded before anything is counted: the zero address on either side, rows from eleven treasury addresses on one Brazilian token, rows from two issuer wallets on another, and one token out of scope by rule. Nobody is named: this work has zero publishable identities, every address label it holds is classification-only, no address appears at any length, and no entity, operator, venue, app or person is named as the party behind one. Two coverage facts travel with their numbers rather than sitting in a footnote: BRZ on gnosis is observed from 1 January to 1 May only, and wBRL on worldchain is a 31.8-day series. And one sentence the whole series rests on: nothing measured here runs on a schedule. The one criterion that tests for a fixed interval flags at most 0.11 % of rows in any of the combinations studied, transaction by transaction, and the only indicator of automated behaviour covers at most 5.0 %.
The method behind the settlement layer is in the companion note, and the monthly series are published with the data. Every number in this post was computed from Teiten's own on-chain transfer data and checked by an independent review and a re-derivation completed on 2026-09-20.