Developing Brazil Turned Off Its Betting Industry and Found 86.2 Million Accounts Holding Less Than a Real Each

Editor's note: TrueEdge builds odds tools and earns affiliate commissions from licensed sportsbooks. This piece argues a position on betting regulation. Weigh it accordingly.

Brazil's licensed betting sites went dark on October 6. Operators had until October 5 to let customers withdraw; Provisional Measure 1.394/2026, signed by President Lula on September 25, banned the operation, offering, intermediation and advertising of fixed-odds betting, and the blocks followed.

Shutting an industry down forces it to count itself. What Brazil's Finance Ministry has now published, in the course of working out whose money has to go back, is the first externally verified picture of who was actually in that market. It does not resemble the market either side has been describing.

The census

Roughly R$1.325 billion — about US$264 million — is owed back to approximately 26.5 million customers. Banks process the returns between October 9 and 14, with the state-owned Caixa Econômica Federal picking up anything that cannot be routed through normal channels.

Then the distribution. Finance Ministry deputy Dario Durigan's numbers: about 1 per cent of bettors held 80 per cent of outstanding balances. Roughly 3.4 per cent of accounts held about 90 per cent. And 86.2 million accounts held less than one real each — about twenty US cents — R$15.5 million between all of them.

Eighty-six million accounts with nothing in them. That is not a rounding error in a dataset. That is the dataset.

A caution before anyone does arithmetic on those two paragraphs, because they do not share a denominator and we are not going to pretend otherwise. The 86.2 million is an account count. The 26.5 million is a count of customers owed a refund. One is larger than the other, which means the ministry is measuring across different bases — most plausibly every registered account at every licensee against the subset with a balance worth returning, with duplicate registrations across operators inflating the first. We asked ourselves whether to compute a percentage from them and decided not to. Nobody publishing a tidy ratio off these two figures has reconciled them either.

Against this, the figures everyone has been quoting: R$37 billion in gross gaming revenue in the regulated market's first full year, 25.2 million people recorded as having placed a bet during 2025. Those are real numbers and they are not in tension with the new ones. They are measuring different things, and the whole Brazilian argument has been conducted as if they were measuring the same thing.

What the number can and cannot tell you

Here is the honest version, and it is the reason this piece is not a victory lap for either camp.

A balance at shutdown measures a stock. It is what was sitting in the account on October 5. It is not a measure of how much someone wagered, how much they lost, or whether gambling damaged their life. An account holding R$0.40 is consistent with a person who deposited R$20 once in 2025, played three times, got bored and left. It is equally consistent with a person who deposited R$8,000 across eighteen months and has nothing left.

We do not know which. The Finance Ministry's figures cannot distinguish them, and anyone telling you they can is selling something. That caveat is fatal to the strongest version of the industry's reading — "look, hardly anyone was really playing" — and it should be stated before the argument, not after.

What the figures do establish is narrower and still consequential: the money in Brazilian online betting was concentrated to an extreme degree, and the policy instrument Brazil reached for was built for a mass-participation problem.

Why the shape matters to the policy

Attorney General Jorge Messias has made the government's case on volume: a 140 per cent rise between 2018 and 2025 in gambling-addiction records in Brazil's public health system. The National Confederation of Commerce, whose 2024 petition became ADI 7721, puts betting-linked retail defaults at roughly R$143 billion between January 2023 and March 2026. Those are population-scale harm claims, and a prohibition is a population-scale remedy.

But the concentration data points somewhere else entirely, and the government's own investigators have followed it there. About forty accounts are now under review for money laundering and possible organised-crime links — some holding more than R$500,000, some over R$1 million. That is not a consumer-protection problem. That is a financial-crime problem, and it is the kind of thing a licensing regime with transaction monitoring is built to find, which is in fact how it was found.

The uncomfortable implication for the ban's architects: the regulated market generated the surveillance that produced these forty names. An unlicensed market, which is what Brazil now has by default, generates no such records. Blocking domains does not close demand; it moves demand to operators who file nothing with anybody. Brazil has run this experiment before, and so has every other country that has tried to switch a large grey market off by decree.

The uncomfortable implication for the industry: if 3.4 per cent of accounts hold 90 per cent of the balances, then the business was never the twenty-five-million-customer mass product the sector's own marketing described. It was a concentrated book with an enormous, nearly dormant mailing list attached. Every sponsorship deal, every stadium board, every statement about democratising entertainment was priced off a user count that the money does not support.

The carve-out that just appeared

On Thursday, Judge Cristiane Farias Rodrigues dos Santos of the 9ª Vara Cível Federal de São Paulo granted Stake Brasil a partial injunction: it may operate until October 25, the blocking of its domains is suspended, and authorities may not treat its authorisation as already extinguished. She refused the larger request — Stake wanted to run to December 31, 2029, when its licence expires — and she attached twelve conditions, among them no new promotional campaigns, no advertising aimed at minors, maintained identity verification, preserved transaction records, continued regulatory reporting, reserves sufficient to pay out, and a bar on wagers funded by social-assistance payments.

Read the conditions rather than the headline. A federal judge has just ordered one operator to keep doing precisely what the licensing regime required of it — reporting, verifying, segregating, not advertising to children — three days after the state switched that regime off for everyone. The injunction is an accidental argument that regulation was working on the operator it was applied to.

October 25 is not an arbitrary date. It is the day of the presidential runoff. Justice Luiz Fux, handling the consolidated challenges at the Supreme Federal Court — ADI 7721 from the CNC and ADI 7749 from the Prosecutor-General among them — gave the government 72 hours on October 2 to answer the suspension requests; the AGU filed at 22:10 Brasília time on October 5, hours before the sites went down. He has not ruled, and the reporting is that he does not intend to until after the runoff.

Congress has its own clock. Amendments to MP 1.394 close at 11:59 p.m. on October 13 — forty had been filed as of Monday — expedited procedure begins November 9, and the measure must be confirmed by November 23 or it lapses on its own. A provisional measure that expires without a vote takes the ban with it.

The thesis

Brazil has produced, by accident, the best dataset anyone has on what an online betting market actually consists of, and it is going to be read by every regulator and every operator in the world over the next month. The finding is not "nobody was gambling." The finding is that the money and the account count have almost nothing to do with each other, and that policy built on account counts — whether it is a ban justified by twenty-five million exposed citizens or a valuation justified by twenty-five million customers — was built on the wrong number.

What would change this reading: deposit-flow data. If the Finance Ministry publishes lifetime deposits and losses by account rather than closing balances, and the 86.2 million near-empty accounts turn out to have cycled significant money through, then the mass-harm case is vindicated and this analysis is wrong in the way that matters. That data exists. The operators filed it. Someone should ask for it before November 23.