iGaming Pulse #53

Three operators reported into the same fortnight and all three describe the same market from different positions. Bally's Intralot grew UK online 11.6% in constant currency in Q2 — accelerating from 10.5% in Q1 — in the first full quarter under the new rate, absorbed a €34m quarterly duty hit and mitigated 65% of it. Its CEO said plainly that consolidation "has been slower to arrive than we anticipated" because "competitors have been more willing to absorb margin compression for longer than their arithmetic suggests they should." Entain told this database on 14 August that the tax rise was helping it gain share. Evoke's CFO said on 13 August that handling the rise was more than a cost-cutting exercise. Three operators claiming outperformance, and no one naming the losers.

Buzz Bingo names the mechanism. It took an £18.4m goodwill impairment on its online business after revising cash-flow forecasts for the 40% rate, and stated that the loss of viability of standalone online acquisition has accelerated its pivot to converting club customers into omnichannel players. That is not mitigation language. That is an operator concluding a channel no longer works and reallocating out of it.

Read that against Superbet in Romania, which on 1 September converted a retail estate that can no longer take bets into a network that can still take cash deposits into online accounts, and against Bally's Intralot's own claim that year-on-year player volumes held and "we did not lose customers to competitors, and we did not lose them to the unregulated market." Both are the same trade: substitute owned physical distribution for bought digital distribution.

Commercially, this means the UK online market is quietly repricing around who owns footfall. If paid acquisition at 40% duty only works above a scale threshold, the survivors are the large digital-only players who can absorb CPA inflation and the omnichannel players who can bypass it — and the squeezed middle is the sub-scale digital-only operator, exactly the cohort everyone predicted would exit and which is instead sweating assets. Reeves' own framing is the best available forecast: "deferred is not cancelled. A 20% to 25% margin business paying 40% of its revenue in duty does not have an indefinite runway."

The Ninth Circuit did not weigh Kalshi's exchange architecture and find it insufficient. It ruled the architecture irrelevant: "Kalshi argues that unlike a traditional casino or house, Kalshi's DCM is not counterparty to any trade, does not set betting odds, does not profit when its customers lose… These are distinctions without differences. Most importantly, none of those distinctions have any connection to the statutory definition of swap." When the structural defence is ruled out, what remains is what the product does — and what it does is take parlays on football.

The geofencing holding compounds it. Nelson's finding that "regulated entities in Nevada use geofencing. Kalshi could do the same; it just refuses to do so and instead seeks a competitive advantage over its competitors" reframes national access from a technical constraint into a commercial choice, and every subsequent state action can cite it.

Meanwhile the sector is hedging through product mix rather than law. Crypto.com and PYMNTS are launching 20 AI-adoption contracts in September with 25 more per quarter, settled against proprietary primary research — non-sports volume with a settlement source the venue controls commercially. That is the shape of a defensible prediction market: one where the sports share is not 78%.

The revenue trajectory is unambiguous: R$8.7bn in taxes from licensed operators January-July 2026 against R$4.9bn in the same 2025 period; R$9.95bn for full-year 2025 at the 12% GGR levy; the rate legislated to 15% next year and 18% by 2028; and ~R$37bn GGR in 2025 with ~R$4bn to statutory public beneficiaries. Housing reform and other multi-billion-real commitments are drawn against it.

Against that, the political construction is equally unambiguous. Lula's 1 September meeting included ten critic organisations and excluded ANJL, IBJR, the SPA and any substantive Fazenda intervention. He described the decision as two-thirds made before hearing anyone. 56% of Brazilians — 62% of women — support a total ban. Elections are 4 October.

HYPOTHESIS, not established causation: the government's preferred landing point is severe advertising restriction, because it is the only intervention that reads as decisive action, polls well, costs the treasury nothing directly, and leaves the tax base intact. Superbet's CEO Alexandre Fonseca has already identified why operators will fight it hardest: "My only differentiation as a regulated betting house in Brazil is being able to advertise widely." That is also, uncomfortably, the argument for why an advertising ban would damage channelisation — the same argument VNLOK is making in The Hague this week.

One causal claim I am explicitly NOT making: the sector's tax contribution rising 76.9% does not itself prove the market grew that much. Licence-fee timing, the first full year of enforcement, and the base effect of a market that only opened in January 2025 all inflate the comparison. The direction is clear; the magnitude is not clean.

The Netherlands: parliament debates a total advertising ban on 3 September. VNLOK's counter is that the illegal market is now estimated to be the same size as the licensed one, on the regulator's own survey, and that duty collected fell after the rise to 37.8% in January 2026.

Brazil: CONAR's revised Annex X now requires media outlets and platforms to be notified when an unauthorised operator advertises; a Senate bill on advertising restriction is under contested debate; ANJL and IBJR are preparing action against big tech over illegal betting ads; and the president has signalled a "drastic" decision.

The UK: not advertising per se, but the same squeeze from the cost side — Buzz Bingo has concluded standalone paid online acquisition is no longer viable at 40% duty, which removes a substantial buyer of certain digital inventory.

Conversational AI: OpenAI's policy update closes the channel twice over — gambling ads are prohibited, and gambling is classified as a context in which no ads should be served at all. The second rule is novel and worth noting for how AI platforms will treat regulated categories generally.

The common thread is not a coordinated campaign; it is four independent responses to the same underlying problem — that illegal operators advertise on the same platforms as licensed ones and platforms have not distinguished between them. VNLOK's Meta complaint, ANJL/IBJR's big-tech action and CONAR's notification duty are all attempts to force that distinction. This is a real mechanism, not a lexical connection: in three of the four cases the trade body is explicitly offering to accept restrictions on licensed advertising in exchange for enforcement against unlicensed advertising.