Robinhood published its August operating data at 16:05 ET on 10 September, and the fastest-growing product in the company went backwards: 4.7 billion event contracts traded, down 23% on July. The obvious question is whether prediction markets have broken. It leads nowhere, because a month is not a quarter and Robinhood books this revenue quarterly. The question that pays is arithmetic. Two of the September quarter's three months are now on the record and cannot change, so what does the third one have to do?
Less than any month since February.
Two months are booked, and September needs 93 million contracts a day
Three disclosed figures. Robinhood traded 13.6 billion event contracts in the quarter ended 30 June, alongside $156 million of revenue. July was 6.1 billion. August was 4.7 billion.
July and August together are 10.8 billion, so September has to beat 13.6 − 10.8 = 2.8 billion for the September quarter to come in above the June quarter.
Convert that to a daily rate, the basis Robinhood's own monthly data uses. September has 30 days, so 2.8 billion ÷ 30 = 93 million contracts a day, against August's 152 million. The daily rate has to fall 39% for the quarter to shrink. The steepest monthly fall in Robinhood's 2026 data is 29%, in February.
| September | The quarter | vs June quarter |
|---|---|---|
| 91m a day (−40% on August) | 13.5bn | −0.5% |
| 117m a day (−23%, another August) | 14.3bn | +5.2% |
| 139m a day (−9%) | 15.0bn | +10.0% |
| 152m a day (flat on August) | 15.4bn | +12.9% |
| 182m a day (+20%) | 16.3bn | +19.6% |
September's rate is the only assumption in that table and the softest figure here; everything else is subtraction. A September that repeats August's fall exactly still leaves the quarter up 6%.
Those averages run on calendar days, not trading days: 4.7 billion over 31 days is the 152 million Robinhood reported, and the same division reproduces every 2026 month it has published.
February fell 29% and the line was 2.7 times bigger four months later
Robinhood has published this volume every month of 2026: 3.4 billion in January, 2.4 billion in February, then 3.0, 3.2, 3.9, 6.4 billion in June, 6.1 in July and 4.7 in August. Those eight sum to 33.1 billion, consistent with the "more than 30 billion" through August Robinhood disclosed on 8 September, and the first three sum to the 8.8 billion reported for the March quarter.
February was a steeper fall than August, and by June the line was 2.7 times its February level. August ran 21% above May, the best month before the tournament, on the total and the daily rate alike.
The World Cup is the obvious candidate for that shape. It ran from 11 June to 19 July, and Robinhood said on 8 September that more than 5 billion contracts traded on it, against 12.5 billion across June and July combined. August had no tournament, and the NFL season did not open until 9 September. Robinhood attributes the decline to nothing — not a sports calendar, not a fee change, not a state regulator — and two numbers moving together is not a cause.
Crypto's 61% month leaves it needing another July
The same arithmetic runs backwards on the line underneath. Crypto notional was $17.5 billion in August, up 61% on July and down 38% year over year — $7.4 billion in the app, $10.1 billion on Bitstamp.
A 61% month sounds like a recovery. Convert it. July was $10.9 billion, so July and August together are $28.4 billion against $40 billion in the June quarter, leaving September to clear $11.6 billion — more than the whole of July — merely to match. The bounce got the quarter back to roughly par.
A monthly percentage tells you almost nothing about a quarter until you convert it into the level the remaining month has to reach. Down 23% is a quarter still growing. Up 61% is a quarter still shrinking.
Volume is not revenue, and the rate is not published until late October
The strongest case against all of it: the contract count can grow while the money does not. Robinhood publishes volume monthly and revenue quarterly, so revenue per contract for July and August is unknown until the quarter is reported. That rate has run just above a cent — 1.147 cents in the June quarter, and why it sits above Robinhood's own 1-cent cap is a separate argument. A quieter month with a different event mix could carry a different rate, and no monthly disclosure would show it. The September quarter almost certainly trades more contracts than June's did; that is not the same as earning more.
None of it moves our forward model, which grows this line 10% a quarter from 13.6 billion. That implies 14.96 billion, needing 4.16 billion in September, or 139 million a day — 9% below August's daily rate, so the print left our assumption looking conservative, not stretched. Its 60-billion-a-quarter ceiling does not bind until 2030. What a real stall would cost is already on the record: taking volume growth to zero removes $11.19 a share, about 12% of fair value. One month does not establish that.
September prints in mid-October on the same monthly cadence. Below 2.8 billion contracts — 93 million a day — the quarter shrinks and this is wrong.
Monthly volumes come from Robinhood's operating data releases of 12 March, 13 May, 9 June, 12 August and 10 September 2026, each stating that current-quarter months are unaudited and preliminary. The June quarter's contracts, revenue and crypto notional are from the second-quarter results of 29 July, and the March quarter's 8.8 billion from the first-quarter results of 28 April. Cumulative volumes and the World Cup figure were released on 8 September with Robinhood's OG.com announcement; sporting dates are the published calendars. Robinhood offers no explanation for August's decline. The thresholds, the daily-rate conversions and the September range are ours, by subtraction from the disclosed totals; the 10% growth rate and the $11.19 sensitivity are our model's, not guidance.
