Robinhood's Event Contracts: 13.6 Billion a Quarter at 1.15 Cents Each, and the Exchange Takes More

Abstract

An event contract is the simplest instrument in retail finance. It pays $1 if something happens and $0 if it does not, it trades in cents in between, and the price is the probability. What is not simple is who gets paid when one changes hands, and the answer is that two different companies charge two different fees on the same trade, computed by two different formulas, and the larger of them usually belongs to the one the customer has never heard of.

Robinhood publishes its half. Since 1 June 2026 its commission has been k × p × (1 − p) × c — 10% of price times one-minus-price per contract, 5% for Gold subscribers — rounded up to the nearest cent and, decisively, capped at $0.01 per contract. The cap is what matters. A quadratic that peaks at 2.5 cents and a ceiling of 1 cent produce a flat line across most of the board: for a customer without Gold, the cap binds on every contract priced between 11.27 cents and 88.73 cents, which is 77.5 points of a 100-point range. Robinhood publishes a curve and charges a constant.

The venue charges the second fee, and it is not capped. Kalshi's schedule, as filed with the CFTC, is round up(0.07 × C × P × (1 − P)) — the same quadratic shape, a lower coefficient, and no ceiling. It peaks at 1.75 cents on a 50-cent contract, against Robinhood's 1.00. Set the two curves on the same axis and they cross twice, at 17.27 cents and 82.73 cents. Between those points — 65.5 points of the range, and where nearly all contested markets trade — the exchange earns more per contract than the broker does. That is the economic fact underneath Robinhood's announcement on 8 September that it is taking an equity stake, of undisclosed size, in one.

Then there is a number that does not fit. Robinhood's commission cannot exceed a cent a contract per trade, and yet the revenue it reports divided by the contracts it reports has been above a cent for four straight quarters: 1.174, 1.188, 1.182 and 1.147 cents from the September 2025 quarter to the June 2026 one. The overage is not noise — it runs 12.8% to 15.8% of the line every quarter, $20.0 million in the June quarter alone. Only two ordinary things explain it. Either a "contract traded" is counted once while commissions are charged on both the opening and the closing leg, in which case the figure implies that at least 14.7% of contracts were closed before settlement — a behavioural statistic about retail prediction-market traders that nobody publishes. Or some of the money is coming back from the venues, in which case Robinhood was already earning at the exchange layer before it bought into one. Robinhood discloses neither, and we are not going to pick.

The rest of this is the mechanism: what the four layers of a prediction market are and which two of them charge, why a fee shaped like p(1 − p) taxes the uncertain trade and a fee shaped like a percentage of stake taxes the confident one, how five venues charging five different shapes turn the identical winning trade into a bill of anywhere between 2 cents and $8, and what a 22-year-old exchange licence that began life as HedgeStreet is now worth to the broker routing football contracts into it.

Table of contents

1. The vertical, in volumes and in dollars
2. Four layers touch one contract, and two of them charge
3. Robinhood's commission is a quadratic that a cap flattens
4. The second fee, and whose it is
5. Where the venue out-earns the broker
6. The same fee read as a share of the money at risk
7. Five fee shapes, five different traders taxed
8. What Robinhood actually books, and why it sits above its own ceiling
9. Twenty-two years of one licence, and what Robinhood bought into
10. Implications

1. The vertical, in volumes and in dollars

Start with the size, because the fee arithmetic that follows only matters at this scale.

Robinhood began listing event contracts in October 2024, with the presidential election. Six quarters of revenue later the line reads: $3 million in the March 2025 quarter, then $10 million, $27 million, $101 million, $104 million, and $156 million in the quarter ended 30 June 2026. That is 52 times in six quarters, and it is why the chart below runs on a logarithmic axis — on a linear one the first three bars would not be visible at all.

The volumes underneath it are public in two forms. Quarterly: 2.3 billion contracts in the September 2025 quarter, 8.5 billion in December, 8.8 billion in March 2026 and 13.6 billion in June. Monthly, from the operating data Robinhood publishes: 0.3 billion in July 2025 at an average of 10 million a day, against 6.4 billion in June 2026 at 213 million a day and 6.1 billion in July 2026 at 197 million. Year on year that is twenty times, on both the monthly total and the daily average.

The cumulative figures Robinhood released alongside the OG.com announcement on 8 September fill in the rest: more than 12 billion contracts in calendar 2025, more than 30 billion in 2026 through August, more than 45 billion since launch about two years ago, and more than 5 billion on the World Cup alone.

Against the company, the line is no longer small. The $156 million in the June quarter is 20.1% of Robinhood's $776 million of transaction revenue and 11.9% of its $1,308 million of net revenues — the second-largest transaction line, behind options and ahead of equities and crypto. Spread across 28.5 million funded customers, the June quarter works out to 477 contracts and $5.47 of revenue per funded customer, and July to 214 contracts each, about seven a day.

Two things in that series are worth marking before the mechanism. The first is that July was the first sequential decline — 6.1 billion against June's 6.4, down 5% — and it came at the end of a World Cup that ran from 11 June to 19 July and contributed more than 5 billion contracts on its own. June and July are the tournament. The second is that August is the one month with no published figure. The disclosed months through July already account for 28.5 billion contracts in 2026, so "more than 30 billion through August" bounds August only at more than 1.5 billion — a floor so far below the recent run rate that it establishes nothing. Robinhood published a cumulative number for the year on the day it announced a new venue, and left out the single month that would say whether the growth survived the football.

2. Four layers touch one contract, and two of them charge

Robinhood's own explainer sets out the stack, and it is worth taking from the company because it is the company arguing for the least of it. Exchanges — "such as Kalshi or ForecastEx" — "sit at the core of prediction markets. Their job is to create the marketplace itself." They "define the event contract — its exact wording, its settlement, and the rules for how the outcome will be resolved." What they do not do is "predict outcomes or set prices." A broker/futures commission merchant, "like Robinhood, acts as the intermediary between customers and the exchange, providing access to markets, safeguarding customer funds," and does not "create event contracts, decide how they resolve, or control where prices move." Market makers quote both sides. Underneath all of it sits a clearing house, which stands between buyer and seller and holds the collateral.

That collateral is the reason the instrument is clean. Both sides of an event contract post their maximum loss up front — a Yes at 63 cents and a No at 37 cents together fund the single dollar that one of them will receive — so the clearing house never carries directional risk and there is nothing to margin-call. It is why the product could be built for retail at all.

Two of those four layers bill the customer. The broker charges a commission. The exchange charges a trading fee. In Robinhood's phrasing on its own fee page: "Exchanges typically charge an additional fee, which varies by exchange and can be up to $0.01 per contract," and "these fees apply to both opening and closing trades."

What Robinhood's explainer never says — and this is the gap the rest of this piece fills — is which of those two layers earns more. Its own article defines what an exchange does without once mentioning that an exchange is paid for doing it.

3. Robinhood's commission is a quadratic that a cap flattens

On 1 June 2026 Robinhood replaced a flat per-contract commission with a formula. The company states it as:

Commission = (k * p * (1 − p) * c)

with "k = tier-based constant (5% with Gold or 10% without Gold)", p the contract price as a decimal and c the number of contracts. Then two clauses that do more work than the formula: "We round up to the nearest $0.01 on all trades with a maximum commission rate of $0.01 per contract."

Read the intent first. The p × (1 − p) term is the variance of a single binary outcome — it is largest at 50 cents, where nobody knows, and collapses toward zero at 1 cent and 99 cents, where everybody does. A fee shaped that way says: we charge you for uncertainty, and we charge you almost nothing to trade a near-certainty. That is a defensible design, and it is the same one the exchanges use.

Now apply the ceiling. Without Gold, the uncapped commission reaches a cent when 0.10 × p × (1 − p) = 0.01, which is p(1 − p) = 0.1, which happens at 11.27 cents and 88.73 cents. Between those two prices the formula wants more than a cent and the cap refuses. So across 77.5 points of the range the commission is not a curve at all — it is a flat 1.00 cent. With Gold at k = 5% the cap binds over a narrower band, 27.64 to 72.36 cents, still 44.7 points.

The published curve therefore describes the behaviour of the fee only in the tails. A contract at 4 cents costs 0.384 cents to trade; a contract at 40 cents and a contract at 60 cents both cost exactly 1.00, and so does everything between them. In practice Robinhood charges a penny.

The rounding rule has a sting of its own at small size. Commissions round up to the nearest cent per order, so a single contract bought at 2 cents — an uncapped commission of 0.196 cents — is billed a full cent. On a 2-cent stake that is a 50% fee. It disappears at scale and it is brutal at one.

4. The second fee, and whose it is

The exchange's fee is not Robinhood's, does not appear in Robinhood's revenue, and is larger.

Kalshi's schedule, in the version filed with the CFTC, sets the general trading fee at:

fees = round up(0.07 x C x P x (1-P))

with the same definitions — P the price in dollars, C the number of contracts, and "round up = rounds to the next cent." Two structural features distinguish it from the broker's. Fees "are only charged for orders that are immediately matched with orders sitting on the orderbook" — the taker pays and the maker does not. And there is no cap. The quadratic runs its full course, so the fee peaks where the formula peaks, at 0.07 × 0.25 = 1.75 cents on a 50-cent contract. Kalshi's own published table prints it: 100 contracts at 50 cents cost $1.75, at 10 cents cost $0.63, at 99 cents cost $0.07.

The coefficient is not universal. Kalshi files a halved multiplier, 0.035, for the S&P 500 and Nasdaq-100 families, and describes the general figure as the baseline from which specific products depart. Sports, the category Robinhood began routing to OG.com, runs the 7% coefficient.

Set against that, the sentence on Robinhood's fee page is oddly small: the exchange fee "can be up to $0.01 per contract." Kalshi's own filed schedule says its fee reaches 1.75 cents. Either Robinhood's routed order flow pays a member rate rather than the retail table, or the venue absorbs the difference, or Robinhood does. Nothing in either document says which, and it is the first thing a disclosure would settle.

5. Where the venue out-earns the broker

Two fees, two curves, same axis. The broker's is a quadratic with a lid at 1.00 cent. The venue's is a shallower quadratic with no lid at all, running to 1.75.

They cross twice. Below 11.27 cents and above 88.73 cents Robinhood's cap is not binding, both fees are quadratics, and the broker's 10% coefficient simply beats the venue's 7% — Robinhood earns more. Between 17.27 cents and 82.73 cents the venue's uncapped fee has climbed past the broker's flat penny and the ordering reverses. In the narrow bands either side of that, 11.27 to 17.27 and 82.73 to 88.73, the broker is capped and the venue has not yet caught up.

The middle band is 65.5 points wide, and it is where contested markets live. A football contract in the week before a game, an election market with a real race in it, a rate decision the committee is split on — all of them price between 20 and 80 cents, and on every one of them the exchange collects more per contract than the broker that sourced the customer, built the app and carries the account. At even odds the split is 1.75 to 1.00, so the venue takes 64% more than the broker on the same trade.

That is the whole commercial logic of the announcement of 8 September, and it does not need the stake to be sized to be legible. Robinhood has spent two years building the demand side of a market in which the supply side has the better fee schedule. Buying into a venue is the obvious response to the shape of that chart, whatever the size of the cheque.

For a Gold subscriber the asymmetry is starker still. At k = 5% Robinhood's uncapped curve sits below Kalshi's everywhere, and the cap only pulls it down further in the middle. On a Gold customer's 50-cent trade the broker earns 1.00 cent and the venue 1.75 — except that Gold cost the customer a subscription, and the subscription revenue lands in a different line.

6. The same fee read as a share of the money at risk

Cents per contract is the venue's view. It is not the customer's. A customer who buys a contract at 5 cents has 5 cents at risk; a customer who buys one at 95 cents has 95. The same absolute fee is a different tax on each.

Divide the fee by the stake and the quadratic turns over. Kalshi's 0.07 × p × (1 − p) divided by p is simply 0.07 × (1 − p) — a straight line falling from 7% of stake on a 1-cent longshot to 0.07% on a 99-cent near-certainty. A factor of a hundred, and it runs in exactly the opposite direction to the per-contract chart.

The all-in cost on Robinhood behaves the same way, more steeply. On a 1-cent contract the commission plus a 1-cent venue fee is 16.83% of the money at risk, one side. At 10 cents it is 15.3%. At 50 cents, 4.0%. At 99 cents, 0.17%.

So the fee shape that looks like it protects the confident trader — charge little where the outcome is nearly known — is, as a share of what the customer actually puts up, a tax that falls almost entirely on longshots. The 2-cent contract on an unlikely outcome is the most expensive thing on the board by an order of magnitude, and it is also the kind of contract a retail prediction market sells most of.

Round-tripping doubles it. A hundred contracts bought at 50 cents and sold before settlement is $50 of stake against $1.00 of commission and up to $1.00 of venue fee on the way in and the same on the way out: $4.00, or 8.0% of stake. The identical trade placed directly on Kalshi, paying no broker, costs $3.50 — 7.0%. Retail intermediation costs a percentage point of the stake on a coin flip, and the broker keeps it.

7. Five fee shapes, five different traders taxed

The interesting thing about this market is not that fees differ in size. It is that they differ in kind. Five venues charge on five different bases, and the base decides who pays.

• A quadratic on probability — Kalshi, and Robinhood's own commission. Taxes uncertainty. Cheap on near-certainties, expensive at even odds.
• A flat fee per contract — ForecastEx, which does not bill separately at all but builds a cent into the spread, so Yes and No sum to $1.01 rather than $1.00. Taxes every trade equally per contract, which as a share of stake means it taxes longshots hardest.
• A percentage of the trade — Polymarket's US venue, at 0.10% taker and nothing for makers. Taxes size, and is neutral to price.
• A percentage of profit — PredictIt at 10% of gross profits, Polymarket's global venue at 2% of net profits on withdrawal. Taxes only winners. A losing trade pays nothing.
• A percentage of the potential payout — FanDuel Predicts, at 2% of what the position could pay. Taxes the size of the claim rather than the size of the stake, which makes longshots dramatically expensive.

Put the identical trade through all five and the spread is not marginal.

A hundred contracts bought at 20 cents that settle at a dollar: $20 at risk, $100 back, $80 of profit. Polymarket's US venue charges 2 cents. PredictIt charges $8.00 — 400 times as much for the same position and the same outcome. Kalshi charges $1.12, ForecastEx a flat $1.00, Robinhood all-in $2.00, FanDuel $2.00 on the payout, and Polymarket's global entity $1.60 out of the profit.

And the ordering inverts when the trade loses. The profit-based venues charge nothing on a loser; the volume-based ones charge exactly what they charged the winner. A trader who is right slightly more than half the time and a trader who is wrong slightly more than half the time face completely different cost structures at the same nominal rate — which is a design choice about what kind of customer the venue wants, dressed as a fee schedule.

ForecastEx adds a wrinkle worth naming because it points the other way entirely: it pays interest on posted collateral, at a rate tracking half a point below Fed Funds. Since both sides of every contract have fully funded their maximum loss, that collateral is real money sitting still, and paying interest on it partially refunds the cost of holding a position to settlement. It is the only venue in the list that gives some of the economics back.

8. What Robinhood actually books, and why it sits above its own ceiling

Now the number that started this.

In the quarter ended 30 June 2026 Robinhood reported $156 million of event-contract revenue on 13.6 billion contracts traded. That is 1.1471 cents a contract. Its own published commission cannot exceed 1.00 cent per contract per trade. Thirteen point six billion contracts at the cap is $136 million. The reported figure is $20.0 million higher.

One quarter is an anomaly. Four is a structure.

Robinhood broke the line out across six quarters in its second-quarter supplement, and the volume figures have been in the monthly and quarterly operating data all along. Where both exist, the arithmetic gives the same answer every time: 1.1739 cents in the September 2025 quarter, 1.1882 in December, 1.1818 in March and 1.1471 in June. The overage against a 1.00-cent ceiling runs 14.8%, 15.8%, 15.4% and 12.8% of the line. It has never once been below the cap.

There are two ordinary explanations. Neither is disclosed, and the choice between them is the most consequential unanswered question about this business.

One: the count and the charge are on different bases. Commissions apply to opening and closing trades. If "event contracts traded" counts a position once while two commissions are charged on a round trip, the excess is arithmetic, not income. Running it backwards: if every commission were the full 1.00-cent cap, the four quarters imply that at least 17.4%, 18.8%, 18.2% and 14.7% of counted contracts carried a second commission. Those are floors, not estimates — every contract priced outside the capped band pays less than a cent, which pushes the true early-close rate higher. Read that way, the figure is a genuine and otherwise unpublished measure of how retail behaves in these markets: the large majority of contracts are held to settlement rather than traded out.

Two: money is flowing back from the venues. Rebates, fee sharing or volume payments from exchanges would land in this same line. Our own reading of the business has carried "exchange and clearing rebates" as one of its revenue streams for months on the strength of nothing more than that the line is bigger than the commission schedule allows.

The difference matters because it decides what 8 September was. If the excess is a counting convention, taking equity in a venue is Robinhood entering the exchange layer for the first time. If it is venue payments, Robinhood has been earning there for at least a year and the equity stake formalises an arrangement that already exists. One line in one filing settles it, and none has been printed.

9. Twenty-two years of one licence, and what Robinhood bought into

The venue Robinhood is taking equity in is the oldest thing in the story.

North American Derivatives Exchange, Inc. holds a CFTC designation as a contract market that traces to 2004, when the entity was HedgeStreet — the first attempt to sell regulated event contracts to American retail, roughly fifteen years before the market existed. HedgeStreet failed. The licence did not. IG Group bought it and renamed it Nadex, which spent a decade as a small binary-options venue. In March 2022 Crypto.com's parent, Foris DAX Markets, acquired it. On 3 February 2026 it was rebranded OG, and by September it had been spun out as OG.com, with a standalone $5 billion mark taken from inside Citadel Securities' investment in Crypto.com Group at a $20 billion valuation.

Held against that, Kalshi is young — CFTC-designated in 2020, first contracts in 2021 — and Polymarket younger still. The single most valuable asset in US prediction markets is a designation issued in 2004 to a company that went bankrupt, and there are only a handful of them.

What Robinhood announced on 8 September is precise and narrow, and worth stating without embellishment:

• It began routing a selection of football event contracts to OG.com, in phases to eligible US customers, from that date. Kalshi, ForecastEx and Rothera all continue; the routing is additive, not exclusive.
• It will hold "initial equity stakes in Crypto.com and OG.com following the latter's spin-off," priced in line with the Citadel Securities investment. The size is not disclosed. Robinhood's own cautionary language refers to "the equity interests we may earn thereunder" — may earn, and thereunder, meaning under the commercial agreement. No current report was filed, which is what you would expect if the amount were immaterial.
• The $5 billion is a valuation embedded in a round that had already happened, disclosed now. It is not a fresh raise and not a price anyone paid this week.

Anyone putting a dollar figure on the stake is inventing one, and this piece does not. What can be said is what the fee curves say: the layer Robinhood is buying into is the layer that earns more per contract across the two thirds of the price range where the volume is.

10. Implications

In descending order of how confident we are:

1. The cap, not the formula, is Robinhood's fee schedule. A probability-weighted commission that is flat across 77.5 points of a 100-point range is a penny with a curve painted on it. Any analysis that models revenue per contract off p(1 − p) is modelling the tails of the distribution and nothing else.
2. The exchange layer has the better economics on contested markets, by 75%. 1.75 cents against 1.00 at even odds, across a 65.5-point band. This is a structural fact about the two published schedules, not a forecast, and it explains vertical integration into venues without reference to any stake size.
3. Revenue per contract has exceeded the commission ceiling in every quarter where both numbers exist. Twelve to sixteen per cent, four quarters running, $20.0 million in the last one. It is either a round-trip counting convention that reveals how retail actually trades, or income from the venues. Robinhood has never split it.
4. Fee shape decides the customer. Volume-based fees charge winners and losers identically; profit-based fees charge only winners. The same 20-cent position costs 2 cents in one venue and $8.00 in another. Competition in this market will be fought on the shape as much as on the rate.
5. As a share of money at risk, these are not cheap products. Eight per cent of stake to round-trip a coin flip through a broker, and up to 17% one-way on a penny longshot. The fee schedules that look gentlest per contract are the harshest per dollar committed.
6. The scarce asset is a 2004 licence. Designation as a contract market, not technology and not an app, is what OG.com carries and what a $5 billion mark is attached to.

What would settle the open questions is specific and dated. August operating data is the first clean month after a World Cup that ran 11 June to 19 July and contributed more than 5 billion contracts; the "more than 30 billion through August" figure Robinhood published on 8 September bounds August only at more than 1.5 billion, which establishes nothing. The September-quarter filing is the first that could size the stakes, and the first in which OG.com volume appears. Any split of commission revenue against venue payments decides section 8. And revenue per contract in the September quarter is the single number to watch: if it rises while OG.com ramps, the venue economics are reaching Robinhood; if it keeps drifting down, the stake is decoration.

Primary sources worth reading

• Robinhood's event contracts fee page — the commission formula, the k values, the round-up rule, the $0.01 cap, the "up to $0.01 per contract" exchange fee, the both-sides rule, the named venues and the 1 June 2026 change from a flat rate.
• Kalshi's fee schedule as filed with the CFTC — the 0.07 × C × P × (1−P) formula, the taker-only rule, the printed fee table that peaks at $1.75 per hundred contracts, and the halved 0.035 coefficient for the S&P 500 and Nasdaq-100 families.
• Robinhood's explainer on how prediction markets are structured — the company's own definition of exchange, broker/FCM and market maker, and a useful demonstration of what it leaves out.
• Robinhood's 8 September 2026 announcement of the OG.com partnership — the routing, the "initial equity stakes" language, the Citadel Securities pricing reference, the $5 billion standalone mark and the named regulated entities.
• Robinhood's second-quarter 2026 results exhibit — $156 million of event-contract revenue, 13.6 billion contracts and the transaction-revenue split.
• Our capture of that quarter and its call, and the earlier deep dive on what these contracts are and who sells them, which covers the history and the competitive field this piece takes as read.

Robinhood's ticker page is at /stocks/hood/.

The commission formula, the 5% and 10% constants, the round-up rule, the $0.01 commission cap, the "up to $0.01 per contract" exchange fee, the both-sides rule and the 1 June 2026 move from a flat rate are Robinhood's own published customer terms; the layer definitions are from its explainer. Kalshi's formula, its taker-only rule, its printed fee table and the halved index coefficient are from the schedule it filed with the CFTC. The routing start, the equity language and the $5 billion and $20 billion valuation references are from the 8 September announcements. The $156 million on 13.6 billion contracts is as Robinhood reported for the June quarter; the earlier quarterly revenue figures are from the six-quarter breakout in its second-quarter supplement and the volumes from its quarterly and monthly operating data. ForecastEx's spread and collateral interest, Polymarket's, PredictIt's and FanDuel's schedules are those venues' published terms as compiled by third parties, not filings we opened. Everything computed here is ours: the cap-binding bands, the crossover prices, the share-of-stake figures, the five-venue comparison, the revenue per contract in each quarter and the implied second-leg floors. Nadex's lineage from HedgeStreet is the CFTC's designation record.