For thirty-nine days this summer, the fastest-growing venue in American sports betting was not a sportsbook.
Kalshi, a federally regulated futures exchange that lets users trade contracts on real-world outcomes, processed $31.5 billion in notional trading volume in June 2026 — an 87.4% jump from May’s $16.8 billion. Polymarket, its crypto-native rival, hit a platform record of $10.8 billion. Combined with Polymarket’s regulated US arm, the sector moved $44.8 billion in a single month, a 75% increase over May.
To put that in context: combined monthly volume across the two platforms was under $5 billion in September 2025, according to a Pew Research Center analysis of data from The Block. June’s total was more than triple the roughly $14 billion average monthly handle of every legal US sportsbook in 2025.
Then the tournament ended, and the number fell off a cliff.
The collapse
Spain beat Argentina 1-0 on July 19. The trading data since then is unambiguous.
Kalshi’s weekly volume peaked at roughly $9 billion in the week of July 5, then fell nearly 55% to about $4 billion by July 19. Polymarket dropped harder — close to 70%, from a peak near $2.3 billion to $740 million. Combined open interest across both platforms fell nearly 20% from its early-July high of about $2 billion.
Sports accounted for roughly 80% of total volume across both platforms during the tournament. When the sports calendar dried up, so did the volume.
Annualised, Kalshi’s July 19 week runs to roughly $16 billion a month — essentially back to its pre-tournament May baseline. On the simplest reading, the World Cup was a spike, not a step change.
The complication
Except the simplest reading leaves out the most interesting number in the dataset.
H2 Gambling Capital estimates that prediction-market activity accounted for around 27% of all legal US sports-betting volume during the World Cup, up from 9% at the beginning of the year. That is not a volume spike. That is share capture.
The app-download data points the same way. Apptopia recorded Kalshi’s daily active users growing 36% between June 15 and June 30 — while DraftKings and FanDuel lost between 36% and 41% off their early-tournament peak in the same window. Kalshi and Polymarket together accounted for 78.5% of June installs across a six-app comparison group.
That pattern reverses the usual logic of a major tournament, where established sportsbooks absorb the casual influx and the challengers get scraps.
So both things are true. Volume reverted almost exactly as a bubble would. Market share moved structurally in a way bubbles don’t. The post-tournament autumn — an NFL season, no World Cup — is the first clean test of which signal was real.
Prediction markets took 27% of legal US sports-betting volume during the World Cup, up from 9% at the start of the year.
Kalshi added daily active users through late June while the major sportsbooks shed them.
What they actually bought
The mechanics matter for interpreting any of these figures. A prediction market contract is priced between $0.01 and $0.99 and pays out $1.00 if the outcome hits, nothing if it doesn’t. The price is therefore the market’s implied probability: a 60-cent contract means the market thinks the event is 60% likely.
Kalshi’s World Cup winner market alone drew more than $832 million, with roughly 35% of participants backing France. Ticker Tracker put total Kalshi World Cup volume at $13.8 billion, with parlays exceeding $19.8 billion. Polymarket’s winner market drew over $4 billion, and roughly $20 billion across all World Cup markets combining its international and US exchanges.
Spain won. France did not reach the final.
Kalshi’s weekly volume peaked at roughly $9bn in the week of July 5, then fell 55% by July 19. Sports accounted for around 80% of tournament volume.
A caution about the headline numbers
Anyone comparing these two platforms directly should know that the industry’s standard metric is not standardised.
Kalshi counts notional volume as contracts traded multiplied by their $1 face value. Polymarket uses taker notional volume — contracts multiplied by the price actually paid at the time of trade. A contract trading at 10 cents therefore registers as $1.00 of Kalshi volume and $0.10 of Polymarket volume. The two figures measure related but structurally different things.
Parlays compound the distortion. Since the World Cup opened, parlays made up almost half of Kalshi’s total volume, and parlays inflate notional volume by construction. Polymarket US has been at an accounting disadvantage without a comparable product, and expects to roll out custom parlays shortly.
There is also a question about who is generating the volume at all. Bloomberg’s analysis of Dune data found that roughly 5% of bot-like wallets generated 75% of Polymarket’s trading volume.
Kalshi’s lead over Polymarket is real. It is narrower than $31.5 billion versus $10.8 billion implies.
The platforms publish free public Prediction Market APIs, which is why prediction market volume is unusually easy to scrutinise — and why nearly every figure in circulation traces back to the same handful of Dune dashboards, with the methodology caveats that implies.
The field got crowded
Eighteen months ago, prediction markets meant Polymarket, Kalshi, and a handful of thinly traded on-chain venues. In September 2025 Kalshi held 62% of global prediction market activity — but a year before that, its share had been 3%. The category was small enough for one platform to go from negligible to dominant in twelve months.
By mid-2026, BeInCrypto’s institutional research screened a starting pool of 34 prediction-market platforms and infrastructure providers, narrowing to a 15-firm longlist spanning regulated exchanges, broker-distributed platforms, on-chain venues and liquidity infrastructure.
The more consequential change is who entered:
- November 2025 — DraftKings acquires Railbird Exchange, launching the DraftKings Predictions app a month later across 38 states
- December 2025 — Polymarket US launches as a CFTC-regulated domestic venue
- January 2026 — Coinbase launches regulated markets via a Kalshi partnership
- February 2026 — Crypto.com’s OG launches, reaching 40-plus states
- March 2026 — Underdog acquires Aristotle Exchange, rebranding as UDX; MEXC launches a zero-fee platform
- June 2026 — Rothera, a Susquehanna–Robinhood joint venture, debuts and takes 7% of US prediction market volume in its first month
FanDuel and Fanatics have launched their own products. Kraken has said it plans to. Interactive Brokers operates ForecastEx, regulated as both a designated contract market and a clearing organisation.
For years the space was led by crypto-native protocols and a couple of regulated pioneers. It is now a venue category that the largest US sportsbooks and brokerages have all entered within nine months.
Concentration remains extreme despite the crowd. Through the World Cup final, Kalshi accounted for roughly 83% of notional trading volume among CFTC-approved exchanges. Behind Polymarket US, only Rothera (4.6%) and Crypto.com-owned Nadex (2.4%) clear 1% share.
Who is actually trading
Not, on the evidence, a broad cross-section of sports fans.
On Kalshi, roughly 3 million of 4 million active users are male. Six in ten are aged 18 to 34.
That age distribution is not incidental — it is partly a regulatory artefact. Kalshi and Polymarket are open to anyone 18 or older, with some state-by-state exceptions. Online sports betting, where legal, is typically restricted to 21 and over. Juice Reel, a wagering-tracking app, found higher uptake of prediction markets in states where sports betting isn’t legal at all.
Prediction markets are arbitraging both a three-year age gap and the state legality map simultaneously. That has drawn attention: NCAA President Charlie Baker wrote to the CFTC asking it to remove college sports from tradeable options until more safeguards are in place, and Truist analysts have suggested college students and teens could be a meaningful driver of the volume surge.
The regulatory basis for the gap is that prediction contracts are classified as financial derivatives rather than bets, placing them under federal CFTC jurisdiction rather than state gambling law. That classification is being tested: Nevada gaming regulators sued Kalshi in February 2026, and the Arizona Attorney General filed suit in March.
The money followed
Capital has moved faster than volume. Kalshi raised over $1 billion at a $22 billion valuation in a round led by Coatue, doubling its December 2025 mark, and secured a licence in March to offer margin trading to institutional clients. Intercontinental Exchange, parent of the NYSE, completed a $2 billion commitment to Polymarket in March, having begun with $1 billion in October 2025.
In the first half of 2026, prediction markets took $1.85 billion of the $7.1 billion raised across crypto’s top ten categories — 26% of the total, making it the most funded category in the sector.
The industry spent heavily to convert World Cup attention into permanent users. Kalshi ran 36 ads during live matches on Fox and Telemundo through the semifinals, roughly 40% of its 2026 national TV spend. Polymarket ran 45. The rest of the US betting industry combined managed 22.
The test
The World Cup was a customer acquisition exercise, and its success will not be measured in June’s $44.8 billion.
It will be measured this autumn, in whether the users who downloaded Kalshi to bet on Spain–Argentina are still trading NFL contracts in November — and whether that 27% share of legal sports-betting volume holds when there is no global tournament propping it up.
The volume data says bubble. The share data says structural shift. Neither has been falsified yet.
Sources: Dune Analytics (queries 5910819 and 5910984, @datadashboards prediction markets dashboard); Pew Research Center analysis of The Block data; The Block; DeFi Rate volume tracker; Bank of America; H2 Gambling Capital; Apptopia; Ticker Tracker; Bloomberg; TRM Labs; CryptoRank; BeInCrypto Institutional Research; iSpot; Sportico; Fortune; CNBC.
Note on methodology: Kalshi and Polymarket calculate notional volume differently, and parlay products inflate the metric. Figures should be treated as directionally accurate rather than directly comparable across platforms.