Key Takeaways
- The category is usually described as Kalshi against Polymarket, which collapses a five-layer stack into a two-horse race and hides where the real risk sits.
- Six venues now matter on volume: Kalshi, Polymarket International, Polymarket US, Rothera, Underdog and Hyperliquid's HIP-4, with PredictIt and Manifold serving separate research and social niches.
- Custody, resolution and regulatory posture differ more between venues than pricing does, and those differences decide what happens when something goes wrong.
- The infrastructure layer beneath the venues, principally UMA and Gnosis, carries integrity risk out of proportion to its size. UMA's token market capitalisation sat near $95M while it adjudicated a market carrying up to $242M of exposure.
- Liquidity rather than demand is the binding constraint for new venues, according to Stratium founder Wasim Fukase, who reports roughly $400M in cumulative volume on his Hyperliquid-based venue.
- Sportsbooks and brokers entering the category changed the competitive picture more than any new onchain venue did.
How to read a landscape that is no longer a duopoly
Choosing a prediction market, or underwriting one, requires knowing which layer you are exposed to. A venue can be deep and well run and still settle your market on the wrong answer, because resolution happens somewhere else. This guide lists the platforms operating as of the September 2026 summary of our prediction market report, what each one is structurally, and which layer of the stack it occupies. Names are illustrative of who occupies each role rather than an endorsement or a complete list.
The venues, by model and regulatory posture
Platform | Model | Custody | Regulatory posture |
|---|---|---|---|
Kalshi | Central limit order book, regulated event contracts | Custodial, centrally cleared, segregated funds | CFTC designated contract market. Preemption litigation split |
Polymarket International | Off-chain matching, onchain settlement on Polygon | Wallet signing, CTF outcome tokens | Offshore. Banned for US users, restricted in a dozen-plus jurisdictions |
Polymarket US | Regulated US platform via the QCEX licence | Custodial, KYC through regulated intermediaries | CFTC DCM. App waitlist removed May 2026 |
Rothera | Central order book, B2B venue for FCMs and market makers | Custodial, centrally cleared | CFTC DCM and DCO. Robinhood and Susquehanna joint venture |
Hyperliquid HIP-4 | Onchain order book for binary and multi-outcome contracts, unified with perps | Self-custodial on Hyperliquid L1 and HyperEVM | Offshore, unregistered |
PredictIt | Centralised political and event exchange, small position caps | Custodial fiat | CFTC no-action relief, amended by Letter 25-20 in July 2025 |
Manifold | Social forecasting, mostly play money | Web app, no onchain custody | Light. A social game rather than a real-money exchange |
Underdog belongs on that list as the third-largest US venue by regulated notional flow, now being acquired by IG Group for up to $1.3B. Newer onchain venues including Opinion Labs, Limitless and predict.fun are pulling meaningful volume of their own, with Limitless near $1.7B and predict.fun-class venues around $1.5B in April 2026 on Dune-based compilations.
The layer most comparisons skip
The venue is one layer of five. Positions are collateralised somewhere, resolution happens somewhere, and disputes are adjudicated somewhere, and each of those can fail independently of how deep the order book was.
UMA runs the optimistic oracle beneath Polymarket, with staking, bonds, liveness parameters and a Data Verification Mechanism vote. Gnosis supplies the Conditional Tokens and AMM contracts behind a range of onchain forecasting apps. Augur, historically the decentralised prediction market, is now a generalised oracle reboot and is commercially marginal as a venue.
The concentration of risk in that layer is the structural finding. A bond of about $750 was enough to propose a resolution on a market carrying nine figures of exposure, and the token securing the process was worth a fraction of the sum it adjudicated. Kalshi resolves through a regulated rulebook instead, and Hyperliquid's HIP-4 uses a staked and slashable deployer with authorised oracle feeds. Three live designs, none of which has yet been stress-tested by a contested nine-figure market on its own rails.
What an entrant is actually fighting
New venues rarely fail for lack of a product. Wasim Fukase, founder of the Hyperliquid-based venue Stratium, was direct about the constraint when we asked what was holding volume back.
"Right now the liquidity is the issue, and this is being honest about it."
He reports roughly $400M in cumulative volume, with the World Cup final on 19 July driving the largest single spike and five or six of his top ten markets tied to the tournament. Outside those, the recurring volume is daily Bitcoin up-or-down contracts. His model prices builder-code fees on a time basis, which he says comes in materially cheaper than Kalshi's or Polymarket's schedules, an entrant's answer to a distribution problem it cannot solve by renting an existing user base.
Limitless chief executive CJ Hetherington made the adjacent point to Bernstein analysts, noting that the missing momentum behind Hyperliquid's push suggests either that building real liquidity in this category is harder than commonly assumed, or that volume on rival platforms is more inflated than it appears.
The competitive threat to the two leaders now comes from the distribution layer rather than from each other. DraftKings, FanDuel and Fanatics all left the American Gaming Association to launch event contracts, CME helped FanDuel launch its product, and Robinhood moved its own flow onto an exchange it part-owns.
The full market map, including the surveillance, data and structured-products layers, is in the Prediction Markets Report 2026. Fukase's full conversation is in our episode on Hyperliquid and Stratium.
None of this is investment advice.