Who Controls Prediction Markets’ $50 Billion?

More than $50 billion was wagered on Polymarket and Kalshi in June 2026; bets on the FIFA World Cup final alone exceeded $5.69 billion. When that demand reaches traders through brokerages, social feeds and AI assistants, what does either prediction venue actually control?

The contract stopped being the product

Prediction markets began with a narrow proposition: dispersed beliefs could be compressed into a tradable probability. A contract defined an event, buyers and sellers took opposing positions, and a final result converted one side’s judgment into a payout. Because the visible artifact was the yes-or-no box, the box looked like the business.

In 2024, Polymarket still looked primarily like a crypto-based election market that had raised more than $70 million and handled more than $350 million in predictions on the U.S. election. Political attention supplied the participants, the participants supplied the trades, and the resulting prices supplied the publicity. The venue appeared to be an application organized around a popular event.

The 2024 election also exposed that design’s weakness. Polymarket open interest fell 64%, from $510.95 million on November 5 to $184.16 million by November 8. The collapse warns against treating every surge in prediction-market volume as durable liquidity. An event can gather a crowd without creating a market that survives it.

During the election, Polymarket also built the machinery for the phase that followed. More contracts gave traders more reasons to keep capital on the platform; more active traders made market makers more willing to quote; more quotes improved execution; better execution attracted larger and more sophisticated positions. The venue could begin each new contract with an existing base of traders, capital and settlement experience.

This is prediction-market platformization: the contract remains necessary, but the defensible asset moves into the coordination layer around it. The venue no longer merely lists questions. It gathers event risk in one place and makes that risk continuously tradable.

Market makers turn attention into liquidity

A market can have millions of interested observers and still be unusable for a serious trader. Interest produces clicks; liquidity requires someone willing to quote both sides of a contract, update those quotes as information changes and absorb an order without letting the price move beyond recognition. Market makers build usable markets by keeping standing offers available as attention shifts.

Professional gamblers have begun adopting Wall Street-style strategies in prediction markets, where available depth, execution price and the ability to enter or exit can matter as much as being directionally correct. A thin market may display a probability; a deep one lets someone act on it at scale.

Polymarket’s U.S. daily volume rose from roughly $50 million in mid-May to more than $200 million by June 20, while the company said annualized revenue had exceeded $1 billion.

Those wagers do not prove that one venue has won. They show that event risk can support pools large enough to attract professional traders, brokerages, AI assistants and regulators at the same time. Traders want depth; market makers want flow; distributors want prices that remain credible before a large audience; regulators want operators who can explain how they handle orders, disputes and payouts.

Market makers and traders can deepen the pool quickly, but they cannot keep it full on their own. The post-election decline showed that volume tied to a single event can evaporate. Sports tournaments, elections and political crises can produce enormous episodic demand without keeping capital in the market between events, so a venue proves durability only when it carries participants from one contract to the next.

Settlement turns judgment into institutional memory

An order book solves only the middle of a prediction contract. Someone must still define the beginning and decide the end, and the real world resists binary formatting. Speeches contain interruptions. Government actions arrive in stages. Statistics are revised. A source may report an outcome before the authority named in the contract confirms it. The contract can be precise and still collide with an edge case it did not anticipate.

A Polymarket dispute over whether a person had uttered a single word exposed the operating burden hidden inside that collision. Polymarket uses Risk Labs’ Optimistic Oracle to decide roughly 200,000 tough-call bets per month. At that volume, resolution is a production system rather than a back-office exception process.

The oracle does not eliminate judgment; it organizes judgment into a procedure that traders can inspect, contest and eventually accept. Each difficult resolution tests whether the written rules, approved evidence and dispute process produce an outcome participants regard as legitimate enough to trade again. A new entrant can copy a contract description and launch an order book. It cannot immediately copy the accumulated memory of thousands of disputed settlements.

Traders who cannot estimate how a venue will interpret ambiguous language quote less aggressively or avoid the contract. Participants who can influence the reference price or evidence around settlement may create a liquidity hole precisely when the market is most exposed. Market makers price that uncertainty into the spread long before the event concludes.

A venue builds an operating moat—not merely a brand—through definitions, escalation procedures, oracle participation and a record of payouts that survived disagreement. Traders revise prices every second, but they will quote the next ambiguous contract only after watching the venue resolve the last one.

Regulators are dividing one global pool into local markets

Most digital networks grow because every additional participant can join the same system. Prediction markets do not have that luxury. A regulator can prevent local demand from reaching a venue, restrict the contracts it may offer or impose a different legal classification on the entire product. Regulators therefore fragment what might otherwise become one global liquidity pool.

France’s National Gambling Authority blocked access to Polymarket, citing concerns about significant user losses and manipulated trading. Portugal ordered the platform to cease operations after activity tied to its presidential election generated more than €110 million in volume. Spain said it was blocking both Polymarket and Kalshi as a precaution while examining possible gambling-law violations over the following three to four months.

France, Portugal and Spain undercut the idea that scale automatically produces one global liquidity pool. A contract may be visible everywhere while legal participation remains local. An internet service provider enforces the block, a regulator decides whether the contract is gambling, and a compliance team determines which account may place an order from which address.

Venues that satisfy local rules can turn fragmentation to their advantage. Regulators determine where market makers may quote and where customers may supply flow. Polymarket’s pursuit of CFTC and NFA approval for U.S. margin trading illustrates the mechanism. Margin could attract more traders by requiring less capital upfront, but that advantage exists only if regulators permit the venue to offer it.

Once regulators make access scarce, an approved venue can concentrate orders that an unlicensed rival cannot reach. The venue can then use each approved product and reporting relationship to pull more orders into its book.

That does not make regulation a clean moat. France, Portugal and Spain show that permissions can be withdrawn, contested or defined differently across borders. The result is an archipelago in which approved venues deepen local pools while regulators control the bridges between them.

Distribution owners can turn venues into wholesalers

Polymarket and Kalshi built direct destinations because traders once had to visit a prediction market to see its prices and place its contracts. Users no longer need to do either. A probability can travel independently of the screen that produced it, appearing inside a social feed, brokerage or AI answer at the moment someone encounters the underlying event.

Meta was reportedly exploring partnerships with Polymarket and Kalshi for Arena, a prediction application targeting 100 million monthly active “predictors” between 18 and 34. OpenAI struck its first prediction-market deal with Kalshi to place FIFA World Cup market data in ChatGPT search results. Both leading venues have also partnered with social-media accounts presenting themselves as breaking-news reporters.

Meta and OpenAI place market probabilities inside systems where people already discover information and make decisions, turning prediction into an input inside another product much as an exchange executes an order initiated through a brokerage.

When social platforms and assistants carry probabilities to users, they expand the market and redistribute power. The platform controls discovery, presentation and the user relationship; the prediction venue controls liquidity, execution and settlement. If several venues can provide comparable contracts, the distributor can demand better commercial terms or route users toward the provider that best fits its jurisdiction, economics or product design. This is platform gatekeeper leverage: the market may own the probability, but the interface owns the moment when someone acts on it.

A venue with deep liquidity can send prices through interfaces it does not own. Those interfaces can compare providers and switch when another venue offers better terms, legal access or execution.

Routing exposes the limit of liquidity lock-in

By routing some World Cup bets to Rothera while shifting away from Kalshi, Robinhood supplied the clearest counter-evidence to a permanent venue moat. The brokerage owns customer demand and can divide it among providers. If enough large distributors behave this way, they can move customer flow between venues and seed competing pools.

But a distributor cannot reproduce a market merely by rerouting demand. The receiving provider still needs market makers, reliable deposits and withdrawals, contract definitions, settlement credibility and legal access. ADI Predictstreet entered with tiny trading volumes and fund-withdrawal bugs. Its difficulties showed that a new provider can copy the trading screen before it can reliably handle customers’ money.

Local entrants can also redefine the wager. In Japan, Miraima and Poyp have used non-monetary loyalty points convertible into gift cards to build alternatives under strict anti-gambling laws. They do not need to reproduce Polymarket contract for contract. They need to coordinate participation in a form that local law permits.

Robinhood can move customer flow when it controls the interface, while Miraima and Poyp can redesign the product when local law blocks the incumbent model. Market makers and traders, however, will follow only if the receiving venue can execute orders and settle payouts reliably.

Neither the largest venue nor the largest distributor automatically holds the strongest position. A venue gains bargaining power when its pool is difficult to replace; a distributor gains it when its flow can make another pool viable. Each side gains leverage only when it can replace the service the other provides.

No company controls the whole contract

Kalshi can place its prices inside ChatGPT. Meta can consider both venues rather than one. Robinhood can route toward Rothera. France can close an access point. Risk Labs supplies the oracle process behind outcomes that Polymarket lists. The company that appeared to own the whole experience increasingly depends on institutions above, below and beside its order book.

A brokerage choosing a provider for that flow now compares depth, settlement history and legal access, not just contract coverage. It can reroute demand, but only to a venue whose market makers and payout machinery can absorb it.

The contract still ends with “yes” or “no.” The business now lives in everything required to make either answer pay.