Institutional AI Dominance Surges 800% as Retail Fades in Prediction Markets

Woofun AI reports that the prediction market landscape is undergoing a structural inversion, shifting from retail-driven volatility to institutional precision as AI agents and proprietary trading firms engage in a high-stakes battle for pricing power. This transformation was starkly evident during the Federal Reserve’s interest rate decision window, where quantitative strategies replaced speculative retail bets. The analysis, compiled by Saoirse and written by Gino Matos for Foresight News, highlights how platforms like Kalshi and Polymarket are becoming the new frontier for institutional liquidity, driven by the need to price uncertainty in real-time. The era of casual retail participation is fading, replaced by a sophisticated ecosystem where market makers, quantitative firms, and AI agents compete to capture the marginal gains in event-driven contracts.

The immediate catalyst for this institutional migration was the Federal Reserve’s policy meeting scheduled from July 28th to July 29th. Traders across bonds, forex, cryptocurrencies, and event contracts positioned themselves to bet on the central bank’s decision outcomes. A Reuters survey conducted on July 21st of 104 economists revealed a near-universal expectation that the Federal Reserve would maintain the interest rate range at 3.50%–3.75%. This consensus was mirrored in the prediction markets, where Kalshi’s July contracts priced an 87% probability for this specific outcome. With trading volume reaching approximately $29.7 million, the market efficiently priced the consensus view.

However, the remaining 13% probability represented a significant tail risk that required counterparties to quote prices, highlighting the critical role of liquidity providers in ensuring market depth even for low-probability events.

To address these liquidity needs, the market has seen the rise of sophisticated counterparties, including market makers, quantitative firms, proprietary trading teams, and AI agents. These entities operate continuously, monitoring prices around the clock and comparing similar contracts across multiple platforms to update the probability of events occurring in real-time. Major financial institutions are actively testing event contracts, while brokers work to onboard liquidity providers to thicken the order book.

Proprietary trading firms are increasingly using settled contracts as a benchmark to select traders, whether human or algorithmic. Those who can price uncertainty more accurately than the market average receive special attention, as their ability to provide liquidity in tail regions is essential for market functionality. The combined efforts of these parties accelerate price discovery, but trading advantages are concentrating among institutions with the fastest infrastructure.

The scale of this institutional adoption is reflected in the surging trading volumes.

Woofun AI data shows that the combined monthly trading volume of Kalshi and Polymarket peaked at $13.7 billion in June, with July volumes already exceeding $11 billion. Kalshi reported that its annual trading volume reached $178 billion, a figure that tripled within six months. More notably, institutional trading volume on the platform increased by 800%, signaling a decisive shift in market composition. To support this growth, major financial players have established access channels: Clear Street helps institutional clients connect to Kalshi, Marex links Kalshi and Polymarket, and Jump Trading assists institutions in directly participating in event market trading.

Additionally, firms like AQR, Susquehanna, and OKX have published job listings for professional prediction market trading positions, underscoring the growing demand for specialized talent in this emerging asset class.

Beyond speculative trading, corporate finance departments are experimenting with event contracts to hedge against tariff risks and regulatory policy exposures. For such hedging strategies to be effective, the market must have counterparties capable of continuously providing large-scale reverse position quotes. This requires suppliers to be willing to quote bid-ask prices, compare similar contracts across platforms, and immediately adjust pricing when significant deviations occur. The ability to provide this liquidity is not just a service but a competitive advantage, as it allows corporations to manage their risk exposures more precisely.

However, the requirement for continuous liquidity provision means that only well-capitalized institutions with robust infrastructure can participate effectively, further marginalizing retail traders who lack the resources to maintain such positions.

Louis Régis, founder of on-chain proprietary trading firm Propr and former Credit Suisse quantitative trader, argues that event contracts offer stricter criteria for selecting traders compared to traditional financial markets. These contracts clearly reflect a trader’s judgment ability, with well-defined and controllable risk boundaries. Since a contract ultimately settles based on a clear outcome, fund providers can directly assess whether a trader can consistently provide probability pricing better than the market consensus. This method of evaluation is far more pure than simply looking at directional trading profit and loss records, which are often influenced by market trends and margin fluctuations. By focusing on the accuracy of probability pricing, institutions can identify traders with genuine skill rather than those who have simply benefited from favorable market conditions.

To quantify this skill, Foresight Arena conducted benchmark calculations showing that confirming a stable 2-percentage-point trading advantage with reasonable statistical confidence requires approximately 350 settled binary prediction contracts. Verifying a 1-percentage-point advantage requires a sample size roughly four times larger. Achieving short-term profits through a few contracts related to Federal Reserve decisions or elections may simply reflect picking favorable trading targets, coincidental position correlations, or luck, rather than indicating long-term capability.

Propr plans to expand this evaluation system to Polymarket, allowing traders and AI agents that pass the assessments to receive up to $100,000 in trading credit per account, with a total limit of $300,000 across multiple accounts. Profit sharing can reach 80%, incentivizing high-performance participants. The company treats each trade as a valid signal, with some signals replicated to the live trading platform as A-book positions, while the rest run internally in simulation as B-book positions. Currently, Propr only sends about 5% of its trading signals to the real market, with the remaining signals kept for internal simulation to accumulate sufficient data and manage funds carefully. Earnings in both A-book and B-book are ultimately settled in USDC on-chain.

Despite the theoretical suitability of prediction markets for AI agents, practical challenges remain. Louis Régis believes that the standardized structure of contracts, real-time price monitoring, and fixed settlement rules create an ideal environment for AI.

However, a benchmark test conducted by Prediction Arena from January 12th to March 9th revealed significant difficulties. Six cutting-edge AI models were each given $10,000 to trade independently on Kalshi and Polymarket. The results showed that the models lost between 16%–30.8% on Kalshi. On Polymarket, the average loss was smaller, but they still recorded negative returns, with an average pullback of 1.1%. Another research paper points out that converting prediction accuracy into stable profits requires a reasonable betting strategy combined with sufficient liquidity to support the strategy. This highlights the gap between theoretical AI capabilities and the practical realities of market execution, where liquidity constraints and transaction costs can erode potential gains.

Looking ahead, the future of prediction markets depends on the balance between liquidity and concentration. Under an optimistic scenario, traders, market makers, and AI agents backed by funds will bring in ample real-money capital, narrowing the bid-ask spread and thickening the order book. A research paper from January 2026 analyzed similar contracts on Polymarket, Kalshi, PredictIt, and Robinhood, finding that when liquidity is high, Polymarket often dominates price discovery. Large-scale one-way order flows determine which platform adjusts prices first, and more real-money capital is expected to widen the lead of leading platforms.

However, under a pessimistic scenario, trading advantages will concentrate in the hands of a few institutions with top-tier infrastructure, leaving ordinary retail traders to lose to more informed counterparts. When the market struggles to price events, liquidity will shrink rapidly, exacerbating the divide between institutional and retail participants.

The battle for pricing power is intensifying as the market prepares for upcoming economic data releases. The U.S. Bureau of Economic Analysis will release preliminary GDP estimates on July 30th, and the July employment report will be published on August 7th. These events, along with CPI and NFP data, will trigger concentrated repricing across all contracts. The real competition lies in the tail regions, those probability ranges outside the consensus view. Whoever can accurately price these data releases first or quickly adjust outdated prices will gain control over trading order flows.

Louis Régis notes that while he is confident in the direction of development, the ultimate scale remains unpredictable. Even if a proprietary trading firm expands rapidly, its trading volume remains limited compared to a market with monthly volumes in the hundreds of billions. The ability to consistently and accurately price such data trends is key for a trader or model to receive real-money capital support from proprietary trading firms, marking a new era where precision and speed define success in prediction markets.