The Oracle Layer
Wall Street is not just watching prediction markets. It is wiring them into its infrastructure — and the wiring changes what the instruments measure.
On January 15, 2026, Goldman Sachs CEO David Solomon told analysts that prediction markets were “super interesting.” He disclosed multi-hour strategy sessions with the leadership of both Kalshi and Polymarket in the opening weeks of the year. He called them “event contract activities” — derivatives language, not gambling language. He noted that CFTC-regulated event contracts “increasingly resemble traditional derivatives” and appear “similar to financial instruments Goldman already trades or clears.”
Goldman is not early. It is late. By the time Solomon spoke, Susquehanna International Group had already become the first market maker on Kalshi, running cross-venue arbitrage algorithms that trade the spread when a Fed cut contract prices at 65 cents on one platform and 68 cents on another. DRW Trading was building a dedicated prediction markets desk, advertising base salaries up to $200,000. Intercontinental Exchange — parent of the New York Stock Exchange — had invested up to $2 billion in Polymarket at an $8 billion pre-money valuation, with a mandate to become Polymarket's global data distributor.
The institutional build-out is not speculative capital chasing a trend. It is operational infrastructure: desks, algorithms, clearing relationships, data pipelines. The question is what happens when that infrastructure creates a feedback loop.
The Loop
The standard narrative runs one direction. Prediction markets produce accurate signals. Institutions recognize the accuracy. Integration follows. The intellectual lineage runs from Hayek through Tetlock through Vitalik. Build the oracle. Let it speak.
The mechanism that is actually operating runs in a circle.
Institutional capital enters prediction markets. Susquehanna's algorithms and DRW's traders narrow bid-ask spreads, reduce noise, and improve price efficiency. More efficient prices attract the next tier of institutional participants. An NBER working paper finds that Kalshi's forecast errors are “almost the same” as Bloomberg consensus — academic validation that these prices are institutional-grade. That validation justifies Goldman's strategy sessions. Goldman's participation deepens liquidity further. Deeper liquidity produces tighter spreads and more accurate prices. More accurate prices justify broader integration into the financial data infrastructure — the six media partnerships signed in 62 days, the Bloomberg Terminal integration, the Google Finance rollout. (These distribution deals are the subject of “The Feedback Loop.”)
Broader integration drives visibility to hundreds of millions of users. Visibility drives participation. Participation deepens liquidity. The cycle compounds.
But the loop has a second face. A probability displayed on CNBC's ticker during a discussion of the March FOMC does not passively describe expectations about Fed policy. It enters the information set of every trader, analyst, and policymaker watching. It shapes the conversation about what the Fed will do. It becomes an input to the decision it claims to measure. The oracle stops being passive the moment it achieves the authority that institutional adoption confers.
The Volume Beneath the Signal
Kalshi cleared $43.1 billion in notional volume in 2025, a 2,100 percent increase year-over-year. The headline is real. The composition complicates it.
Ninety-one percent of that volume is sports. Fee revenue tells the same story: $234.6 million of $263.5 million in total fees came from sports contracts. Non-sports fee revenue — the entire commercial footprint of “information finance” — was less than $30 million.
Polymarket's reported volume ranges from $9 billion to $33.4 billion depending on methodology. A December 2025 Paradigm analysis documented significant double-counting in Polymarket's on-chain reporting: a sale of YES tokens for $4.13 gets recorded as $8.26 because both maker and taker events are counted. Adjusted figures cluster around $9 to $16 billion.
The macro contracts that Wall Street cares about are a thin layer riding atop a sportsbook. Fed rate decision contracts for the March 2026 FOMC generated over $120 million in volume. The Kevin Warsh Fed Chair nomination moved $368 million across platforms. December 2025 alone saw $394 million in Fed-related volume on Kalshi. These are real numbers. They are also a fraction of the headline figure. The platform that Goldman Sachs is evaluating as an institutional data feed derives less than 9 percent of its volume from non-sports markets.
This does not invalidate the accuracy claims. The NBER paper and Kalshi's internal research pertain specifically to macro contracts. But the liquidity underpinning the oracle layer is thinner than the integration deals imply. The institutional infrastructure is being built atop a narrow base.
The Accuracy Evidence
The accuracy case rests on two studies. One is independent. One is not.
The independent study is NBER Working Paper No. 34702, authored by Anthony Diercks, Jared Dean Katz, and Jonathan Wright. One author is from the Federal Reserve. The paper found that Kalshi's modal forecast maintained a perfect record on Fed rate decisions from 2022 through June — a “statistically significant improvement over the Fed funds futures forecast.” The critical case was September 2024, when Kalshi correctly weighted a 50-basis-point cut while traditional forecasters remained divided. It was the only FOMC decision during the study period where traditional forecasters got it wrong. The paper has not been peer-reviewed.
The second study is Kalshi's own. “Crisis Alpha,” published by Kalshi Research in December 2025, claimed market-based CPI estimates recorded a 40 percent lower mean absolute error than Wall Street consensus over 25 months. An 85 percent win rate from one week out. A 67 percent outperformance during “shock” events. The authors acknowledged a small sample of shocks. This is a study from the platform's in-house research arm. It is not independent evidence.
The divergence between prediction markets and traditional tools is currently live. On the March 2026 FOMC, CME FedWatch prices a 48 percent probability of a rate cut. Kalshi prices 62 percent. Polymarket prices 71 percent. The 23-point gap between FedWatch and Polymarket is the widest sustained divergence observed in the prediction market era. Somebody is wrong. The resolution will test whether the oracle layer's authority is earned or premature.
The Institutional Build-Out
The capital entering prediction market infrastructure follows a consistent pattern: not speculative positions, but permanent operational commitments.
ICE's $2 billion investment in Polymarket is not a trade. It makes the owner of the NYSE a global distributor of prediction market data — embedding event-driven probabilities into the same distribution network that carries equity prices and fixed-income benchmarks.
Susquehanna, Kalshi's first market maker, is recruiting traders for cross-venue arbitrage. DRW is building a dedicated desk. The Coalition for Prediction Markets — Kalshi, Robinhood, Interactive Brokers — formed in December 2025 to lobby for a regulatory framework that preserves federal jurisdiction. Kalshi raised $1 billion in its Series E at an $11 billion valuation, led by Paradigm with Sequoia, Andreessen Horowitz, and ARK Invest.
Robinhood, which drives more than half of Kalshi's volume, closed its acquisition of MIAXdx on January 20, 2026, and rebranded it as Rothera — its own CFTC-licensed exchange, expected to launch in Q2 2026. (The competitive dynamics of that migration are the subject of “The Convergence” and “The Wallet Wars.”)
The regulatory environment is cooperating. CFTC Chairman Michael Selig, in his first public speech on January 29, 2026, withdrew the proposed ban on political and sports event contracts and directed staff to draft new rulemaking with “clear standards.” He reframed prediction markets as tools for “price discovery” and “information aggregation.”
Every element reinforces the loop. Regulatory legitimacy broadens access. Broader access deepens liquidity. Deeper liquidity improves accuracy. Improved accuracy justifies further institutional adoption. The infrastructure is being built to compound.
What the Oracle Produces
In eighteen months, prediction markets have built a parallel data infrastructure for macroeconomic expectations, attracted institutional capital at a scale that makes the markets more liquid, and produced accuracy results that are — by the available evidence — roughly on par with the existing consensus machinery.
The evidence is real but provisional. The NBER paper is not peer-reviewed. The Kalshi study is promotional. The 91 percent sports composition means the macro signal is produced by a thin layer of volume. The 23-point spread between CME FedWatch and Polymarket on the March FOMC will resolve, and the resolution will matter.
The structural tension is not between accuracy and inaccuracy. It is between measurement and intervention. Institutional adoption makes the oracle more accurate, which makes it more trusted, which makes it more powerful, which makes it less passive. Goldman's strategy sessions, DRW's trading desk, ICE's distribution deal — each deepens the oracle's authority. Each degree of authority increases the oracle's influence on the outcomes it measures.
No one building the institutional layer has modeled this dynamic, measured it, or disclosed it. The loop is running. It does not have a governor.
Dawson Smith writes Reflexivity, a newsletter on prediction markets as reflexive systems.