A comprehensive analysis of the three prediction market paths: Why are investors all rushing in?

1. Introduction

Today, I was chatting with a friend, and he brought up a very interesting question: “If Crypto × Fintech truly creates gains over the next 10 years, who will be the biggest winners? In other words, which companies will be part of the ‘Mag7’ in this space?”

Revolut, Robinhood, Coinbase, Stripe… these are obviously the first names that come to mind. Over the past decade, they have proven their ability to redo certain parts of traditional finance.

But as we talked, I suddenly realized: my previous way of thinking had a fundamental misconception. I kept asking—“What parts of traditional finance haven’t been rebuilt yet?” Essentially, this logic is still about finding gaps on an old map.

But the real question should be: which companies are not just digitizing old finance, but creating an entirely new financial market?

Within this framework, one name is almost always assumed—Polymarket. Not because it has risen rapidly, nor because it’s frequently cited by the media recently, but because what it does is completely different: it doesn’t reform banks or payments; it redefines “events” themselves. It turns events into assets, and probabilities into prices.

Coincidentally, prediction markets have surged again over the past year. So naturally, we ask another more important question: why are prediction markets becoming one of the “most worth studying tracks” in 2024–2025? And in this wave of revival, what paths do Polymarket, Kalshi, and Opinion each represent?

2. Why Will Prediction Markets Heat Up Again in 2024–2025?

Explaining this wave of enthusiasm solely through “US elections” or “celebrity events” is insufficient. There have been numerous hot topics in recent years, but prediction markets haven’t surged like this before. This time is different. Underlying this are several deeper structural changes.

1) AI makes “probability” relevant again

In the past, large models provided judgment sentences; now, more and more scenarios output probabilities. Predictions like CPI, interest rate cuts, corporate events, policy directions—all these involve probabilities. Once probabilities emerge, a demand naturally arises: probabilities need prices, prices need markets. So, prediction markets are now integrated into AI workflows, rather than being mere “speculative tools.” The influence of this shift will far surpass current discussions.

2) Media starts using it as a “real-time sentiment indicator”

Over the past year, a noticeable change: more mainstream media outlets are citing Polymarket. Why? Because it’s faster than polls and more transparent than expert judgments. Media coverage → user growth → market depth expansion—this is a simple but effective flywheel. Previously, prediction markets weren’t big enough because they didn’t enter mainstream narratives; now, they have.

3) High event density but lack of “corresponding tools”

The world in 2024–2025 features unprecedented information density: elections, geopolitics, macro policies, tech regulation, corporate events (especially AI-related). The problem is: these events have huge impacts but lack corresponding financial tools for trading.

You can buy gold, US stocks, bonds—but not: “the probability of a Fed rate cut in December.” “Will a certain CEO resign this quarter?” “Will a specific regulation be implemented?” Prediction markets fill exactly this gap. Essentially, they create a new asset type: event assets.

4) Slight but significant regulatory attitude shifts

CFTC once penalized Polymarket, but at the same time, Kalshi obtained a CFTC license. This sends a very real signal: some prediction markets can be permitted; some can operate within compliance; gray areas are being delineated. For institutional investors, “reducing uncertainty” is a growth signal.

5) User composition has changed

In the past: mainly entertainment-oriented users, with dispersed liquidity; products resembled “information apps.” This wave is clearly different: more institutional accounts, professional traders doing indicator predictions, hedge funds starting to use it for hedging, AI companies referencing it. As user structure shifts from “spectators” to “traders,” the quality of the market will undergo a qualitative change.

Summary

Prediction markets didn’t suddenly catch fire. They are the result of multiple forces: AI demand, media citations, macro environment pushes, user structure shifts, and gradually clarified regulatory boundaries—all contributing to a structural uplift. This wave isn’t driven by short-term events. It’s more like prediction markets gaining their first “use case in the era.”

3. Three Completely Different Paths: Polymarket, Kalshi, Opinion

These three companies all operate prediction markets, but their routes are entirely different. They solve different problems and target different user groups. Viewing them together reveals a potential layered structure of this track’s future.

1) Polymarket: Turning Events into Assets

Polymarket’s approach is straightforward: turn events into assets, turn probabilities into prices. It’s not a traditional “prediction tool,” more like a real-time event price display. The higher the social attention, event density, and media citations, the faster its market response. Its low understanding threshold and strong emotional influence are reasons for its rapid growth. Its advantage is speed; its challenge is regulation. To summarize: the entry point for event assetization.

2) Kalshi: A Regulated Event Derivatives Exchange

Kalshi represents a more financialized route. It offers event contracts that can be defined and captured within regulation: CPI, unemployment rate, yields, FOMC, etc. It attracts a different user base: macro traders, hedge funds, quant teams. This results in more stable, scalable trading structures than Polymarket.

The political markets you see on Kalshi do not mean it’s the same kind of product as Polymarket—politics is just one category of regulated events and doesn’t define its growth logic. To summarize: an event derivatives exchange, the financial infrastructure of prediction markets.

3) Opinion Labs: The Model Consensus Layer in the AI Era

Opinion follows a third route: it’s not aimed at retail traders nor institutional traders. It seeks to build a “probability consensus layer” for AI models: aggregating, citing, and ultimately market-pricing probabilities output by different models. Its target isn’t humans but models. It’s not about “letting users bet,” but “giving models a readable, tradable probability interface.”

This path has a longer time horizon and is more forward-looking. Compared to the other two, Opinion is at a much earlier development stage.

It already has a trading interface (opinion.trade), but restricts access in regions like the US and China, leading to inconsistent user experiences across networks. Public info is limited; its main external contact remains Twitter. The underlying infrastructure is rapidly iterating; branding and official websites are lower priorities.

This isn’t an “immature website experience,” but typical of early-stage foundational projects: first build the core mechanisms, then gradually move toward external stability.

To summarize: Opinion has a product but remains in very early stages, more like a foundational piece for future AI ecosystems rather than a player in current user-scale competition.

Polymarket, Kalshi, and Opinion may all seem to be doing prediction markets, but their directions, product structures, compliance paths, and future positioning are completely different: Polymarket captures “attention and sentiment.” Kalshi captures “risk and pricing models.” Opinion captures “AI’s understanding of the future.”

They correspond to the three levels of prediction markets: mass audience, financial infrastructure, and model layer. Because these three paths coexist, this wave of prediction market growth isn’t like the past—where one product suddenly became popular—but the market is gradually taking shape.

4. My Observation on This Track: AI Creates Noise, Web3 Differentiates Noise

I don’t want to make a definitive “what the future holds” statement about prediction markets, because I haven’t deeply studied this track. But over the past year, across different projects and product forms, I’ve repeatedly seen one thing: the integration of AI and Web3 is happening faster than we think, and in a very clear direction.

AI’s strength lies in “generation”—generating text, judgments, predictions. But as its output increases exponentially, a new problem becomes more evident: AI is creating noise. Judgments, explanations, probabilities, inferences—each is expanding rapidly. More information → more noise → higher costs.

Web3’s role, right after the noise, is to distinguish it: Web3 provides not “content,” but: immutable, settleable, verifiable, incentive-aligned, price-formation mechanisms.

The combination of these two will gradually become natural in financial markets:

• AI responsible for generating views of the future;
• Web3 responsible for embedding these views into markets, allowing them to be tested through prices, time, and incentives.

Prediction markets are just a very intuitive example. They turn “AI-generated probabilities” into “financially usable prices.” From this perspective, they’re more like an interface than an application. I’m not sure what this track will ultimately look like, but what I do see is: AI is making the future more blurred, Web3 is making it more verifiable. And in the financial realm, these two will naturally complement each other.