PRIM3 Brief #13: Prediction Markets Became an Asset Class — and the Trade Isn't the Front-End
Prediction markets stopped being a political-betting sideshow sometime in the last year. The tell isn't the headline volume. It's who's now on the other side of the trade.
Combined platform volume scaled to roughly $21 billion a month in 2026, according to TRM Labs, with Polymarket and Kalshi taking most of it. Kalshi's institutional trading volume grew about 800% over six months, and institutions now drive close to 40% of its flow. Hyperliquid, a perps venue rather than a betting site, launched its own prediction markets this year. And U.S. regulators are now reviewing prediction-market ETF structures, which is not a thing that happens to a fad.
So the category is real. The question every allocator and founder is asking is the wrong one: Kalshi or Polymarket? That's a front-end question. The more useful question is what compounds once event contracts become a permanent venue type. And the answer sits one layer down from the two names everyone knows.
The State of Prediction Markets in Mid-2026
Two platforms define the public picture. Polymarket runs the larger book by 30-day volume, on-chain and global, strongest in politics, crypto, and fast-moving events. On the regulated side sits Kalshi, the CFTC-registered U.S. venue that's become the default for American institutional desks that can't touch an offshore book. Between them they've turned event contracts from a curiosity into a liquid market that prints nine figures a day.
The institutional adoption is the part that changed the math. When roughly 40% of a venue's volume is institutional, the product is no longer "retail bets on elections." It's a hedging and expression instrument that sits next to options on a desk's screen. Kalshi now offers dedicated API access and tailored fee tiers to traders clearing six figures of monthly volume, the same playbook every maturing exchange runs when it starts courting flow it can rely on.
This matters beyond the two incumbents. Once a market type attracts institutional liquidity, it attracts competition for that liquidity. Hyperliquid's entry is the signal here. A high-performance on-chain venue building prediction markets natively tells you the category is now worth fighting over, and that the fight will happen partly on-chain, where settlement and composability are the edge.
Why the "Kalshi vs Polymarket" Framing Misses the Point
Most coverage treats this as a two-horse race. Pick the winner, own the category. That framing has a short shelf life, and the options market is a useful analogy for why.
In options, the venues that capture attention (the apps, the brokerages, the front-ends) are not where the durable margin lives. The margin lives in the plumbing: clearing, settlement, market data, the pricing and risk layer. Front-ends compete away their take rate because switching is cheap and users chase liquidity and fees. The infrastructure underneath gets stickier as volume grows, because everyone has to route through it.
Prediction markets are early enough that the plumbing is still being built in public. Every event contract needs a few things the front-end doesn't provide on its own: a resolution mechanism that's credibly neutral, a settlement rail that can move collateral without trust, and a data layer that lets desks price and audit the market. Get the resolution wrong and the whole contract is worthless. A market that can't be trusted to pay out correctly has no institutional bid, full stop.
That's the unglamorous part nobody tweets about. It's also where the defensibility is.
What the Data Actually Says
A few numbers are worth holding onto, and each points away from the front-end and toward the infrastructure.
First, the volume curve. Scaling to ~$21B in monthly volume in 2026 from a standing start is the kind of growth that breaks tooling. Resolution disputes, oracle latency, and settlement throughput all become real engineering problems at that scale — problems the incumbents would rather buy than build if a credible provider exists.
Second, the institutional share. The +800% growth in Kalshi's institutional volume over six months isn't retail euphoria; it's desks integrating event contracts into existing workflows. Desks demand exactly the things front-ends are weakest at: clean APIs, auditable resolution, reliable settlement, and data they can plug into risk systems. Demand from sophisticated counterparties pulls infrastructure into existence.
Third, the on-chain migration. Hyperliquid building prediction markets natively, plus Polymarket's on-chain settlement, means a growing share of this volume settles on public rails. We flagged the broader version of this in Brief #5 on the on-chain volume shift: when activity moves on-chain, the value capture moves toward whoever owns the settlement and data primitives, not the interface.
Put those together and the pattern is the same one that showed up in our Q1 funding analysis. Capital crowded into the legible front-end names (Kalshi raised $1B, Polymarket $600M) while the layer that actually makes the category work stayed comparatively unfunded. That gap is the opportunity.
What This Means for Founders
If you're building in this space, the instinct is to launch another venue. Resist it unless you have a genuine distribution wedge. The two incumbents have liquidity and brand, and a third general-purpose prediction-market front-end is a hard, capital-intensive fight you probably lose.
But the more fundable positions are in the plumbing:
- Resolution and oracle infrastructure. Build the neutral mechanism that settles disputed outcomes. This is the single highest-trust component in the stack, and the venues would rather integrate a credible third party than carry the reputational risk themselves. If your community is small but your resolution design is rigorous, lead with the design. That's what gets the meeting.
- Settlement and collateral rails. On-chain venues need collateral to move across chains and contracts without bridge risk. This is adjacent to interoperability, a problem our portfolio company Kima Network (a settlement layer that moves value across chains without bridges) has been working since 2023.
- Market-integrity and data tooling. Desks won't allocate into markets they can't audit. On-chain investigation and analytics, the kind of work Bubblemaps (a PRIM3 portfolio company that turns on-chain activity into visual investigations) does for token flows, map directly onto detecting manipulation, wash trading, and resolution gaming in event contracts.
A specific framing we give founders: if you can't get to meaningful liquidity in your first two quarters, you're probably building a front-end when you should be building a pick-and-shovel. The infrastructure layer doesn't need viral liquidity to prove itself. It needs one venue integration and one institutional reference. And the cleanest early proof point isn't a user-growth chart at all — it's a signed integration with a venue that already has the liquidity, because that one relationship validates the technology, the trust model, and the commercial wedge in a single move.
Where PRIM3 Is Looking
We don't have a prediction-market front-end in the portfolio, and we're not chasing one. What we're watching is the infrastructure thesis: the resolution, settlement, and data primitives that every event-contract venue will eventually depend on.
The reasoning is straightforward. Front-end venues are where attention concentrates and margins compress. Infrastructure is where attention is thin and switching costs compound. When PRIM3 evaluates a sector that's institutionalizing this fast, what we weight heaviest is who owns the trust-critical component, whether sophisticated counterparties are pulling the product into existence, and whether value capture survives commoditization of the front-end. On all of those, prediction-market infrastructure scores well.
What would have to be true for this to be wrong? If one venue achieves enough liquidity dominance that it vertically integrates resolution, settlement, and data into a closed stack, the independent-infrastructure thesis weakens. That's the real risk — a winner-take-most front-end that swallows its own plumbing. We don't think it plays out that way, because institutional desks prefer neutral, multi-venue infrastructure over single-venue lock-in. But it's the scenario we're underwriting against.
The category crossed the line from curiosity to asset class in 2026. The capital, so far, went to the logos. The compounding is one layer down.
FAQ
How big are crypto prediction markets in 2026? Combined platform volume scaled to roughly $21 billion per month in 2026, per TRM Labs, led by Polymarket and Kalshi. Kalshi's institutional volume grew about 800% over six months, with institutions now driving close to 40% of its flow.
Are prediction markets an asset class now? In 2026 they started trading like one. Institutional desks run event contracts as a distinct exposure with dedicated API access, and regulators are reviewing prediction-market ETF structures, both signs the category has moved past retail political betting.
Where is the investable value in prediction markets? PRIM3's view: the front-end venues will compete on fees and commoditize, while durable value accrues to the infrastructure beneath them — resolution and oracle layers, on-chain settlement, liquidity routing, and the data and integrity tooling that keeps markets honest.
That's where PRIM3 is looking in this category. More on our thesis and portfolio at prim3.vc.