The court fight is not the story. On Friday the 6th U.S. Circuit Court of Appeals ruled against Kalshi, siding with Ohio and Tennessee on whether sports-event contracts can be regulated under state gambling law. A three-judge panel was unanimous: Kalshi did not adequately show that its sports-event contracts are “swaps” under the Commodity Exchange Act. In August the 9th Circuit reached a similar conclusion. In April the 3rd Circuit went the other way, letting Kalshi keep operating in New Jersey while its appeal proceeds and saying it was likely to succeed on federal preemption. State lawmakers have now filed an amicus brief asking the Supreme Court to settle it.
Everyone is treating that as the main event: a jurisdiction war that decides whether prediction markets live or die in the United States. I think that is the wrong frame, and a comfortable one for an industry that would rather argue about who regulates it than about whether the thing it built can actually be operated.
The problem is not who regulates these venues. It is whether their settlement layer survives contact with a real position size. A prediction market sells exactly one thing: a contract that settles at $1 if an event happens and $0 if it does not. Everything that determines whether that promise is worth anything happens after the trade, in the layer nobody is pricing.
Thirty years of moving the hard part to the back end
Trace the history and this stops being mysterious. Single machine, then client/server, then browser/server, then middleware, then distributed services, then cloud. Not one of those steps removed complexity. Each one relocated it — off the client, into a back end the user cannot see or touch. That is the whole pattern of our industry: the front end gets simple, and the hard part moves behind a wall.
Prediction markets are doing precisely that to gambling’s messiest parts. Instead of a bookmaker’s discretion and a floor manager’s judgment, you get a published rule and a payout function. Architecturally, that is a genuine improvement. But it is only an improvement if somebody operates it, and “operate” is where this sector has not yet done the work.
I have used the same three gates for years to judge where technical difficulty actually lives. The first gate is business function: does it work at all. The second is business performance: does it work reliably, at scale, under load, with a runbook. The third is business intelligence: does it improve decisions. Most of this sector is standing on the first gate and talking about the third.
What actually got built
Set the pitch aside and look at the product. A contract settling at $1 or $0 turns its own price into an implied probability — trade at 60 cents and the market is saying roughly 60 percent. That is a legitimate information product, and it is not what is being traded.
Roughly 90 percent of prediction-market liquidity and turnover is sports contracts, and by the industry’s own account a large share of users treat the venues as an alternative sportsbook. That number is not a side note. It is the product. I have seen many companies do this: they describe the business they intend to build, then ship the one that gets traffic, and eventually the traffic becomes the strategy.
The financials say the same thing. Kalshi has raised at a reported $40 billion valuation and Polymarket closed a round at $20 billion, while DraftKings — the closest listed comparison, and an actual sportsbook — is worth around $13 billion. Analysts at Bernstein project prediction-market volumes could reach $1 trillion by 2030 with revenue near $10.8 billion. Those are private marks measured against one public comparable. I trust the comparable more than I trust the story attached to the marks.
The case for
Let me put the strongest honest version of the bull argument, because it deserves to be made properly:
- There is real risk that has no venue. A conference organizer in New York can do everything right and still lose a quarter’s revenue to a hurricane. No conventional instrument prices that exposure well. A clean $1-or-$0 contract does. That is a legitimate hedging use case, not gambling with a costume on.
- It is a genuine research product. Desks want to know whether market odds read the world better than polling does. That is information aggregation, and the demand is real regardless of what happens in court.
- Institutions are starting to show up. Morgan Stanley joined the NEXTPredict summit in New York as a strategic partner and is leading a panel on institutional capital, the first bank to attach its name to a public-facing initiative in the category. It also took part in Kalshi’s $1 billion Series F in May.
- The primitive is simple enough to standardize. One settlement rule, $1 or $0, is the kind of boring interface that scales — and it is natively expressible onchain, which is why crypto keeps circling this sector.
The case against
- Ninety percent of the flow is sports. You do not get to call yourself an institutional risk venue when your customers are using you as a sportsbook. The hedge-your-hurricane-exposure story may be true and still be irrelevant to where the money actually is.
- Legality depends on the state line. Two of three circuits that have ruled have now said these contracts are not clearly swaps under federal law. The 3rd Circuit disagreed, which means a product’s legal standing changes when it crosses a border. Anything that only works inside one circuit is, by definition, a special case.
- The state lane is not slowing down. New York sued Polymarket for running what it calls an unlicensed gambling operation, seeking forfeiture, restitution to users and fines equal to three times what it earned — after suing Kalshi for $36 billion in July, and Coinbase and Gemini in April, with Kentucky and Illinois following. Separately, CFTC staff have warned that “mention” contracts invite manipulation. The pressure is coming from both directions at once.
- The institutions that matter are explicitly waiting. JPMorgan’s chief executive said in April that the bank was weighing an entry while describing most of the activity as closer to gambling than investing, and ruling out sports and politics outright. Goldman Sachs said in January that it had met both large operators, and nothing followed. A precondition that does not exist yet is not a moat. It is a blockade.
Read the circuit split as an operability signal
The usual reading of the split is legal: the circuits disagree about preemption, so the highest court will eventually pick a winner. Read it as an engineer instead. The 3rd Circuit let Kalshi operate on a likely preemption win; the 6th Circuit, unanimously, said the contracts were not shown to be swaps; the 9th Circuit agreed with the states. That is not just a map of judicial opinion. It is a map of where your resolution rules stop working.
This is exactly how tech debt accumulates in a system. Nobody writes the debt down on purpose. It builds because a rule that holds in one environment does not hold in another, and every integration downstream quietly carries the difference. For a venue whose entire value proposition is a deterministic $1-or-$0 outcome, a settlement rule that varies by jurisdiction is not a compliance problem bolted on at the end. It is the core product failing the second gate.
And note that the federal side of this is not standing still — the SEC, CFTC and the Fed have been writing rules at speed all year on the crypto side of the house, as we tracked in our review of that rule sprint. Regulation arriving quickly does not make a venue operable. Operability is your problem, not the regulator’s.
The crypto angle is the resolution layer
- A $1-or-$0 contract is native to a blockchain. Deterministic payout, no clearinghouse, no counterparty discretion. This is one of the few places where crypto’s settlement pitch meets a non-speculative, mass-market use case.
- The hard part is the one crypto already knows. Who decides whether the event actually happened? That is the oracle problem, pointed at a consumer product with real money on both sides — and it is the same manipulation surface the CFTC flagged when it warned about “mention” contracts. The industry’s oldest unsolved weakness is now the deciding feature.
- Tokenized event exposure would spread the debt. If these contracts get wrapped and traded onchain, the circuit split stops being a legal footnote and becomes a composability and compliance boundary inside DeFi. Every protocol that touches the token inherits the ambiguity.
My standard, and what I rule
My standard is operability, not ideology. A venue is real when a position can be opened, priced and — above all — resolved to cash reliably, at size, under rules that do not change at a state line. That is the standard, and by it the verdict is split.
I rule for the idea and against the current product. The hedging case and the probability-feed case are both legitimate, and the primitive underneath them is sound. But a market that is 90 percent sports, resolves inconsistently across circuits, and needs a court to decide before a bank will quote it is a special case wearing a mainstream costume. It has passed the first gate and is nowhere near the second.
The verdict does not turn on who wins in court. It turns on resolution mechanics. Make outcomes deterministic, non-manipulable and portable — one settlement rule that means the same thing in Ohio, New Jersey and New York — and the legal fight becomes a detail you can litigate from a position of strength. Fail to do that, and even a Supreme Court win leaves you holding an un-operable product. Remember where the cost sits: eighty percent of the bill arrives after launch, and here the maintenance cost is not servers. It is resolution disputes and jurisdiction drift.
I would apply the same test I applied to AI agents earlier this month: verify what actually happened, not what everyone hoped was intended — a point we made in our postmortem on agent permissions. A market needs a verified answer to a factual question. Everything else about this sector is marketing on top of that.
A market that needs a court to tell it where it is allowed to resolve has already told you something. Not about the law — about whether anyone built it to be operated.
Related reading on BBVN Markets: Crypto got a spec instead of a law and the sandbox was never the safety boundary.
