Image credit: Source: SpeedLabs company announcement. Never imply stock depicts the actual event.
The next race in AI betting markets is not pricing odds faster, it is inventing markets that did not exist a second earlier, and a New York startup just raised 6.5 million US dollars to build the engine that does it. SpeedLabs closed the seed round to launch Momentum Markets, an AI system that generates entirely new in-game wagers from live sports moments rather than repricing a fixed menu of bets, with a first product due in summer 2026.
The round was led by Parlay Capital, with participation from Bullpen Capital, TA Ventures, and EdgeEquity, according to the company's announcement. Founder and chief executive Nick Meader is building it as B2B infrastructure, sold to sportsbooks and prediction-market platforms rather than run as a consumer app. The round is small; the idea is not, because it points at the part of the betting product that has been static for twenty years.
The old race was speed. The new race is creation
For two decades, in-play betting improved along one axis: how fast a book could reprice a fixed set of markets. A goal goes in, the model recalculates, the odds on next-goal and match-result update in milliseconds. Trading teams and data suppliers competed on latency and accuracy, and the menu of things you could bet on stayed roughly the same from one season to the next. Better pricing, same markets.
SpeedLabs is attacking the menu itself. Parlay Capital chief executive Greg Buonocore put the distinction plainly in the announcement: "SpeedLabs is doing something we have not seen before: using AI to create the market itself, not just price markets that already exist." Momentum Markets reads live game flow and spins up fresh, priced wagers around the moments that are actually swinging a contest: the injury that shifts a game, the run that flips a lead, the sequence that changes what fans are watching for. When the moment passes, so does the market.
That is a different product category, not a faster version of the old one. A repricing engine works from a catalogue a human defined in advance; a generation engine writes the catalogue in real time from what the game is doing. The first is a calculator, the second closer to an author.
Why market generation changes the economics of in-play
Betting is an inventory business. An operator's live revenue is bounded by how many distinct, credible wagers it can offer a customer during a game and how much margin each one carries. The industry has spent years expanding that inventory manually, first through micro-betting on the next pitch or possession, then through same-game parlays that bundle many correlated outcomes into one high-margin ticket. Both were attempts to multiply the number of things a customer can bet on inside a single event.
Automated market generation is the same expansion by another route, and a steeper one. If an AI can create and price a credible new market from any notable in-game moment, the ceiling on live inventory stops being what a trading team pre-built and starts being what the model can safely price. That multiplies wagering opportunities per game, and more priced opportunities, held at a sensible margin, is more revenue from the same broadcast minute. This is why the generation approach is being framed as infrastructure. It sits underneath the sportsbook and feeds it markets, in the same layer that data and trading services already occupy.
The margin question is the whole game. Generated markets are only worth having if they can be priced with enough edge and liability control to survive contact with sharp customers. A market invented on the fly, on a moment no trader vetted, has thin historical data and real model risk. Whoever solves the pricing and risk side of automated generation, not just the creation side, owns the category. Creating markets is the demo; pricing them safely at scale is the business.
The collision with prediction markets
The timing is not incidental. A generation engine that mints tradable contracts on live in-game events looks a great deal like the machinery underneath a prediction market, which is why SpeedLabs is aiming at both sportsbooks and prediction-market platforms as customers. This publication has argued that prediction markets are becoming a structural rival to the sportsbook, not a novelty. Real-time market generation is the technology that could blur the two into one surface: a live feed of continuously created contracts a user can trade, whether the operator calls it a bet or a market.
For prediction-market platforms, an AI that generates fresh contracts from live sport solves their thinnest problem: having enough timely, liquid things to trade during a game. For sportsbooks, the same tool defends the in-play franchise prediction markets are circling. The supplier that can feed generated markets to both sides sells into a fight rather than picking a winner, which is the more durable place to stand.
The data-rights and integrity problem generation creates
Automated generation raises the stakes on the two issues already straining live betting: data and integrity. A market invented from an in-game moment is only as trustworthy as the data feed describing that moment, in real time, with no human in the loop to catch a bad signal. That pushes even more value toward official, low-latency data and makes the fight over exclusive data rights sharper, because a generation engine is worthless on a slow or disputed feed. The market-generation layer and the data-rights layer are joined at the hip.
Integrity is the harder question. Manually built markets are vetted; a trader decides that a given proposition is safe to offer. When markets are generated automatically from live moments, some will inevitably land on outcomes that are easier to manipulate or that a single participant can influence, the exact niche markets integrity bodies already watch most closely. A system that can create thousands of micro-markets per game can also create thousands of small, thinly-traded targets. Any serious generation product has to ship with its own suppression logic, a way to decline to create markets that fail an integrity test, and regulators will want to see it. Expect that to become a licensing question, not a technical footnote.
Three signals to read
Three signals will tell whether AI-generated markets become core infrastructure or a niche feature. The first is a named operator deal. Momentum Markets is due to launch in summer 2026, and the moment it goes live inside a licensed sportsbook or regulated prediction-market platform, the category stops being a pitch and starts being a product. Watch for the first integration and how the operator describes the margin.
The second is the incumbents' response. The established trading and data suppliers already sit in the infrastructure layer SpeedLabs is entering. If they move to build or buy market-generation capability rather than treat it as a fringe idea, that is the tell that they see it eating the pricing franchise they have owned for years. A supplier that recently hired operator-grade commercial leadership will be reading this space closely.
The third is the regulators. The first jurisdiction to write rules for machine-generated markets, on how they are approved, how integrity suppression is proven, and how a market that no human authored is held accountable, will set the template everyone else copies. The technology is arriving before the rulebook, as it usually does. In AI betting markets, the companies that win will be the ones that build the rulebook into the product rather than waiting to be handed one.
Related coverage: The micro-betting product race | Same-game parlays as the sportsbook margin engine | Prediction markets as the sportsbook's structural rival | The sports-data exclusivity debate