When Zuckerberg's takeover talks with Kalshi CEO Tarek Mansour collapsed last year, Meta pivoted to building Arena, a standalone app that strips out real money wagering and lets AI generate the questions and resolve the outcomes.
Zuckerberg wanted to buy the prediction-market category leader. When that deal died, he did not walk away from the category. He rebuilt it without the part that made it worth buying in the first place.
That is the throughline of an NPR report published Tuesday, which describes how Meta CEO Mark Zuckerberg personally sat down with Kalshi CEO Tarek Mansour last year to discuss a takeover of the fast-rising prediction-market platform. The conversation never produced a formal offer. A year later, Meta is shipping its own version: a standalone app called Arena that lets users wager on real-world outcomes with play money, not real money, and routes both the questions and the resolutions through artificial intelligence.
Prediction markets are exchanges where users trade contracts tied to real-world outcomes, including who wins an election, whether a central bank cuts rates, and who takes home the NBA title. Prices fluctuate as traders bet, and the consensus price is treated as a crowd-sourced probability forecast. The category has split between regulated, real-money venues like Kalshi, which operates under US Commodity Futures Trading Commission oversight, and offshore or restricted-access competitors such as Polymarket, with a long tail of synthetic and analytics-only products tracked across industry dashboards.
Why the Kalshi talks stalled is the part NPR's sources cannot agree on. One person familiar with the meeting told the outlet that Mansour was simply unwilling to sell, even after Zuckerberg expressed interest. Another source said Meta walked because executives concluded the legal and ethical questions around Kalshi, a US-regulated venue where Americans can wager real dollars on everything from presidential elections to Federal Reserve decisions, were too messy to absorb into a consumer platform the size of Facebook or Instagram. NPR's own write-up calls the explanations "competing narratives" and does not pick a side. Either way, the talks never progressed to terms.
What Meta is building instead is structurally different from anything Kalshi ships today. Internal documents reviewed by NPR describe Arena as a standalone prediction-market product that uses "play money" rather than real-money contracts, sidestepping the CFTC-regulated derivatives framework that gives Kalshi its legal footing. The app leans on AI to generate the markets themselves and to resolve outcomes, an architecture NPR's June 24 reporting on Meta's prediction-market plans described in more detail. The bet, presumably, is that Meta can capture the social and engagement upside of betting on the news without inheriting the gambling-regulator footprint, the age-gating complexity, or the perception risk that comes with letting users win and lose actual dollars on whether the Fed cuts in September.
Stripping out real money is not a small product choice. It is the feature that gives Kalshi its current $22 billion valuation, arrived at after a recent $1 billion fundraise, and the regulatory standing that lets the company surface market odds in mainstream venues. Kalshi has already pushed into consumer-platform distribution, with its market data already appearing inside Meta-owned Threads, which is part of why a consumer social company might have wanted to own the asset in the first place. A play-money version can copy the look of the product. It cannot copy what makes the product worth $22 billion.
Columbia Law School professor Tim Wu, who has tracked Meta's category-shopping for years, offered a sharper read on the pattern to NPR. NPR's account paraphrased Wu as describing Meta as "clutching at every shiny object" — a characterization Wu paired, in the same reporting, with references to Meta's past forays into the metaverse and the failed Libra stablecoin project, a frame that reads as constructive rather than dismissive.
That is the part worth watching. A real-money prediction-market app launched by Meta, under Meta's brand, would be a category-shaking event on its own. A play-money, AI-resolved knockoff that Meta controls end-to-end is something else: a content surface dressed up as a market. The next tell is whether Arena ever lets users cash out, partners with a regulated venue, or simply becomes another feed Meta can monetize with ads.