AINEWS
Why Meta built a fake-money prediction app instead of buying Kalshi
When Zuckerberg couldn't acquire Kalshi, the leading U.S. event betting platform, Meta built Arena, a play money version where AI picks the questions and decides who wins.
When Zuckerberg couldn't acquire Kalshi, the leading U.S. event betting platform, Meta built Arena, a play money version where AI picks the questions and decides who wins.
When Meta couldn't buy Kalshi, the leading U.S. event-betting platform where users wager real money on real-world outcomes, it didn't settle for second place. It built a structurally different product: Arena, a standalone app where Meta's own AI generates the questions and decides who wins, all using fake currency instead of cash.
According to three sources familiar with the discussions who were not authorized to speak publicly, Meta CEO Mark Zuckerberg met with Kalshi CEO Tarek Mansour last year to discuss a possible takeover. Neither Kalshi nor Meta commented when asked by NPR.
What happened next is reported in two competing versions. Some of those sources said Mansour would not move forward with a sale. Others said Meta walked away because it judged Kalshi's legal and ethical exposure too messy. NPR, which first reported the acquisition approach, attributes both accounts to anonymous sources without resolving which is closer to the truth.
The acquisition would have given Meta immediate ownership of the regulated real-money market. Without it, Meta is taking a different route. Internal Meta documents reviewed by NPR show the company is building Arena as a standalone prediction-market app that runs on "play money" rather than real wagers. That makes Arena fundamentally different from Kalshi, which is overseen by U.S. commodities regulators, and from Polymarket, a crypto-based competitor, both of which let users win or lose actual cash.
The structural gap is the point. Prediction markets have grown rapidly under a permissive U.S. regulatory environment, and a play-money product sidesteps the licensing, compliance, and disclosure regimes that govern real-money event trading. Arena is not a cheaper Kalshi. It is a different category, one that does not require a federal license.
According to those same internal documents, Arena's AI systems will both generate the questions and determine winners and losers based on whether the events in question actually occur. The product is not yet public, so feature claims remain hedged to what the documents describe and to what NPR has reported.
The category itself has drawn Meta's interest because prediction markets have attracted unusual growth since mid-2025, when a permissive federal posture toward event-betting opened the door to consumer products that would have struggled a year earlier. The New York Times reported this month that Meta was racing to ship Arena, and followed up with a separate look at Zuckerberg's posture toward both Kalshi and Polymarket.
Columbia law professor Tim Wu, a former White House tech-policy adviser, told NPR the move fits a familiar Meta pattern: "clutching at every shiny object." The framing matters after the company pulled back from a metaverse-first strategy and abandoned its Libra cryptocurrency project. Meta is not just curious about prediction markets. It tried to buy the leader, failed, and built a structurally different product rather than wait for a real-money license.
What to watch next: whether Arena ever graduates from play money to real wagers, whether U.S. regulators treat an AI-resolved event-betting app as a peer to Kalshi's regulated exchange, and whether Meta's audience responds to a knockoff that pays out in fake currency when competitors pay out in dollars.