The competition among leading companies in the field of artificial intelligence has gradually transformed into a massive prediction market. On platforms like Kalshi, Polymarket, and similar sites, traders and enthusiasts place multimillion-dollar bets, trying to predict which AI model will perform best over a given period.
This is reported by Business • Media
Growth of Volumes and Strategies in the Prediction Market
The monthly trading volume on contracts related to AI model races reached $20 million in August 2025 — ten times more than at the beginning of the year. On these platforms, bets operate on a contract basis: the price reflects the probability of a specific outcome, and participants can sell their predictions before the final settlement, profiting from changes in odds.
Experienced traders actively utilize insights from social media, technical signals, and leaderboards. Notably, renowned trader Foster McCoy earned $10,000 in just a few hours by betting on Google’s Gemini advantage over GPT-5. According to him, constant monitoring of platforms like Discord and leaderboards helps gain an edge. Since the beginning of the year, he has traded $3.2 million through Kalshi and made a profit of $170,000.
Impact of Social Media and Insights on the Market
Activity in the prediction market directly responds to events on social media. For example, after Elon Musk’s posts about the advantages of Grok, the price of the “Grok to Win” contract surged by 500% within a few hours but soon plummeted sharply. Traders track hints in posts from leaders at OpenAI, Google, as well as changes in GitHub code and mobile applications to anticipate the release of new models in advance.
Harvard student Rishab Jain also took advantage of such insights, earning $3,500 by early detection of signs indicating a rapid release of Gemini. He analyzes technical changes in the backend of Google products that may signal updates to the AI model.
George Mason University economics professor Robin Hanson believes that the rapid increase in turnover enhances the value of reliable information, while random bets become increasingly risky. Alongside professional players, more ordinary users are entering the market, eager to profit from the active races of AI models that are in the media and social media spotlight.
“I just love technology and can spend 10 minutes checking the market,” confessed investor James Cole.
The latest update to GPT-5 has made this model “warmer and friendlier,” which, according to OpenAI, aimed to make communication with ChatGPT more natural and accessible. These changes were prompted by numerous user feedback regarding the excessive formality of GPT-5. Additionally, the Claude chatbot from Anthropic recently received a feature to complete conversations with users.