The Big Discount on Intelligence (arunabhdastidar.substack.com)

A recent exploration into AI pricing models reveals a dramatic shift in the release cadence of frontier AI models, with average intervals shrinking from 38 days in 2023 to as little as 4.5 days today. This quick succession indicates not only a mechanization of model production but also highlights the substantial financial investments being made in reinforcement learning (RL) environments, with companies like Anthropic and Surge reportedly spending and earning over a billion dollars annually. The overall low subscription costs for these AI models, juxtaposed with exorbitant API rates, suggests companies are providing affordable access to potent intelligence as a strategic move to gather user data and insights—essentially, trading low prices for valuable usage data that shapes future model development. Moreover, the implications of these pricing strategies raise questions about long-term market positioning. As companies like Anthropic eye a $30 trillion addressable market, the rush to capture new customers at discounted rates may not be sustainable. Analysts suggest that the true value lies not in selling subscriptions but in wielding the intelligence models for competitive advantage. As firms build AI-native alternatives across various industries, this pricing strategy could reshape entire market landscapes and determine who holds the dominant knowledge bases essential for future innovation. The narrative emphasizes that decisions made by users while utilizing these AI tools, rather than the data itself, create unique value propositions that are increasingly critical in a rapidly evolving technological ecosystem.