For years, at Categorical we’ve repeated one simple truth: if you’re trying to build a new category, giving your product away for free is one of the most powerful growth strategies available. It’s how habits are formed, markets are created, and winners emerge.
But every land grab eventually ends, and the LLM industry may be reaching the moment where “free” starts becoming very expensive.
As the old saying goes, if the product is free, you’re the product. That idea once belonged to social media, broadcast television, radio, and free-to-play games. Now it increasingly applies to artificial intelligence. Only this time, it’s not just your attention that’s valuable – it’s your prompts, your data, and increasingly, the staggering number of tokens you or your company consumes.
The economics are starting to bite. Enterprises are discovering that AI token usage can spiral far beyond expectations. Workers are generating eye-watering compute bills. Companies like Google are already placing limits on resource-intensive features such as video generation. The honeymoon period of limitless AI may be over and that has profound implications for both compute use and employment.
At the same time, an uncomfortable competitive reality is emerging. While Western AI companies wrestle with the cost of delivering ever-more capable models, a growing wave of cheaper – and often free – models, many from Chinese developers, are offering businesses a compelling alternative.
If intelligence becomes commoditised, can premium pricing survive?
Then there’s Europe. Regulation, privacy laws, and fragmented markets are already reshaping how AI products launch and scale. Apple’s latest AI features and Siri upgrades have been delayed or withheld in Europe, raising a bigger question: could European policy become the force that fundamentally changes Big Tech’s ambitions to build global AI categories?
These aren’t isolated stories. They’re signals that the AI market is entering a new phase – one where pricing, regulation, competition, and economics may matter just as much as model performance.
The real question isn’t whether AI will get better. It’s whether the business model that brought us here is about to change – and what that means for everyone building, buying, or betting on the next generation of AI.