AI Use Surges, But So Do Costs, Hinting at Limits to Automation
Companies deploying advanced AI models are encountering significant expenses, potentially indicating a ceiling for purely automated tasks.
The adoption of artificial intelligence has seen a dramatic increase in 2026, with companies leveraging AI's ability to process information in units called "tokens." These tokens, small chunks of text used for understanding and generating language, are being consumed in unprecedented quantities. Despite a declining "per token" cost, the sheer volume of usage has led to substantial expenditures.
Top companies utilizing AI have implemented "token leaderboards" to incentivize employees to maximize productivity gains from advanced AI models. This surge in usage is largely attributed to "agentic use," where AI agents perform tasks automatically. However, the immense bills generated by these operations have prompted many firms to begin rationing AI model usage.
This trend suggests potential limits to the extent of work that can be automated. In some instances, the cost of virtual AI workers may be exceeding that of human employees, depending on the specific task.
An additional factor influencing the AI landscape is the shift by some companies, including Western ones, towards cheaper AI alternatives. These are often derived from Chinese models that are being made freely available on the market. While significant uncertainty remains, these developments point to clear emerging trends in the practical application and economic viability of artificial intelligence.