Token Usage of AI Agents Increases 14-Fold Since February

The token usage of AI agents has increased 14-fold since February, reaching approximately five times that of human users. As the use of generative AI shifts from simple conversations to task-executing agents, token usage has emerged as a key metric that determines the costs of AI infrastructure. Andreessen Horowitz revealed this fact on the 21st in their 'Charts of the Week' based on data from OpenRouter. Tokens are the basic units of text that AI models read and generate, serving as the measurement for model API billing and inference infrastructure usage. Human users typically interact with the model through a question-and-answer format, while AI agents increase token consumption by repeatedly reading, writing, and executing to achieve goals, recalling the same context and maintaining task states. This figure reflects trends observed on OpenRouter rather than the absolute scale of the entire AI market. With the rise of agent-based usage, cost management has become a major challenge for platforms, and OpenRouter has introduced features to automatically categorize requests by department, task type, agent complexity, and cost center. In South Korea, the spread of AI agents is related to issues of infrastructure costs, security, and data control, with companies like Rainer and Rebellion pushing forward with domestic NPU-based AI full-stack solutions.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.