Bots consume more tokens than humans on OpenRouter
Data from OpenRouter paints a clear picture: since February 6, AI agents have officially surpassed human users in token consumption. Agentic usage exploded with a 14x increase, completely overshadowing human growth which sits at a modest 2.8x.
This overtake marks a fundamental shift for AI system architectures globally. Nearly 70 percent of tokens consumed by agents come directly from cached prompts. This drastically reduces the actual operational costs for developers compared to estimates based on standard pricing.
Fabrizio's Analysis:
I expected this exact scenario for months. I always maintained that the true killer app for artificial intelligence is the agent-to-agent infrastructure, and these metrics prove my point entirely. Humans are slow, writing short prompts sporadically. Bots work in continuous loops, generating massive volumes of structured requests.
I note the 70 percent prompt caching metric with extreme interest. I personally optimized my workflows recently by leveraging long context caching, slashing inference expenses on repetitive tasks. Those building simple B2C wrappers are targeting the wrong market entirely. The real business today involves providing stable, scalable APIs for machines talking to other machines. This completely changes platform revenue models.
Daily digest and summary generated with AI from RSS sources.
The AI news that actually matters
I filter dozens of sources every day to find only what has real impact on work. Subscribe for a curated selection.
Sector Intelligence
Work Better with Artificial Intelligence
If this news is useful, turn it into action: practical prompts and concrete workflows to apply AI in your daily work.