AI agents are consuming significantly more tokens than humans, with one platform, OpenRouter, showing agents using 7.3 trillion tokens compared to humans' 1.4 trillion. This trend is projected to increase, potentially reaching a 10x ratio or higher. A substantial portion of this agent token usage, over 85%, comes from rereading cached prompts, which, while cheaper than processing new information, still drives up demand for high-bandwidth memory (HBM). This escalating demand for memory is contributing to existing shortages and is expected to worsen, impacting both AI data centers and potentially PC buyers. AI
IMPACT Accelerates demand for high-bandwidth memory, potentially leading to shortages and increased costs for AI infrastructure and consumer hardware.
RANK_REASON Analysis of AI agent token usage trends and their impact on hardware, not a direct release from a frontier lab.
- Andreessen Horowitz
- Claude Code
- Daniel Newman
- DeepSeek
- McKinsey & Company
- Micron
- Nvidia
- OpenRouter
- Peter Walker
- X
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →