Goldman Sachs forecasts monthly token use could reach 120 quadrillion by 2030, with Qualcomm CEO Cristiano Amon projecting a 40x increase in global token demand to 1.27 trillion every 10 seconds by 2030. This surge is driven by AI agents' multi-step planning and tool-use, consuming 10x to 50x more tokens per request than chatbots, creating a significant cost test for the AI boom.
AI agent activity is shifting compute demand from human-paced to agent-paced, creating a massive, persistent background infrastructure cost for every useful action.