This Is the New Oil and Most People Have No Idea: Why Tokens Become the Currency of AI

A single server with eight Nvidia H100 chips cost somewhere between 200 to $400,000. Looking at the most expensive models, GPT4.5 as an example, when it first came out, that was $75 per million input tokens and $150 per million output tokens. And by October 24, one could get GPT3.5 level performance for 7 cents per million tokens from 20 bucks to 7, a 280 times collapse in under 2 years. Enterprise AI spending surged from 11.5 billion in 2024 to 37 billion in 25. The average monthly AI budget per organization in 2025 was $85,000, up 36% from the year before. A coding agent solving a bug might chew through 50,000 to 500,000 tokens in a single session. Overall, this points to specialized systems and vertically integrated players shaping what comes next far more than surface narratives suggest.

Check the video here.