The problem is not using many tokens; it's that most of them are being wasted

Companies are pouring money into AI tokens, but much of the usage is wasted on low-value tasks and fixes instead of real production work.

The article argues the problem is not token consumption itself, but inefficient use: only a small share of AI spending reaches lasting business value.

Executives from firms like Uber and Amazon are cited as examples of teams overusing large models for trivial requests and rewrites.

Agentic coding tools and monthly AI plans such as Claude Code, ChatGPT Plus, and similar products are sharply increasing token consumption.

As costs rise without matching productivity gains, companies may respond with tighter access controls, usage limits, or higher pricing.