Hi Dotikers!

A chart is making the rounds this week, and it deserves a moment of your attention. OpenRouter, the popular routing platform for AI models, has seen its weekly token consumption jump from 0.5 trillion in January 2025 to 126.2 trillion today. That is a growth of more than 25,000 percent in under a year. Impressive? Absolutely. Proof that AI adoption is exploding? Not so fast.

Here is the uncomfortable part. A big chunk of that growth does not come from millions of new users suddenly discovering AI. It comes from reasoning models that burn through tokens like teenagers through a data plan, and from AI agents that loop, retry and re-prompt their way to astronomical consumption, often without anyone optimizing anything. In other words, we are measuring activity, not necessarily value.

This matters because token volume has quietly become the industry's favorite vanity metric. Providers point to these curves to justify sky-high valuations and massive infrastructure spending. But if the demand is inflated by inefficiency rather than genuine usage, we are looking at a market that congratulates itself for being wasteful. And that, frankly, is the textbook definition of a bubble mechanic: growth that looks organic but is actually structural noise.

The honest reading is somewhere in between. Yes, real adoption is happening, and reasoning models do deliver better results for complex tasks. But the industry needs to stop treating raw token counts as a health indicator. The companies that will win the next phase are not the ones consuming the most tokens, but the ones doing the most with the fewest. Efficiency is about to become the new frontier, and the current chart, as spectacular as it is, might age like a 2021 crypto graph.

Let's dig into what this means for builders, investors and anyone trying to separate signal from very expensive noise.

Aym.