If you're blindly aping into every $AI token narrative, stop now.
A lot of traders got burned chasing the AI hype cycle. Prices ran fast, expectations ran even faster, and now people are stuck wondering if the demand behind the narrative is actually sustainable.
Here’s the twist in the latest data. Some global companies are reportedly starting to pull back on AI spending after costs ballooned. Training and running large models isn’t cheap, and many businesses are realizing the bill shows up long before the profits do.
But at the same time, usage data tells a different story. On OpenRouter, Chinese AI models are now consuming around 18.5 trillion tokens per week, compared with roughly 6.0 trillion for US models. That’s more than 3x the activity across the top 9 models tracked on the platform, a rough proxy for real-world demand. If usage keeps accelerating like this, the infrastructure layer tied to projects like $FET and $AGIX could still see long-term tailwinds.
So which side wins: companies cutting AI budgets, or exploding real-world model usage driving the next wave of demand for $AI infrastructure?