Last week a few enterprise teams quietly started trimming their AI budgets after their usage bills came back far higher than expected.
For crypto investors, this is a familiar trap. When a narrative gets hot, money floods in fast. Then the real costs show up, and suddenly the market has to reassess what “demand” actually looks like.
A useful signal came from OpenRouter, a platform that aggregates access to multiple AI models. Tracking the top 9 models, weekly token consumption shows Chinese AI systems burning through about 18.5 trillion tokens per week, more than triple the roughly 6.0 trillion used by US models. On the surface that looks like explosive growth and dominance.
But here’s the catch. The same surge in usage is exactly what’s triggering cost alarms for companies experimenting with large-scale AI deployments. When token consumption spikes this aggressively, infrastructure and compute expenses scale just as fast. Some firms are already dialing spending back while they rethink ROI.
That matters for the crypto side of the AI trade too. Narratives around decentralized compute and AI infrastructure tokens like $FET, $AGIX, and $RNDR often assume demand only moves in one direction. If enterprises start optimizing or cutting usage after the first wave of experimentation, growth expectations across the whole sector could get repriced.
So the real question is whether this surge in AI token usage signals durable demand, or just the expensive trial phase before budgets tighten. What do you think?