Is the Era of "Expensive AI" Over?

US enterprise market data shows a massive shift in real-time: companies are beginning to "abandon" top-tier, ultra-expensive US models in favor of Chinese alternatives like DeepSeek and Zhipu GLM.

The numbers are actually wild. According to OpenRouter, the usage share of these Chinese models has surged, hitting peaks of 46% recently. This isn't just a minor blip; it's a structural change in how AI is being deployed.

In my opinion, this isn't about "better" vs "worse," but rather "smart" vs "wasteful." Most enterprises realize they don't need a trillion-parameter powerhouse to summarize a meeting or write simple code. Why pay a premium for a sledgehammer when a small hammer will do? Chinese open-weight models are currently priced 60% to 90% lower than their US counterparts, and the performance gap has shrunk to just a few months.

We are seeing startups save millions of dollars simply by rerouting their workflows to these high-efficiency models. It’s a classic case of "good enough" technology disrupting a high-margin market. The era of blindly paying for the most famous model is ending; the era of cost-effective, task-specific AI is officially here.

What do you guys think? Are you seeing this shift in your own workflows?