When Silicon Valley giants can no longer afford tokens

When Silicon Valley giants can no longer afford tokens, China’s large AI models are reshaping the global AI pricing landscape!

Key Highlights:

Cost Disruption:
Uber used Anthropic’s Claude for coding and burned through its entire 2026 AI budget in just four months. Consulting giant Accenture urgently restricted unnecessary employee use of AI. Meanwhile, an AI startup switched its business from Claude to DeepSeek and saved millions of dollars within a few months.

Data Speaks:
China’s large AI models reached 18.81 trillion token calls per week, surpassing the United States for eight consecutive weeks and firmly ranking first in the world. Among the global top five models by usage, four are from China.

New Cost-Saving Approach:
More companies are adopting a “model routing mechanism” — using cheaper Chinese models for simple tasks and reserving top-tier models for complex problems. On the OpenRouter platform, open-source model token usage has surged to 65%, doubling from earlier this year.

Why China? Not Just Subsidies:
The advantages are structural rather than purely subsidy-driven:

The “Eastern Data, Western Computing” initiative reduces computing costs by about 25%
Inference costs reduced by around 10%
Cheap electricity in western regions solves over 60% of electricity cost pressure
A large pool of young engineers enables low-cost experimentation and high-density R&D

Expert Prediction:
The era of a single “all-purpose model” is coming to an end. In the future, companies should not rely on just one universal AI tool in their toolbox.