A significant shift is occurring in the global artificial intelligence landscape as China’s large language models (LLMs) achieve a new milestone in operational scale. According to the latest data from OpenRouter, a leading global aggregator for AI model APIs, China’s weekly token usage reached 4.69 trillion as of mid-March. This figure marks the second consecutive week that China has outpaced the United States in aggregate AI demand, signaling a transition from a race of theoretical capabilities to one of massive industrial application.
In a striking display of market dominance, the top three positions for global model calls are currently held by Chinese-developed architectures. This surge in volume suggests that Chinese enterprises and developers are integrating AI at a faster rate than their international peers, leveraging a combination of aggressive pricing, local infrastructure optimization, and a rapidly maturing ecosystem of consumer-facing applications.
Financial institutions are taking note of this exponential trajectory. J.P. Morgan recently projected that China’s consumption of AI inference tokens will skyrocket from approximately 10 quadrillion in 2025 to a staggering 3,900 quadrillion by 2030. This represents a 370-fold increase over just five years, highlighting a future where AI becomes the foundational utility for the world’s second-largest economy.
The implications of this data extend beyond mere statistics. While the United States has historically led in the development of frontier models, the 'Inference Era' belongs to the players who can deploy at scale and minimal cost. China's lead in token consumption indicates that its domestic industry has successfully cleared the hurdle of adoption, moving AI from the laboratory into the fabric of daily digital life and industrial workflows.