As China’s use of artificial intelligence (AI) surges, daily AI token consumption has increased more than 1,000-fold in about two years.

Cryptopolitan, a blockchain media outlet, reported on July 9 that China’s daily AI token usage rose from 100 billion in early 2024 to 100 trillion by the end of that year, and exceeded 140 trillion in March 2025.

Tokens are the basic unit used when AI models process and generate text and data. Most services, from chatbot replies to AI video generation, calculate costs and usage around this unit. The Chinese industry views expanding token use as forming an economic foundation in which AI services are bought, sold and priced in token units.

China’s AI expansion is spreading beyond the domestic market and into Western corporate markets. On OpenRouter, a platform that distributes requests across multiple AI providers, Chinese AI models have accounted for at least 30 percent of enterprise AI traffic since February 2026. By mid-2026, that share rose to 46 percent based on a CNBC tally. That is a sharp increase from an average 11 percent a year earlier and 4.5 percent in early 2025.

OpenRouter’s overall usage has also climbed quickly. Weekly token throughput increased from more than 5 trillion in April 2025 to more than 20 trillion in April 2026. Against that backdrop, the rise of Chinese models is attributed mainly to price competitiveness.

Justin Somerville of OpenRouter said open-source-based Chinese models are typically 60 percent to 90 percent cheaper than top-tier products from Anthropic and OpenAI. As of June 2026, OpenAI’s GPT-5.5 input token price was $5 per 1 million tokens, but DeepSeek V4 Flash was $0.14. Based on that price gap, Chinese models processed 4.12 trillion tokens in the week from Feb. 9 to 15, surpassing U.S. models on OpenRouter for the first time.

DeepSeek is currently the largest enterprise provider on OpenRouter. It processes 17.6 percent of all routed tokens, or 5.13 trillion a week. Alibaba’s Qwen followed with 13.9 percent, or 2.77 trillion a week. By country, Chinese models accounted for 46.4 percent of all tokens on the platform, versus 35.7 percent for U.S.-based models. Anthropic, the largest operator in the United States, was tallied at 14.8 percent.

Similar signals are emerging in corporate settings. Ara Karadjian of Ramp, a chief economist, assessed that cost is the biggest driver. He said DeepSeek has risen to become the fastest-growing software provider in the company’s index, and that such technology is moving beyond pilot projects into actual corporate spending items.

U.S. access restriction policies are also influencing changes in choices. As Washington allows access rights to its most sophisticated model, GPT-5.6 Sol, only to about 20 approved institutions, general users can access Terra and Luna, which are provided in stages, but Sol’s most powerful features require federal approval. Some in the industry view the United States as blocking access to its top-tier models and pushing companies to choose cheaper and more accessible overseas alternatives.

At the center of this demand expansion is Alibaba Cloud. Kenneth Fong of Citi Research and UBS raised his forecast on July 8 for Alibaba Cloud’s revenue growth in the first quarter of fiscal 2027 to more than 45 percent from 40 percent. Earlier, in the fourth quarter of fiscal 2026, Alibaba’s Cloud Intelligence Group posted 40 percent growth in revenue from external customers. AI products accounted for 30 percent of external cloud revenue at the time, and triple-digit growth also continued for 11 consecutive quarters.

Risks are also becoming more prominent alongside the pace of growth. Under China’s National Intelligence Law, the possibility that companies may have to disclose data to the government is heightening privacy concerns. U.S. regulation is also a factor that can disrupt AI access. Under U.S. export controls, Anthropic’s Fable 5 and Mythos 5 models were suspended on June 12 and restored on July 1.

As a result, companies and governments now need to weigh the low cost of Chinese AI models against privacy concerns, geopolitical tensions and uncertainty in U.S.-China trade relations.