China’s use of artificial intelligence has surged faster than expected, with daily AI token usage rising more than 1,000-fold in just over two years. The growth is also expanding into Western markets.

According to China’s National Bureau of Statistics, daily AI token usage increased from 100 billion at the start of 2024 to 100 trillion by the end of the year, reaching over 140 trillion by March 2025. That’s roughly 100,000 tokens per person across China’s 1.4 billion population.

Tokens are the small units of text and data that AI models use to process and generate content, forming the basis of everything from chatbot responses to AI-generated videos.

Industry experts in China say token use has grown so large that it is becoming the basis of a whole economy, with AI services increasingly bought, sold, and priced by the token.

Chinese models gain ground in the West

China’s AI industry is also gaining ground internationally. Since February 2026, Chinese AI models have accounted for at least 30% of corporate AI traffic on OpenRouter, a platform that routes AI requests to different providers.

According to CNBC, the share peaked at 46% by mid-2026, up from an average of 11% over the previous year and just 4.5% in early 2025.

OpenRouter’s overall AI traffic has also grown rapidly, increasing from more than 5 trillion tokens per week in April 2025 to over 20 trillion by April 2026.

The change is being driven by price. Open-source Chinese models are usually 60% to 90% less expensive than the best products like Anthropic and OpenAI, according to Justin Summerville of OpenRouter. Chinese AI models are also much cheaper.

As of June 2026, OpenAI’s GPT-5.5 charged $5 per million input tokens, compared with just $0.14 for DeepSeek V4 Flash. The price advantage helped Chinese models overtake US models on OpenRouter for the first time during the week of February 9–15, processing 4.12 trillion tokens.

With 17.6% of routed tokens, or 5.13 trillion each week, DeepSeek is presently the biggest provider on OpenRouter by firm. Next is Alibaba’s Qwen, with 13.9% and 2.77 trillion tokens per week.

Chinese models generate 46.4% of all tokens routed through the platform, while US-origin models produce 35.7%. The top US provider, Anthropic, owns just 14.8%.

Cost is the primary motivator, according to Ramp’s senior economist, Ara Kharazian, who notes that DeepSeek has emerged as the top trending software provider on Ramp’s own index, indicating that these technologies are now appearing in actual corporate spending rather than simply test projects.

It’s possible that US policy is also encouraging businesses to choose Chinese choices. Currently, only roughly 20 authorized organizations have access to Washington’s most sophisticated model, GPT-5.6 Sol. On July 9, GPT-5.6 became live in tiers: consumers can access Terra and Luna, but federal authorization is required for Sol’s most potent features.

Some in the business contend that the US is pushing consumers toward less expensive, more accessible alternatives from overseas by blocking access to its top models.

Alibaba’s cloud growth and the risks ahead

At the center of this change is Alibaba Cloud. Both Citi Research and UBS analyst Kenneth Fong increased their growth projections for Alibaba Cloud’s fiscal first-quarter 2027 revenue from a previous 40% estimate to over 45% year over year on July 8.

That builds on strong recent numbers: in fiscal fourth-quarter 2026, the company’s Cloud Intelligence Group posted 40% growth in external customer revenue, its eleventh straight quarter of triple-digit growth in AI products, which by then made up 30% of external cloud revenue.

There are concerns associated with Chinese AI’s fast growth.

Companies may be forced to disclose data to the government under China’s National Intelligence Law, which raises privacy issues.

However, access to AI may potentially be hampered by US limitations. For instance, US export regulations halted Anthropic’s Fable 5 and Mythos 5 models on June 12, then reinstated them on July 1.

As a result, businesses and governments must balance the lower cost of Chinese AI models against privacy concerns, geopolitical tensions, and the uncertainty of the US-China trade relationship.

The smartest crypto minds already read our newsletter. Want in? Join them.