Surge in Model Token Use to Send China's AI Market Value Soaring, Industry Expert Says
Listen to the full version
China’s public cloud service providers are expected to see a sixteenfold increase in the use of their large model tokens in 2025, driving the value of the country's artificial intelligence (AI) market up 30% to more than 1.2 trillion yuan ($170 billion), an industry expert said.
The number of large model tokens used this year may exceed two quadrillion, signaling a massive ramp-up in the adoption of generative AI technologies in China, according to an estimate made Saturday by Wei Kai, head of the China Academy of Information and Communications Technology (CAICT)’s AI Institute. Tokens are units of data processed by AI models during training and inference, enabling prediction, generation and reasoning.
Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- RELATED
- DIGEST HUB
Digest Hub
Back
- China’s AI market is projected to grow by 30% in 2025, reaching over 1.2 trillion yuan ($170 billion), driven by a sixteenfold surge in large model token use.
- Google’s Gemini 3 Pro outperformed OpenAI models, with a 30% performance increase over o1 and 50% over GPT-4o in multimodal tasks.
- AI agent accuracy improved from 28% to 50%, with rapid adoption in manufacturing, finance, and healthcare, though challenges remain in scalability and integration.
- According to CAICT's testing data, Google's Gemini 3 Pro model, released this month, showed a 30% performance improvement over OpenAI's o1 model. In the multimodal category, Gemini 3 Pro also demonstrated a 50% capability increase compared to OpenAI's GPT-4o.
- OpenAI
- OpenAI's "o1" model, released in December 2024, was considered the strongest of its time and was outperformed by Google's Gemini 3 Pro by 30%. In the multimodal category, OpenAI's GPT-4o, launched in May 2024, showed a 50% lower capability than Gemini 3 Pro.
- May 2024：
- OpenAI’s GPT-4o launched.
- December 2024：
- OpenAI’s o1 model released.
- 2024-2025：
- AI models showed significant gains in reasoning, coding, emotional intelligence, tool usage and long text processing; AI agents rapidly advanced and adoption increased.
- 2025：
- China’s public cloud service providers are expected to see a sixteenfold increase in the use of their large model tokens.
- 2025：
- The value of China’s AI market is expected to rise 30% to more than 1.2 trillion yuan ($170 billion).
- 2025：
- The number of large model tokens used in China may exceed two quadrillion.
- 2025：
- 27 robot training facilities established nationwide in China for embodied AI data collection.
- 2025：
- Significant progress in motion control for embodied AI industry in China.
- 2025：
- Global AI industry described as transitioning from technical innovation to industrial application.
- June 2025：
- Google’s Gemini 3 Pro released.
- MOST POPULAR
Sign in with Google
Sign in with Facebook
Sign in with 财新