Digital News

　　The Cyberspace Administration of China (CAC) recently revealed that, to date, over 490 large models have been officially registered with the agency. The total number of individual user registrations for large models applications has exceeded 3.1 billion, while the number of Application Programming Interface (API) users has surpassed 159 million. These figures mark a historic milestone for China’s AI industry: the world’s largest, most diverse, and fastest-evolving AI testing ground has taken shape. 

　　The number of registered individual users for large AI models in China is more than twice the country’s population, suggesting that it’s common for individuals to register multiple accounts. What benefits could such a massive number of registered users bring to China’s AI industry development?

　　Data Benefits. Every day, 3.1 billion users produce a continuous stream of conversations, images, and videos—essentially providing “super fuel” for AI, driving the ongoing evolution of large models. For example, an elderly fruit farmer and a graduate student from an agricultural university would phrase the question “How to grow strawberries” very differently. These differences help AI evolve to better understand your questions and even anticipate what you might ask next.

　　Scenario Benefits. From factory assembly lines to live-stream sales, from medical imaging in hospitals to UAVs for farm use, China has turned nearly every industry into an AI “training ground.” For example, factories in Shenzhen use AI for quality inspection, while farmlands in Henan employ AI to predict pests and diseases. This kind of real-world training is something foreign AI systems can only envy.

　　Ecosystem Benefits. A large user base means a vast market, and open-sourcing has become the common choice for leading companies aiming to expand their footprint. Open-source large models are like “shared power banks”—giving smaller companies affordable access to cutting-edge technologies. This “technology supermarket” model is fueling a wave of innovative applications.

　　Certainly, such vast benefits are accompanied by equally considerable challenges.

　　Computing Power Challenges. Training large models is like “feeding” AI. AI has an enormous appetite, so the “kitchen” equipment must be powerful enough. Right now, the global AI chip market is like a restaurant dominated by NVIDIA, the “foreign master chef.” With recent reports of security vulnerabilities in their chips, isn’t it even more important to rely on domestic alternatives? Huawei, Cambricon, and other domestic companies have already brought their AI chips to the table, but our “cooking skills” are still a work in progress — there’s plenty of room to improve!

　　Security Challenges. Behind 3.1 billion AI users, there are many hidden “bombs.” Chat logs and photos could be snooped on — that’s a privacy bomb. Questions asked in dialects often get poor answers — that’s an algorithm bias bomb. AI-generated celebrity product promotion videos can be indistinguishable from real ones — that’s a deepfake bomb. AI is developing so rapidly that regulatory frameworks and technical standards are still struggling to keep pace. Company self-inspections and government regulation often occur independently, highlighting the need for better coordination mechanisms.

　　Scientific Research Challenges. Compared with the U.S., China still lags in core large model technologies. In fundamentals such as algorithm design and development frameworks, the U.S. maintains a stronger foundation. In terms of basic R&D investment, the U.S. is like a seasoned player constantly pouring in resources. The good news is we’re closing the gap much faster than they anticipated! The AI Index Report 2025 released by Stanford University states that the performance gap between China’s and the U.S.’s leading large AI models has sharply narrowed from 17.5% in 2023 to just 0.3% in 2025, almost a “zero-generation gap.”

　　3.1 billion users not only reflect the market’s scale but also serve as the ultimate stress test of the nation’s innovation capabilities. Transitioning from having the “largest user base” to achieving the “strongest overall strength” is no easy or automatic step. We still need to focus on the following four areas.

　　First, we must strengthen the computing power infrastructure and accelerate the large-scale commercial deployment and iterative upgrades of domestic AI chips. Second, we need to strengthen basic research by fostering collaboration among universities, research institutes, and enterprises. Third, we should refine the governance framework to balance innovation-driven development with effective risk management. Finally, we aim to establish a global open-source hub that promotes collaborative innovation among developers worldwide, leveraging Chinese wisdom to help define new international standards.

　　When AI meets the real needs of a vast user base, the potential impact could far exceed everyone’s expectations. Whoever leads in turning the “largest testing ground” into the “most powerful innovation hub” will gain the upper hand in the next global AI race. We hope that 3.1 billion users will be more than just a numerical miracle, but the true driving force behind the AI revolution.