China AI Developers Drive Commercial Adoption as Costs Decline

Chinese artificial intelligence model developers are accelerating commercial adoption of their systems as training costs fall. Open-source AI models from China are seeing increased use across various industries, according to research firm Frost & Sullivan.

Major developers like DeepSeek, Alibaba Group Holding, and Baidu have shifted their focus. They now prioritise usability, cost efficiency, and broad ecosystem support to boost industry adoption.

Neil Wang, global partner and Greater China chairman at Frost & Sullivan, shared these insights in a Thursday interview with the South China Morning Post. The change coincides with a substantial decline in model training costs.

These costs have dropped 90%, lowering the barrier to entry for new AI projects, Wang added. This development reflects how Chinese AI companies are closing the gap with their United States counterparts.

They achieve this through an open-source approach, which makes the source code of AI models available for third-party developers to use, modify, and distribute. This strategy allows for wider accessibility and innovation.

Frost & Sullivan's latest report on China's AI market, released Thursday, highlights the nation's rapid progress in adopting AI applications. The report notes AI applications are evolving towards scenario-specific deployment from general capabilities.

China's finance, government, telecommunications, and healthcare sectors each boast an average AI penetration rate exceeding 60%. The government sector shows the highest adoption, reaching 95%.

The finance sector follows with a 78% adoption rate. Open-source models from start-ups like DeepSeek and MoonshotAI, along with those from Big Tech firms such as Alibaba AI unit Alibaba Cloud and Baidu, are widely adopted.

Their appeal stems from low-cost options and innovative features. For example, lower training costs enabled DeepSeek to focus on manageable costs, adequate quality, and a user-friendly toolchain, rather than just higher benchmark scores, Wang said.

Wang pointed out that DeepSeek's V3.1 model features a more flexible hybrid reasoning mode. This allows the system to automatically switch between deep thinking and fast response modes based on task complexity.

China's vast market in industrial AI application presents a clear advantage over the United States. Domestic AI firms are extremely fast at trial and error and technology iteration, creating opportunities for swift commercial adoption, Wang explained.

Despite this, bottlenecks persist in mainland China's AI model segment. Many developers still heavily rely on foreign technology providers for foundational development software, including AI frameworks and compilers, Wang stated.

---

• Chinese AI developers are increasingly focusing on usability, cost efficiency, and broad ecosystem support.

• Model training costs have declined 90% this year compared to 2024, reducing the barrier to entry for AI projects.

• China's finance, government, telecommunications, and healthcare sectors show over 60% AI penetration, with the government reaching 95%.

Source: SCMP