▲ Alibaba Group of China
China's Alibaba is reportedly pushing to introduce a new billing policy that requires revenue sharing for companies commercially utilizing its next-generation open-source AI model, "Qwen 3.8-Max."
According to a report by Reuters on August 7 (local time), Alibaba plans to introduce the revenue-sharing billing policy starting next week, similar to the approach applied by its competitor and Chinese startup Moonshot AI to its AI model "Kimi K3."
While open-source AI models generally involve no cost for small-scale use, Moonshot AI is known to demand up to a 30% revenue share from companies with annual revenues of 20 million dollars (approx. 28.4 billion KRW) or more when providing its model in a service format.
Alibaba has also previously charged usage fees only for the Qwen models operated on its own cloud service.
On the other hand, it did not charge separate fees when other companies installed and utilized the models in their own data centers.
This policy shift has led to analysis that Chinese AI companies, which have expanded their market share by leveraging affordable yet high-performing open-source models, are now moving to improve profitability through a freemium strategy.
The freemium strategy is a monetization approach commonly used in enterprise software and mobile gaming industries, where the core service or content is provided for free, and monetization is selectively applied based on usage patterns such as large-scale commercialization.
Chinese IT service firm ChinaSoft International has already disclosed that it signed an agreement with Moonshot regarding AI revenue sharing, and U.S. cloud company DigitalOcean and others are also known to have signed similar revenue-sharing agreements while using Chinese AI models.
Open-source AI is characterized by making the AI's architecture, weights (computation pattern) data, and other elements publicly available so that anyone can easily adopt and modify the model, a method widely adopted by the Chinese AI industry to expand its customer base.
Such open AI can expand user ecosystems much faster compared to closed models like Anthropic's Claude, but the monetization means for developers can be limited.
Reuters added that prominent U.S. AI companies such as Thinking Machines Lab are also joining the open-source AI market in large numbers, which is expected to further intensify monetization competition within the open-source industry.
(Photo: AP, Yonhap News)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.