Chinese artificial intelligence developers, long celebrated for their commitment to open-source distribution, may soon begin charging cloud platforms commercial licensing fees to host their open-weight models. According to a new analysis by Goldman Sachs reported by the South China Morning Post, this strategic shift is being driven by the soaring global use of Chinese AI and the growing need for developers to capture more revenue from their increasingly sophisticated and commercially valuable creations.
The assessment comes at a critical juncture for the Chinese AI industry. Models such as Moonshot AI’s Kimi K3 and Zhipu AI’s GLM-5.2 have recently reached performance levels that are just a fraction behind top U.S. rivals, and their global adoption is accelerating rapidly.
Yet the current distribution model presents a significant revenue gap for the companies investing billions in research and development. Goldman Sachs previously urged investors to go long on China’s AI value chain amid a “triple mismatch” between capability, adoption, and monetization, and this latest analysis suggests that mismatch may be approaching resolution.
The Open-Source Revenue Gap
Currently, most Chinese models are distributed under permissive software licenses, such as the MIT License, which originated from the Massachusetts Institute of Technology. This allows foreign platforms and developers to freely download, modify, and host the software’s core “weights”, the underlying parameters that encode the AI’s intelligence, even for commercial purposes. While this approach has rapidly accelerated global adoption, it limits the direct financial return for the companies investing enormous sums in research and development.
Keung suggests that Chinese model builders could generate substantial revenue by requiring third-party providers to purchase commercial licenses to serve these models (a process known as inference) on their own infrastructure. “Domestic growth in adoption is very fast, and we think there’s considerable adoption among small and medium enterprises in the global market, and even larger [companies] are starting to consider using [Chinese models],” Keung noted. The scale of this adoption has been dramatic: a CNBC investigation found that Chinese-origin models peaked at 46 percent of U.S. enterprise token traffic, up from just 4.5 percent a year earlier, a tenfold increase that has alarmed U.S. national security officials and delighted Chinese developers in equal measure.
The precedent for this shift already exists in the broader software industry. Red Hat built a multi-billion dollar business on commercial licensing and support for Linux, an open-source operating system. MongoDB and Elastic both transitioned from permissive open-source licenses to more restrictive models when cloud providers began offering their software as a managed service without contributing back to development. Chinese AI developers are watching these precedents closely, and the logic of the transition is compelling: once a technology becomes critical infrastructure, the market will pay for reliability, support, and guaranteed access.
Shifting the Monetization Strategy
The potential shift toward paid weights reflects a maturation of the Chinese AI sector. As these models transition from experimental tools to critical enterprise infrastructure, the financial models supporting them must also evolve. If Chinese developers begin enforcing commercial licensing, it could significantly alter the economics of the global AI market. Cloud providers and enterprises that have built services on top of free Chinese models would face new costs, potentially leveling the playing field with U.S. closed-source providers who already charge premium API fees.
This move would also signal a confidence among Chinese developers that their models are now indispensable enough to command a price tag in the international market. DeepSeek founder Liang Wenfeng has acknowledged that while open-source is a core principle, the company is also pursuing commercialization in parallel. The tension between open-source idealism and commercial sustainability is not unique to China, Meta’s Llama models, for instance, carry commercial restrictions for large enterprises.
The timing of Goldman Sachs’ analysis is notable: it arrives precisely as the debate over whether to ban Chinese AI models in the U.S. is intensifying. If Chinese developers begin charging for their models, the economic argument against a ban becomes even more complex, American enterprises would be paying licensing fees to Chinese companies for technology that U.S. national security officials simultaneously argue poses a threat. That paradox may ultimately force a resolution to the open-weight debate that neither pure market logic nor pure security logic can achieve on its own.
For Chinese developers, the transition to paid weights would also represent a significant strategic maturation. It would signal that they are no longer content to be the world’s AI infrastructure provider on a charitable basis, that they intend to capture the commercial value of the capabilities they have built.
The question is whether the global market, having grown accustomed to free Chinese AI, will accept this repricing. Given the performance levels now demonstrated by models like Kimi K3 and the impending DeepSeek-V4 release, the answer may well be yes. The era of free Chinese AI may be drawing to a close, and the era of Chinese AI as a commercial force may be just beginning.