Chinese AI firms may pivot to licensing model, Goldman Sachs says
Analysts suggest developers could monetize open-weight models through commercial agreements with cloud platforms.
Chinese AI developers may shift from fully open-weight model releases toward commercial licensing, according to Goldman Sachs research cited in a July 28, 2026 exclusive by the South China Morning Post. The move would mark a strategic pivot for firms that previously built global adoption through permissive open-source releases.
Ronald Keung, head of Asia internet research at Goldman Sachs, indicated Chinese AI developers could begin charging cloud platforms licensing fees to host their open-weight models. The SCMP report used the term "paid weights" — framing specific to that article rather than established industry language.
The potential shift affects Chinese AI companies including DeepSeek, Qwen, Kimi, and MiniMax, which have released models under permissive licenses such as MIT and Apache 2.0. These open-weight releases gained significant global traction and developer adoption, establishing Chinese AI capabilities internationally.
From open access to monetization
According to Goldman Sachs, the initial open-weight strategy served multiple purposes: building global brand recognition while circumventing compute constraints by allowing third-party providers to handle inference costs. Now that these models achieved widespread adoption, the incentive structure is changing.
The research discusses broader Chinese AI monetization strategies including subscription revenue and API pricing, suggesting licensing arrangements with cloud platforms represent another potential revenue stream as the sector matures.
Implications for the AI ecosystem
This pivot would mark a notable shift in how AI intellectual property is managed by Chinese developers. The open-weight approach previously allowed these companies to compete globally despite hardware restrictions, letting international cloud providers and developers absorb the computational burden of running inference.
Industry observers note that transitioning to a licensing model could affect the availability of truly open-source high-performance models. What began as a loss-leader strategy for brand growth may evolve into a more traditional software licensing framework once market position is secured.
The SCMP report rests primarily on Goldman Sachs research findings. Specific details about which cloud platforms would be targeted, what licensing fees might look like, and when any transition would occur were not independently corroborated by other major outlets at time of publication.
The development comes as global competition in AI intensifies, with Chinese firms seeking sustainable revenue models beyond API calls and enterprise contracts. How this strategy unfolds could influence open-source AI development patterns worldwide.