Editor’s note: This article was written by Lucia, a TechNode reporter.
At a Sept. 1 online media briefing for the 23rd UBS Securities A-Share Seminar, discussion around China’s AI sector centered on a practical question: can lower costs and wider AI use translate into real business value?
UBS Securities China internet analyst Xiong Wei highlighted three themes to watch in China’s large-model sector: model capability, “token ROI” and monetization. While Chinese developers continue to improve their models — particularly in coding and agentic capabilities — companies using AI are becoming more selective about how much intelligence they actually need for a given task.
That marks a shift from “token-maxxing,” or encouraging more AI usage, toward “token optimization,” Xiong said. After rising AI bills made it harder for enterprises to measure the economic value generated by their token consumption, buyers have begun paying greater attention to the balance between performance and price. Xiong said that trend favors Chinese open-source models, whose improving capabilities and lower costs make them increasingly viable for repetitive or lower-risk workloads.
Xiong estimated that some leading Chinese models may cost less than one-tenth as much to develop as overseas peers, while average API pricing can be around 10% to 20% of international competitors. However, cheaper AI does not automatically translate into more revenue.
Kenneth Fong, UBS’s head of China internet research, said China’s major internet platforms face a more fundamental constraint: user traffic and time spent on mobile devices are no longer growing much. AI can lower content-production costs and make advertising or recommendation systems more effective, but consumers still have limited time and attention. He used AI-generated short dramas as an example: producing more shows at lower cost does not mean they will capture more of viewers’ time and attention.
A new experiment from Mango TV illustrates both sides of that equation. The Later Journey to the West, an AIGC-produced fantasy series, debuted on Mango TV and in Hunan Satellite TV’s prime-time slot on Aug. 31, becoming China’s first AIGC long-form series to reach a satellite-TV prime-time slot.
Adapted from an anonymously written late-Ming or early-Qing fantasy novel of the same name, the story follows a new generation of characters who reunite for another journey to obtain Buddhist scriptures after the originals are misunderstood.
Its first season is planned to run for 30 episodes of about 40 minutes each. The series is being made with Mango Lingchuang, Mango TV’s in-house AIGC production platform, which had served more than 40,000 professional users and supported over 3,900 projects by mid-2026.
The platform generated 109 character assets and 143 scene assets for the series, while the project is also testing a “produce, review and broadcast in parallel” model, allowing later episodes to remain in production as earlier ones air. The series also showed some early audience traction: real-time ratings for its Aug. 31 premiere ranked first among provincial satellite channels in the same time slot, according to ITHome, while ChinaTimes reported that it had logged 27.57 million plays on Mango TV by Sept. 2.
Yet whether those efficiencies and early audience gains can translate into a sustainable business remains unclear. Before the series aired, Hunan’s broadcasting regulator asked the project team to explore both a workable technical path for AI-powered long-form storytelling and a commercialization path for AIGC seasonal dramas.
That challenge echoes points made by both Xiong and Fong: as AI becomes cheaper to use and lowers production costs, the focus is shifting from how much AI can produce to whether its use can generate sufficient economic value.