China's AI Strategy Shifts Focus to Monetization and Efficiency

Recent Developments in China's AI Industry

During a recent online media briefing on September 1, the dynamics of China’s AI industry were brought into sharp focus, addressing one pivotal question: can lower costs and broader AI adoption truly create tangible business value? This inquiry is central not just to investors and policymakers but also to tech firms navigating an increasingly complex landscape influenced by rapid technological changes and market demands.

Key Themes Shaping AI's Evolution

UBS Securities analyst Xiong Wei pointed out three critical themes shaping the future of large-model AI in China: model capabilities, “token ROI,” and monetization aspects. Model capabilities have indeed progressed significantly, especially in areas like coding and autonomous functionalities. This evolution reflects the ongoing push towards more sophisticated AI applications. Yet companies are now more discerning about the level of intelligence needed for specific applications.

This marks a significant shift from what Xiong termed “token-maxxing,” which encouraged widespread AI usage without deeper consideration of individual application needs, to a trend dubbed “token optimization.” Rising operational costs associated with AI have precipitated this change, prompting businesses to scrutinize the performance versus expenditures. The financial implications are substantial; if you’re working in this space, understanding how companies balance investment in AI and expected returns is vital. Moreover, the increasing favor for Chinese open-source models is notable. These models can be particularly appealing when deployed for repetitive or lower-risk tasks, providing enhanced capabilities at reduced costs.

Cost Efficiency and Market Realities

He further estimated that the development costs for leading Chinese models could be as low as one-tenth compared to their international counterparts. This stark difference positions Chinese companies to potentially gain significant market share in various AI segments. With API pricing hovering around 10% to 20% of that of foreign competitors, it’s a compelling proposition for firms operating under tight budgets. However, cheaper AI doesn't guarantee increased revenue, and this highlights a nuanced market reality. The economics of AI are complex; the implications suggest that lower costs might not translate into better financial performance without a concurrent increase in product or service demand.

Challenges of User Engagement

Kenneth Fong, head of China internet research at UBS, emphasized a fundamental challenge: user engagement on major internet platforms isn't growing appreciably. While AI can effectively reduce content production costs and create efficiencies in advertising, consumers’ limited time and attention remain significant bottlenecks. There's a distinct danger that companies will produce more content merely for content's sake, which could dilute viewer engagement. Citing AI-generated short dramas as an example, Fong highlighted that simply producing more at lower costs does not inherently drive increased viewer interaction. This is more significant than it looks; audience fatigue could ultimately undermine revenue streams, leading to a classic case of supply outstripping demand.

Mango TV's AI-Generated Series: A Case Study

The debut of a new series, The Later Journey to the West, produced by Mango TV, exemplifies this dual dynamic well. Premiering on August 31, this AIGC-driven fantasy series has become China’s first AI-generated long-form series to air during prime time. With a basis in a classic late-Ming or early-Qing fantasy novel, the show features a fresh cast of characters on a journey to acquire Buddhist scriptures amidst comical misunderstandings.

The inaugural season aims to span 30 episodes, each lasting around 40 minutes, crafted through Mango TV’s in-house AIGC production platform, which had already supported over 3,900 projects and over 40,000 professional users by mid-2026. This initiative reflects Mango TV's ambition to harness AI for more than just efficiency; they're actively pushing the envelope of what's possible with automated content generation.

Interestingly, the platform developed 109 character assets and 143 scene assets specifically for this series, employing an innovative “produce, review, and broadcast in parallel” approach. This method allows the continuous production of later episodes even while earlier ones are airing—a model that could radically change traditional content production timelines. However, (and this is the part most people overlook) initial audience numbers have been promising; the premiere garnered the highest ratings among provincial satellite channels in its timeslot, with 27.57 million views on Mango TV by September 2.

The Outlook: Sustainability Concerns

Yet, despite these early successes, the sustainability of this model remains uncertain. Before its release, the project team navigated guidance from Hunan's broadcasting regulator to ensure they prioritized both the technical feasibility of AI-generated long-form storytelling and the commercialization roadmap for seasonal AIGC dramas. The overarching challenge mirrors insights shared by Xiong and Fong: as AI becomes more affordable and production costs decrease, the industry must pivot from focusing merely on output volume to ensuring that AI can yield adequate economic returns. Companies will need to innovate not just in producing content but in crafting viewer experiences that resonate and engage, or risk creating a cycle of unsustainable production. What this means for you as an observer or participant in this field is clear—technology alone isn’t the answer; understanding human behavior is paramount.