国内AI赛道竞争转向 降本与商业化成核心方向

This article outlines a key shift in China's AI industry, where cost reduction and commercialization have become the core priorities of competition. Core takeaways for general readers include:

1. The industry's current focus centers on whether AI cost reduction and application adoption can translate into tangible commercial value. UBS Securities analysts have identified three key priorities for China's large model sector: model capability, token input-output ratio, and commercialization.

2. The sector has shifted from maximizing token usage for AI inference to optimizing token efficiency. As corporate AI costs continue to rise, buyers increasingly prioritize the balance between performance and price. Domestic open-source models, with improved capabilities and lower costs, are particularly well-suited for repetitive or low-risk tasks.

3. Commercial use cases have already emerged in China: Mango TV produced *Journey to the West Afterstory*, the first AIGC-produced long-form drama to air on a satellite TV primetime slot. It ranked first in viewership ratings in its time slot during its premiere, with over 27 million views, though it remains unclear whether the project can develop a sustainable business model.

The recent shift in China's AI sector offers valuable insights for brands looking to integrate AI and align with consumer trends. Key takeaways include:

1. In terms of industry trends, AI development has shifted from prioritizing output scale to focusing on tangible commercial value. AI cost reduction does not automatically translate into higher revenue. Today, both corporate AI buyers and brands building AI capabilities prioritize the balance between performance and price. Data shows that R&D costs for leading Chinese large models are less than one-tenth of comparable overseas products, and average API pricing is only 10% to 20% of international competitors, creating significant cost advantages.

2. In content marketing, AI does lower content production costs and improves the efficiency of advertising and recommendation systems. However, growth in consumer time and attention has nearly stalled. Producing more content at lower cost does not guarantee capturing more user engagement, so brands should not rely solely on AI to scale content volume and must focus on converting user engagement into value.

3. Brands can draw on Mango TV's implementation experience: the company has built a mature in-house AIGC production platform serving more than 40,000 professional users and supporting over 3,900 projects, validating that AI can improve content production efficiency. Brands can adapt this approach to deploy their own AI applications.

The new shift in China's AI sector brings new opportunities and risk alerts for sellers. Key takeaways are as follows:

1. On risk alert: the industry previously encouraged higher AI usage and maximum token volume, but has now shifted to token optimization. As corporate AI bills continue to rise and measuring return on investment grows increasingly difficult, blindly scaling AI investment does not guarantee corresponding returns. Sellers should control AI investment costs and avoid pursuing scale for scale's sake.

2. On opportunities: domestic open-source large models have continued to improve in capability while costing far less than overseas alternatives. R&D costs for leading Chinese large models are less than one-tenth of overseas peers, and API pricing is only 10% to 20% of international competitors. Sellers building AI-related businesses should prioritize domestic open-source models, which offer higher cost-effectiveness for repetitive, low-risk work requirements.

3. Sellers focused on content-related businesses should note that growth in user traffic and engagement time has nearly stalled. AI-driven cost reduction and efficiency improvement do not automatically equal higher revenue. Regulators are currently encouraging exploration of commercialization paths for AI applications. Sellers need to prioritize exploring sustainable monetization models on top of cost reduction to capture new growth opportunities.

China's AI sector's shift toward cost reduction and commercialization offers key insights for factories advancing digital transformation and capturing new business opportunities. Core takeaways include:

1. For product manufacturing and digitalization needs: businesses are now adopting AI more rationally, abandoning the previous blind pursuit of larger model parameter counts and maximum inference volume, and instead prioritizing the balance between performance and price. Factories do not need to blindly follow trends to adopt high-end large models. Domestic open-source models, which have growing capabilities and far lower costs than overseas alternatives, can fully meet the needs of many repetitive, low-risk digital use cases for factories, such as design assistance, production quality inspection, and customer service responses.

2. For business opportunities: AI cost reduction does not automatically equal higher returns. When advancing AI-powered digital transformation, factories should first calculate return on investment, prioritize implementation in use cases that deliver direct cost reduction and efficiency improvement, and avoid blindly scaling investment across the board.

3. For factories in cultural creativity, film and television production, regulators are currently encouraging exploration of commercialization paths for AI-generated content. Mango TV already has mature implementation experience with an AIGC production platform, which has validated AI's ability to improve production efficiency. Relevant factories can follow this lead to explore AI transformation paths suited to their own operations and capture new business opportunities.

The latest changes in the AI industry help service providers clarify industry trends, pinpoint customer pain points, and optimize their solutions. Key takeaways are summarized below:

1. In terms of industry trends: China's AI sector has moved beyond the stage of competing on model scale and inference volume, and now competes on cost reduction capability and commercial implementation. Customer demand has shifted from "can we use AI" to "can we afford AI, use it effectively, and generate tangible returns from it."

2. In terms of customer pain points: many businesses previously expanded AI investment blindly, leading to inflated AI bills and difficulty measuring the economic value generated by investment. Today, buyers are more cautious in their AI adoption, prioritize the balance between performance and price, and demand for high-cost large models has cooled significantly.

3. In terms of solution directions: service providers can address customer pain points by prioritizing domestic open-source large models, which have continuously improving capabilities, far lower costs than overseas alternatives, and are well-suited for customers' repetitive, low-risk work requirements. Providers should also help customers optimize token investment and clearly track return on investment. For content production clients, providers can reference Mango TV's mature AIGC production platform model to offer ready-to-implement cost reduction and efficiency improvement solutions.

The AI sector's shift toward cost reduction and commercialization offers many insights for platform operation management and risk mitigation. Core takeaways are as follows:

1. On the core challenges platforms face today: growth in mobile user traffic and engagement time for leading Chinese internet platforms has nearly stalled. While AI can lower content production costs and enable more output, this does not automatically translate to capturing more user engagement. Platforms should not rely solely on AI to scale content volume, and must prioritize exploring commercial implementation paths for AI applications.

2. On operational adjustments for large model platforms: the industry has shifted to token optimization, and customers now prioritize cost-performance. Platforms need to adjust their strategic focus: instead of competing on maximum token inference volume, they should shift to optimizing token input-output ratio, highlight the cost advantage and improving capabilities of domestic open-source models to attract price-sensitive customers, and can proactively source and onboard relevant AI application projects.

3. On risk mitigation: platforms should not overstate the revenue gains from AI-driven cost reduction, and must recognize that lower costs do not automatically equal higher revenue. While Mango TV's AIGC long-form drama project has validated that AI can improve production efficiency, its sustainable business model remains unclear. Platforms should guide onboarded merchants to explore commercialization alongside implementation, rather than stopping at cost reduction and efficiency improvement, to mitigate the risk of spending without returns.

This article outlines the latest developments in China's AI industry, offering high reference value for research on AI industry development, policy orientation, and business models. Core insights are as follows:

1. On new industry developments: China's AI sector has completed a phased shift, moving from past priorities of expanding model parameter counts and maximizing token inference volume to focusing on cost reduction, efficiency improvement, and commercial implementation. The entire industry now centers on the core question of whether AI applications can generate sufficient tangible commercial value, and large model competition currently focuses on three core areas: model capability, token input-output ratio, and commercialization.

2. On policy and regulatory developments: regulators have begun guiding AI applications to explore viable technical and commercialization paths. For example, Hunan Radio and Television required Mango TV's AI drama project to explore both technical and commercial tracks, providing a clear case study for research on AI industry policy orientation and regulatory direction.

3. On business model research: AI cost reduction has already validated efficiency gains in multiple sectors including large model services and content production. However, the sustainable business model for new AI applications such as AIGC-produced long-form drama remains unclear. At the same time, Chinese large models hold far stronger cost advantages than overseas alternatives, and it remains an open question whether this localization advantage will spawn entirely new business models, making these valuable directions for further in-depth research.