China’s AI-video race is moving into a more industrial phase. Cities and industrial parks are competing to attract filmmakers, while Shanghai is building policy support around computing access, data, public services, talent, copyright help, and overseas distribution for AI-powered micro dramas. Reuters reported that local offers to AI filmmakers include subsidies, subsidized rent, living allowances, computing support, and technical assistance.
The competition is not only about which company releases the most capable video model. It is about where creators work, how they access computing resources, who funds early production, and whether a short-form project can find an audience or an overseas channel. That is a different problem from a model demo. It requires an ecosystem of studios, platforms, production tools, distribution partners, and policy support.
Shanghai’s policy is unusually explicit about that broader infrastructure. A Shanghai Municipal Government page identifies Xuhui, Yangpu, and Minhang as locations for AI micro-drama clusters. The stated measures include support for renting intelligent computing capacity, using third-party large-model APIs on cloud platforms, purchasing outside corpora for vertical-model training and agent development, and building a service center for filming coordination, marketing, data analysis, copyright protection, financing, and technology.
The policy does not prove commercial success. It does show that a major Chinese city sees AI-generated content as an industry that needs operating support, not merely a cultural trend. EastFrontier previously tracked AI’s reshaping of China’s micro-drama industry for U.S. audiences. The newer development is the effort to build physical and institutional clusters around that production wave.
Production costs are falling faster than audience attention spreads
Reuters reported that 221,900 new AI shows appeared on Douyin in the first half of 2026. Only 1,055 of those drew more than 100 million views, according to DataEye figures cited by Reuters. The two numbers belong together. They describe a surge in output, but also an attention market in which a small fraction of projects clears a very high visibility threshold.
That imbalance is one reason municipal support does not automatically solve the business problem. Easier generation can lower the cost of producing material, but it can also create more competing titles. Distribution, marketing, character design, script structure, and viewer retention remain scarce. A low production cost may make it easier to experiment; it does not guarantee that an audience will watch.
Reuters reported that CCTV put the cost of producing AI short dramas at about 5,000 yuan per minute early in 2026 and only a few hundred yuan per minute in the first half. It also cited a filmmaker’s comparison of a conventional wedding scene costing 60,000 yuan versus 1,400 yuan with AI. The larger cost trend is reported by Reuters and CCTV, while the scene example is a filmmaker’s claim. Neither should be converted into a universal industry price.
Caixin Global’s June report provides a separate signal of rapid adoption. Its accessible summary said ByteDance’s Seedance 2.0 launched in February 2026 and that more than 95% of 128,000 new micro dramas in the first quarter were AI-made. Caixin also summarized reported production costs below 20,000 yuan for some 100-minute AI dramas. Those figures refer to a different period and measurement from Reuters’ first-half Douyin data. They should corroborate momentum, not be merged into one statistic.
Shanghai links model access to services and distribution
Shanghai’s policy matters because it addresses the costs around the model. A studio may need access to computing power and APIs, but it may also need data, filming coordination, copyright-protection support, marketing, and a route to foreign markets. The government page says the city will support these elements and establish a one-stop service center covering consultation, filming coordination, marketing, data analysis, copyright protection, financing, and technology.
The approach looks like cluster policy applied to a content-production sector. Shanghai’s plan lists technology applications, content creation, platform incubation, and overseas services as cluster areas. It also says it will use events such as the Shanghai TV Festival, Hong Kong International Film & TV Market, MIPCOM Cannes, and the Asia TV Forum & Market to help projects reach international markets, while building an overseas-distribution base in Pudong.
These are announced support measures, not evidence that projects have already secured export contracts or earned revenue abroad. The policy does, however, provide a more detailed answer to the recurring question of what happens after a video model becomes capable enough to make short scenes. Someone must still organize the project, distribute it, track performance, handle rights, and find a market.
Reuters’ reporting connects this policy logic with local competition. It described Chinese cities and industrial parks trying to attract AI creators. It also reported that director Cao Yiwen premiered an AI-animated film at the World AI Film Festival in Cannes in April. That kind of festival appearance can create visibility, but it is not a substitute for a repeatable production and distribution system.
The film push carries regulatory and rights questions
The speed of production has brought legal and quality questions into view. Reuters reported that China requires explicit labeling of AI-generated content but has not introduced clear copyright rules for AI work. That creates uncertainty for creators, platforms, and investors. A studio may use a model to generate an image, voice, or sequence, yet still need clarity about training data, authorship, licensing, and potential claims from rights holders.
The policy support in Shanghai includes copyright protection as a service area, which signals that local officials recognize the issue. It does not resolve it. The municipal page describes a support mechanism, not a settled legal framework. A production cluster can make it easier to access support, but the policy does not establish that a particular work will be cleared for distribution in every market.
There is a comparable commercial tension in China’s earlier AI-video coverage. ByteDance’s Seedance 2.0 push into AI video production focused on the capabilities of a model. The current story shifts attention to the surrounding production economy: as models lower the barrier to making a clip, cities are trying to create the conditions for companies to make, sell, and export many clips.
That strategy has risks. Government incentives can attract studios before a stable market forms. Low production costs can create a flood of content. A small number of hits may capture attention while many projects never reach a large audience. Reuters’ Douyin numbers illustrate that concentration. Shanghai’s plan may reduce some operating friction, but it cannot guarantee creative success or international demand.
The direction is nevertheless clear. China’s AI-film push is no longer defined only by model releases and viral clips. Municipal policy, shared services, computing support, and overseas-distribution plans are becoming part of the story. The next test is whether those clusters produce durable businesses with rights management, audience demand, and repeatable revenues, rather than a large volume of inexpensive content competing for the same limited attention.