China's First Licensed AI Feature Film Made 1.2 Million Yuan in Four Days. Its Stock Was Bid Up for a Month First.

Sanxingdui: Future Past, a 100-minute science-fiction feature produced by Bona Film Group, opened in Chinese theaters on October 23 as the country's first AI-generated feature film to receive a public screening license from the National Film Administration, the government approval required for theatrical release. Four days in, the numbers are sobering. Cumulative box office including presales stood at about 1.216 million yuan, roughly $170,000, per Beacon data as of Monday afternoon, against an AI-projected lifetime total of about 2.609 million yuan.

The daily picture is worse than the cumulative one. Opening day brought in 301,000 yuan. On Monday the film grossed 37,100 yuan across 8,461 screenings, an average of 0.1 viewers per showing, with seat occupancy at 0.1 percent and the film holding just 2.6 percent of national screening share. Total admissions across the four-day run have reached about 41,000. For a release that carries the label "first," theaters are effectively playing it to empty rooms.

The stock market had priced in a very different outcome. Bona Film, listed in Shenzhen as 001330.SZ, hit its daily limit-up the day after the release date was announced in late September, then hit limit-up again on October 9 and October 12 as AI-video enthusiasm swept the market. On Monday, as the box office data landed, the stock fell limit-down to 5.54 yuan, valuing the company at 7.615 billion yuan, well below its year high of 13.72 yuan. Peers in the AI-drama content trade fell with it: Ciwen Media also hit limit-down, while Huazhi Digital Media, Guomai Culture, COL Digital and Zhangyue Technology all declined.

The production itself was neither cheap nor quick. Bona says the film took more than two years, was generated on its in-house Boka Film Cloud Set with AI support from Vidu developer Shengshu Technology, consumed more than 1.2 million base images and went through more than a hundred full-film iterations, with 17 professional craft roles working in a virtual production environment. Producer Jiang Defu, Bona's chief executive, says the film cost tens of millions of yuan, against at least 300 million yuan for an equivalent live-action production, noting that a single effects shot used to cost several million and was often sent abroad. Every character is an original digital creation with no digital replicas of real actors, and the film carries an on-screen notice that its visuals and part of its audio are AI-generated.

For Bona, the AI slate is a survival narrative as much as a bet. The company's first-half 2026 revenue fell 31.9 percent year over year to 458 million yuan, with a net loss of 169 million yuan; it has lost money every year since relisting in 2022, with net losses of 76 million, 553 million, 867 million and 1.464 billion yuan from 2022 through 2025. The film's premise, an archaeologist decoding binary messages hidden in Sanxingdui bronze artifacts as a superintelligence envelops the world, mirrors the pitch the company has been making to investors: AI collapses the cost of spectacle that traditional production can no longer fund.

The wider industry is still moving toward the screen even as this release stumbles. February's Reunion Order was billed as China's first full-process AIGC animated film, and in late August the AI-produced series Hou Xiyouji became the first long-form AI drama to run on satellite-TV prime time, topping its slot on Hunan TV. ByteDance's Volcano Engine started enterprise trials of its Seedance 2.5 4K model on October 22, supplying twenty studios and directors, as video-model makers chase theatrical-grade output and professional workflows.

Early audiences, for what it is worth, have not panned the film: Chinese media have quoted viewers praising character consistency and visuals that exceeded expectations. The problem is that almost nobody bought a ticket to check. The lesson landing on AI-content stocks this week is an old one wearing new hardware: a technical first is a news cycle, not a business model, and capital that bid up the "first licensed AI film" before opening day had not priced in that audiences still choose stories, not production methods.