Bona Film's "Sanxingdui: Future Past" Struggles at the Box Office: Predictions and Candlestick Charts for AI Theatrical Feature Films

By Qian Shui Yu X

The much-hyped film "Sanxingdui: Future Past" has finally been released.

Calling it a hype film isn't harsh. In early September, Bona announced that the film was about to hit theaters, and its stock price hit the daily limit; on September 27, a release date was officially announced, and the next day the stock hit the limit again; on October 9 and 12, there were two more limit-ups. A film no one had seen yet had already played out its climax on the candlestick chart. Exploding before it even aired—even the entertainment industry's would-be breakout kings would have to call it a senior.

Then the wolf came, and it really was released.

In four days, the total box office was just over 1.2 million yuan, with an average of two people per screening on the 24th and 25th. The producers say the cost was in the tens of millions, while Beacon's final box office forecast is only 2.6 million yuan. At this attendance rate, in many auditoriums the real people sitting there are probably fewer than the digital people on the screen.

(Image source: Weibo @Cailian Press)

There are bad films every year, and ordinarily this wouldn't warrant a dedicated article. But that round of predictions this spring and the current situation are worth a callback.

In March, Zhang Xiaobei answered a question on Zhihu, saying that at the latest next year, and at the earliest the second half of this year, at least one 90-minute AI film in China would attempt a theatrical release. At the online audiovisual conference in April, Tencent and iQIYI, longtime rivals, rarely reached a consensus on the same afternoon: this year could be the first year of AI long-form dramas. Sun Zhonghuai said the platform was using AI for the entire workflow of a 90-minute film, which might be released in the third quarter; in interviews around those days, Gong Yu gave a more specific timetable: at the earliest this summer, at the latest this autumn or winter, a purely AI-generated feature film would definitely appear, "and it will be a hit feature film."

Now it's late October, and autumn/winter has arrived. The prophecy has come half true: an AI feature film did enter theaters, with a 100-minute dragon mark in hand. What didn't come true is the word "hit"—the day the feature film arrived, what arrived was a dud.

"Sanxingdui: Future Past"

As for that predicted timetable, it has been moved at light speed to next year. Brokerage research reports have already marked 2027 as the "key year" for AI film and television, and Bona's next AI film is also set for release in the first half of 2027.

The greatest advantage of prophecy is that it costs nothing. If it comes true, you're a prophet; if not, you just change the year and keep talking—the next bus is always on the way. This craft was originally the tech industry's signature skill: roadmaps, launch events, "this time next year," using stories about the future to raise money in the present. What Little Fish didn't expect was that in the year or two since AI entered entertainment, the level of the works hasn't improved first, but this set of skills—self-prophecy and self-empowerment—has been learned first.

According to the official account, the clue to the story of "Sanxingdui" is seven carved symbols on bronze vessels; they are actually a string of binary code. In the future, a super AI sweeps the globe, humanity is driven to a dead end by algorithm dependence, and the key to breaking the deadlock is hidden in ancient Shu symbols from three thousand years ago. Colleagues who have seen it added more details (Little Fish is truly a practical person and doesn't want to spend this money): the ancient Shu king rode an elephant on a visit to Egypt and conveniently passed off the Sanxingdui mask as the Eye of Horus; there's also a wormhole, with jumps back and forth between 2045 and 2095.

"Sanxingdui: Future Past"

An AI-made film about humanity being driven to the brink by AI and saved by a code carved in advance by their ancestors. Even the plot itself is a prophecy: people three thousand years ago buried the answer and waited specifically for later generations to dig it up. Did the creators realize this layer of self-reference? Little Fish thinks most likely not.

Audience feedback, on the other hand, is easier to understand than the plot. "Too much AI feel," "plastic feel," "pseudo-human feel"—these phrases appeared repeatedly on social media. The lip movements are exaggerated, the line delivery is forced, the pores on the protagonist's face are visible, but the supporting characters' faces are blurry. One can only say that AI has, to some extent, also taken the dregs of the entertainment industry and discarded the essence, treating (AI) people according to their status. One viewer on the first day put it quite fairly: online it would count as a pass, in the cinema it's a fail.

(Image source: Xiaohongshu)

The producers themselves admit this. Chief producer Jiang Defu said after the premiere that images that look detailed enough on a phone all expose their problems once projected onto the big screen. But viewers already have plenty of this kind of image on their phones, and for free. A ticket costs twenty or thirty yuan, yet it's being benchmarked against free AI short dramas—that's called a pricing anchor error. Yu Dong once gave Bona's AI strategy a name: "top of the pyramid + base of the pyramid": the top is theatrical films, the base is one-click-generated AI short dramas. However, as the masses have tested, this wave's top and base are actually the same height.

Of course, perhaps how viewers see it was never very important in Bona's business calculations.

Looking at Bona's financial reports, in the first half of this year, revenue from the AI film and television drama business was 1.9323 million yuan, with a gross margin of 96.83%, accounting for 0.42% of total revenue. 96.83% is a number that can be written into research reports, but 0.42% is its true position in the business. And the company lost money for four consecutive years from 2022 to 2025, with cumulative losses of about 2.96 billion yuan. That's roughly half the box office of "The Battle at Lake Changjin."

"Sanxingdui: Future Past"

This isn't just Bona's game. For veteran series company Huace, the largest source of revenue in the first half of this year has already shifted from dramas to computing power. And to keep selling computing power, it plans to spend no more than 3.3 billion yuan on servers, more than its entire revenue last year. Ciwen Media's revenue in the first half of the year was 18.92 million yuan, down 90% year-on-year; it then turned around and signed a base for two thousand AI comic dramas, and in October its stock price enjoyed five consecutive limit-ups, rising more than 60% within the month.

After the painful struggles of the first three quarters, this batch of entertainment companies has probably realized one thing: making films is no longer very profitable; the ones making money are still the shovel sellers, but the ones rising the most are the storytellers. Shovels are hard to make, but they are actually better at storytelling than tech companies. Since that's the case, better to strike late and learn on the job—so-called storytelling. As for who the customers of this business are, Xiaoyu thinks it's already very clear: the audience is not the target, the stockholders are. So what is sold to stockholders doesn't need to be a good film; it only needs to be a credible prophecy.

Bona has been selling this prophecy for three years now.

"Sanxingdui: Future Past"

At the end of "Sanxingdui" there is an AI-generated behind-the-scenes clip, letting a group of nonexistent actors frolic and interact on a nonexistent set. This is almost the entirety of Bona's approach to AI: first make it look like something "has already happened," then wait for it to happen. One might call them a master of manifestation in the entertainment industry.

The specific method is also very simple: always claim a "first" first.

At the 2024 Shanghai International Film Festival, Yu Dong said that the upcoming "A Legend" would be "truly the first big-screen presentation of a digital film worldwide." The so-called digital film's core was using AI to recreate a young Jackie Chan. At the same forum, Wang Changtian directly retorted: Yu Dong might need to go back and learn more about it—this is just a software upgrade; it could be done without AI. "A Legend" ultimately grossed just over 80 million yuan, failing to beat "The Myth" from nineteen years earlier. But the list of "firsts" keeps going: first AIGC sci-fi short drama, first AI feature-length theatrical film to obtain a public screening license, first AI hyper-realistic theatrical film. "First" itself is a kind of prophecy; it offers no quality guarantee for the work, only declares a certain self-reported form of lead.

"A Legend"

Wang Changtian has also had his hot moments. In 2023 he said AI could raise the efficiency of animated films by thirty percent, and even made a poster for "To Your Island" using AI. At this year's Shanghai International Film Festival, he reviewed it himself: at the time his head was overheated, he felt he had to keep up with the trend; that poster was actually better when drawn by a person, and using AI was only to appear to keep pace with the times. But Wang Changtian has cooled down, while Yu Dong is still hot.

"To Your Island"

After claiming the firsts, the next step is to build up results for those firsts. The AI short drama "Sanxingdui: Future Revelation" launched on Douyin in 2024, and its view count rose all the way in press releases: 160 million in April this year, and by October it had become "over 200 million." It also received a "Special Mention in the Best Drama Feature Unit of the Cannes World Film Festival in France"—except that this "Cannes World Film Festival" has nothing to do with the Cannes everyone is familiar with. The film version did brush against the real Cannes, though; the press release wrote that it "appeared during the Cannes Film Festival." "During" and "appeared"—these few words are worth savoring.

Once enough results have been accumulated, the final step is to record them as assets. In 2025, Bona's AI production center was upgraded into a wholly owned subsidiary. Early this year, Yu Dong summed up the strategy as that pyramid, and also launched a product for one-click generation of AI short dramas. As early as the 2024 Shanghai International Film Festival, Yu Dong was talking about how important AI is for "digital asset accumulation"; in January this year, Qu Ji Xiaojiang, deputy general manager of Bona Film and Television Production, said it even more bluntly in an interview with Shanghai Securities News: with the accumulation and reuse of digital assets, film companies are gradually evolving into "asset companies."

That statement is more honest than all the publicity. Assets, naturally, are meant for valuation. Therefore, even if the film flops, the prophecy cannot stop. If it stops, the assets will depreciate.

To translate: obtaining the Dragon Mark, getting scheduled into theaters—once the progress bar moves forward one notch, the stock price should rise a stretch. Whether the audience buys tickets was never part of the assessment from the very beginning.

Another brokerage's tone shifted even more smoothly. In June this year, it said in a research report that the AIGC sector in 2026 was "waiting for a dark horse," with theatrical AI films represented by "Sanxingdui"; recently the wording changed to: the AI content industry has already moved from competing on quantity to competing on quality, and high-quality content companies are expected to be revalued in the second wave. This wave actually made Xiaoyu laugh: the first round's dark horse never arrived, so they're waiting for the second wave. When it comes to rhetorical skill, you still have to look at the brokerages.

"Off Plan: Guanghan Operation"

However, if films are treated merely as market trends to be rushed along, then the trouble is that market trends, when the public outcry is too low, will also fall back.

On October 22, the day before release, Bona hit the limit down once, as if the market had voted on behalf of the audience first. On the day of release it closed up 4.4 percent, which could be counted as a final show of support. On October 26, with the four-day box office laid on the table for the first time, Bona hit the limit down again, closing at 5.54 yuan, with a total market value of 7.6 billion yuan and a cumulative decline of more than 27 percent for the year. On the same day, Ciwen, which had enjoyed five consecutive limit-ups, also hit the limit down.

This type of AI film doesn't need an audience; after all, they are only prophecies. And the only audience for a prophecy is the candlestick chart. This time, even the candlestick chart got up and left.

(Image source: Xiaohongshu)

But on a train full of anxiety, no one listens to doom-saying. The people on the train don't care where it's headed; they only fear being thrown off.

If the front row stands, the back row can only stand too; and standing alone isn't enough—they have to run on their feet as well. No one can see the big screen clearly; they can only keep running, or they'll fall. Bona is simply the one that fell the loudest.

What keeps everyone running together is probably one sentence: the AGI era will eventually arrive. This sentence has now become a tacit immunity for the entire industry. The film is bad—the technology isn't there yet; the box office flops—the audience isn't ready yet; and it doesn't matter if the prophecy falls through, the next stop will always come. Anything can be put on credit first, with the bill charged to the future. As for future generations, they may be the humans three thousand years later in "Sanxingdui," digging up the pits left by their ancestors—but modern people, well, their values have always been: after me, the deluge. Besides, they can also time-travel.

When AGI will arrive, Xiaoyu doesn't know. But if the arrival of the AGI era means prophets standing everywhere, then this era might as well not come. Xiaoyu just wants to shout three times: Back off! Back off! Back off!