At the iQIYI World Conference in April 2026, Gong Yu stood on stage and proclaimed, "Live-action or Heritage in the Making."

A large screen on the stage showcased the "AI Artist Pool," where virtual humans tirelessly changed expressions and outfits. Gong Yu announced that iQIYI had signed contracts with over 100 AI artists. Through the use of AI, the production time would be significantly reduced, and post-production work could be completed using digital avatars. He also introduced the commercial launch of "NaDou Pro," an AIGC film and television production platform aimed at small and medium-sized creators, providing AI tools, a new revenue sharing model, and an unlimited revenue promise.

This event was presented as a technological extravaganza, with music, lighting, and the presentation all supporting the narrative: the future is here, embrace it.

However, outside of this grand event, on another large screen was a number: $1.41.

This was iQIYI's closing price on the NASDAQ screen, with a market capitalization of only $1.361 billion, precariously close to the delisting threshold. When the company went public in 2018, its stock price reached a peak of $46, achieving a peak market cap of over $31.2 billion. From $31.2 billion to $1.3 billion, a 97% drop in less than eight years.

While the outside world debates whether AI can act well, the capital market is only concerned with whether AI can tell a good story. Over the past decade, iQIYI has burned through hundreds of billions of funds but has yet to find a sustainable path to profitability.

In 2025, the content costs of this streaming giant still amounted to $15.45 billion, accounting for 56% of the total annual revenue; a full-year net loss of $206.3 million, moving from profit to loss; operating profit was only $640 million, plummeting by 73% compared to $2.36 billion in 2024. Just two years earlier in 2023, iQIYI's net profit reached a peak of $1.93 billion since the company's inception.

From peak profit to loss, it only took two years.

Even more challenging is that iQIYI's free cash flow has plummeted from over $3.3 billion in 2023 to less than $10 million, showing a rapid deterioration in its ability to generate internal funds.

The company issued convertible bonds twice in 2023 and 2025, totaling nearly $1 billion, with the interest burden continuing to increase. After the U.S. stock financing function essentially failed, iQIYI secretly submitted its listing application to the Hong Kong Stock Exchange at the end of March 2026 and rolled out the AI Artist Pool and "NaDou Pro" in April.

A company seeking a Hong Kong stock market "blood transfusion" needs a new and eye-catching narrative to impress investors. Phrases like "Marginal Cost Trending to Zero" and "AI Cost Reduction and Efficiency Enhancement" are more valuable in today's capital market context than any blockbuster TV series.

This is not about art, but merely about preserving the shell.

But when a company no longer relies on content quality but on "cost reduction" to please the capital, who will be the first batch of fuel under this cost-cutting machine?

Hengdian

The public opinion focus of this AI storm was soon taken over by several legal letters.

Zhang Ruoyun was the first to speak out, stating that he had never authorized any platform to create an AI avatar of himself and reserved the right to pursue legal responsibility. Yu Hewei, Chen Xiao, Wang Churan, and others followed suit, successively debunking the rumors. Yiyang Qianxi Studio released a statement pointing out that multiple AI short films had unauthorizedly used his image and voice, with the controversial shorts garnering nearly 75 million views. Tan Jianci, Gong Jun, Yang Zi, and other actors also spoke out one after another to defend their rights.

For a time, the "AI Celebrity Database" became a hot potato, and iQIYI was forced to come out to explain that the AI celebrities in the database were all "authorized AI celebrities" and not digital replicas of real stars. "Artists joining the Nado Pro artist database only represent an intention to contact AI film and television projects." Gong Yu also responded multiple times publicly, emphasizing that the platform respects the rights of actors and that AI is a tool rather than a substitute.

A top celebrity's statement can trend, but in the places where the camera cannot reach, there are also 134,000 extras in Hengdian.

In recent years, with the explosion of the short film market, the demand for extras in Hengdian has increased. With many short film crews, short production cycles, and fast-paced filming, there is a high demand for extras. In a long drama, extras are just part of the background with no lines or names; in short films, they have the opportunity to show their faces, speak a few lines, and even become main characters. The daily wage for an ordinary extra ranges from 150 to 300 yuan, and if they work hard, they can earn four to five thousand yuan a month. Not much, but enough to get by.

After the arrival of AI, this "enough to get by" has also begun to waver.

In the first quarter of 2026, over 85% of Hengdian's short film crews had fully switched to using AI actors to play supporting roles and extras, causing the full-time employment rate of human short film actors to plummet from 82% in 2025 to 29%. Some crews have stopped recruiting extras altogether, only needing to film the main characters in front of a green screen, leaving the background crowd entirely to be generated with the click of a button by an algorithm. AI-generated extras do not need to eat, rest, be late, demand a raise, and will never complain.

Hengdian has a registered group of about 134,000 extras, but can only provide around 800 effective positions per day. The competition is more intense than the civil service exam. Some actors used to maintain a rhythm of filming for over 20 days a month, shooting hundreds of short films. However, in the month following the Spring Festival, they didn't film a single day. After this year's Spring Festival, the production volume of live-action short films was directly cut by 50%.

While top stars defend their image rights on the hot search lists, small extras have disappeared under the onslaught of AI. Top stars have talent agencies, lawyers, and Weibo hot searches. Their rights protection can become news, generate public opinion pressure, and force platforms to compromise. But extras have nothing. Their disappearance will not be news; it will only be seen as the increasingly shorter queue outside a certain crew's door in Hengdian.

The anthem of technological revolution is always composed with the sighs of the grassroots workers. But in this cost reduction game, what role is the platform playing? Is it really driving a technological revolution, or is it using a beautiful name to shed its heaviest burden?

From "Landlord" to "Fee Extractor"

iQIYI announced at the conference that it would transition to "decentralized social media," launch "Nadou Pro," and modify its revenue-sharing rules.

The positioning of "Nadou Pro" is a content platform for small and medium-sized creators, providing AI creation tools, supporting users to produce short dramas, and earning revenue through revenue sharing. Gong Yu said this is about "returning content creation power to the public" and is a "decentralized" content revolution. The new revenue-sharing rules promise creators "uncapped" revenue sharing, and AIGC content and Chinese drama content will receive additional subsidy incentives.

It sounds like empowering creators and giving the stage to the public. But it's just what it sounds like.

In the past, long video platforms were the "landlords" who spent money to buy dramas. At its peak, iQIYI's annual content procurement and production expenditure exceeded 20 billion yuan, with investments of billions in an S-level blockbuster, and the platform bore the financial responsibility. The logic of this model is: I spend money, I buy content, I attract users to subscribe, and I earn membership fees. By 2023, iQIYI's membership service revenue reached 20.3 billion yuan, and its membership exceeded 120 million, marking the company's peak.

However, this logic has a fatal flaw: it assumes that the platform can consistently judge what is good content, and good content can continuously attract enough subscription users.

iQIYI has been losing money for ten years. The number of members has been declining from the peak of 120 million, with membership service revenue dropping to 16.81 billion yuan in 2025, a year-on-year decrease of around 5%. What's more troubling is that one of iQIYI's cost reduction measures is to vigorously promote a movie revenue-sharing model, where the platform no longer pays upfront licensing fees but shares revenue with partners based on actual views. While this has indeed reduced upfront cash outflows, the side effect is that the partners willing to accept revenue sharing mostly provide mid-level content. The truly popular IPs that can drive member payments still prefer high-price buyouts. As a result, iQIYI's content library is filled with "quantity over quality" mid-level works, lacking blockbusters. Users have noticed, and the platform's stickiness and willingness to pay have also declined.

This model has reached its limit.

Now, iQIYI doesn't want to operate at a loss; it wants to make a stable profit. The platform has provided AI tools to numerous individual creators, allowing them to create content and try different approaches. For those who become popular, the platform takes a large cut; for those who fail, the platform bears no impact. Similar to Uber not employing drivers and Airbnb not owning real estate, iQIYI no longer wants to invest in exclusive content.

This logic in the internet industry has a more academic term: "platform capitalism." Platforms provide infrastructure, attract a large supply-side, and then shift the risk to the suppliers through algorithms and fee structures, keeping the profits for themselves. YouTube and TikTok are successful examples of this logic. However, the success of these platforms is built on their position as traffic entry points with a large existing user base.

Can iQIYI replicate this logic? That is a doubtful question. iQIYI's monthly active users are not on the same scale as TikTok, and its content consumption scenario is more about "sitting down to watch a show" rather than "scrolling through content." When transforming a subscription-based platform into a UGC content platform, what will make users stay?

But more importantly, when long-video platforms begin to act as "water extractors," transferring all content production risks to individual creators, what does this mean for the entire content industry? What are those short-video platforms, which originally thrived on quick, engaging content, doing?

Two Approaches

While iQIYI is betting on AI, Douyin (TikTok) is investing 500 million RMB to support "live-action short dramas."

On April 15, 2026, at the 13th China Internet Audiovisual Conference, Douyin Group announced its latest plan to support live-action short dramas, allocating 500 million RMB to continuously promote innovation and exploration of real-life themes in short drama content. At the press conference, Red Fruit Short Drama's Chief Editor Li Le emphasized that live-action short dramas have a significant advantage in user retention, completion rates, and long-term willingness to pay.

Douyin's 500 million RMB support plan is specifically aimed at "short dramas featuring real actors," explicitly excluding AI-generated content. Douyin's rationale is that AI-generated content on the platform has become overwhelming, causing users to experience aesthetic fatigue. Real faces, genuine emotional expressions, and authentic body language have become rare commodities. Live-action short dramas contribute the vast majority of the platform's views, likes, interactions, and conversion to paid content, as AI-generated dramas significantly lag behind in user retention and completion rates.

After reaching the peak of traffic, short-video platforms face a core issue of user retention and conversion to paid services. Low-quality AI content can fill the timeline but cannot establish user stickiness or drive payments. Genuine emotional resonance is the key. Therefore, Douyin is now willing to pay a high price for "authenticity."

What was supposed to be a long video platform representing high quality and in-depth content has, due to financial despair, actively jumped into the cheapest industrial assembly line.

This misalignment is a rational choice under two different financial conditions. Backed by ByteDance, Douyin has abundant cash flow and can make long-term investments in content quality; iQIYI's cash on hand is no longer sufficient to support the long-term gamble of high-cost content, so it must find a "low-cost, high-output" path.

In contrast, Tencent Video, backed by Tencent Holdings, with the parent company's revenue reaching 557.3 billion yuan in the first three quarters of 2025, a year-on-year increase of 14%, maintains its lead in the long video market with over 340 million monthly active users and 114 million members. Mango TV, relying on Hunan Broadcasting System's content supply chain and low-cost in-house production capability, continues to maintain a stable profit. iQIYI lacks Tencent's social ecosystem support and Mango's cost advantage, making its situation the most challenging among the three long video platforms.

However, the "low-cost, high-output" path has never been a good road in the content industry.

Data shows that the proportion of AI simulated human short dramas in the Top 100 Manga and Drama Rankings has surged from 7% in 2025 to 38%. The daily newly licensed AI short dramas may number in the thousands. With a 3-person team, a few thousand yuan in computing costs, and 48 hours of work, a drama can be produced, showcasing the production efficiency of AI short dramas. However, in this race, 90% of companies are operating at a loss, over 60% of works have low viewership, and profitability is challenging. Among the tens of thousands of works released, the number of blockbusters with over a hundred million views is scarce.

Low barriers have never been a competitive advantage and will only make the race overwhelmingly crowded.

"The Jazz Singer"

On October 6, 1927, the first sound film, "The Jazz Singer," premiered in New York.

Prior to this, every movie theater had a band dedicated to accompanying silent films. Large theaters had bands of twenty to thirty people, and even small theaters had at least one pianist. They would improvise their performance based on the on-screen plot, playing low-stringed music for tragedies and lively piano tunes for comedies. This was a craft and also a stable livelihood. There were about 22,000 cinema musicians across the United States, with their own union, industry standards, and professional dignity.

After the advent of sound films, almost all these musicians were unemployed within a few years. Between 1927 and 1930, the number of American cinema musicians plummeted from about 22,000 to fewer than 5,000. Their union had once initiated a boycott, organizing musicians to refuse to play in theaters equipped with sound devices, but to no avail. The tide of technology would not stop just because some opposed it.

As early as 1906, John Philip Sousa wrote "The Threat of Mechanical Music," predicting that the phonograph and mechanical instruments would replace live musicians, turning music into an industrial product and depriving performers of their livelihood.

Today, Gong Yu's statement "Live-action heritage" echoes what film companies once said to theater musicians, seemingly conveying the same message.

Every technological revolution requires a narrative of "heritage" to complete the handover, glorifying the obsolete as "precious tradition," and then effortlessly sweeping it into museums. The term "heritage" itself is a form of appeasement, acknowledging your past existence and value, while also declaring the end of your era.

Talkies and AI actor libraries may seem similar on the surface, but they have a fundamental difference.

Talkies represent a new form of expression. They enable films to tell more complex stories, convey richer emotions, and ultimately offer more to the audience.

Theater musicians lost their jobs, but the art form of film itself became more complete and powerful.

What is an AI actor library? It is not a new form of expression; it is a cheaper replication method. It does not enhance the completeness of film and television; it only makes them more affordable. The logic of talkies is to use better technology to tell better stories, attracting more people to the cinemas. The logic of AI actor libraries is to produce more content at a lower cost, spending less money.

These are not the same.

If iQIYI's AI gamble succeeds and it wins, what about the audience?

Stimulation Replacing Empathy, the Folded Reality

The ultimate outcome of the proliferation of AI-generated films and television is not about replacing anyone but quietly altering the human mechanism for receiving stories and perceiving genuine emotions.

Neuroscientific studies suggest that when humans watch live performances, mirror neurons in the brain are activated, creating emotional resonance. To a certain extent, we can feel the emotions the actors are experiencing. This is an ability developed by humans over tens of thousands of years, forming the basis for understanding others and establishing social connections. This is also why a good movie or series can linger in one's mind long after it ends, even altering a person's perspective on certain things.

AI-generated performances can achieve a high level of visual realism at a technical level, but they lack real-life experiences. When an AI-generated character is crying, it is because the algorithm has determined that this scene requires tears, not because the character has genuinely experienced something. The human brain can perceive this difference to some extent, which is why the completion rate and conversion rate of paid content generated by AI are much lower than those of live-action content, even if the visuals may be more refined and the plot more "exciting."

However, this perceptual ability is a case of "use it or lose it."

When a generation grows up accustomed to AI-fed absolute smoothness and high-frequency stimulation, their brains will gradually adapt to this stimulation pattern, and the threshold for perceiving complex, subtle, and nuanced real emotions will become higher. Just like people who have been eating heavily seasoned food for a long time, they will gradually lose the taste for bland food.

McLuhan said, "The medium is the massage," meaning that the medium is not just a channel for delivering content; it is shaping the way we perceive the world. When the logic of content production shifts from "telling a true story" to "calculating the stimulation the audience wants using an algorithm," the medium ceases to be a massage and becomes an anesthetic.

This is not a eulogy for "live-action filming," nor is it to say that AI is inherently bad. Technology itself is neither good nor evil; the issue lies in how it is used and under what kind of commercial logic it is widely promoted. A company that resorts to AI out of financial desperation and a true believer that AI can enhance artistry may use the same technology, but with completely different motivations, leading to vastly different outcomes.

Gong Yu said that AI can enable an actor to take on four projects a year instead of two. The implication of this statement is that an actor's value is not based on what they experience on set but on how many times their face is rendered.

But real acting has never been about the face. It is about a person at a specific moment, making an unpredictable choice due to some genuine experience. This is something AI cannot calculate.

However, when the audience is already accustomed to algorithmic feeding, will they still crave the unquantifiable?