Is Half of China’s AI Making Slop?

Zilan Qian is a research scholar at GovAI researching AI development and regulation in China. On October 21, the People’s Daily published a commentary titled “Turning ‘Gunpowder’ into ‘Fireworks’? Beware the Claim That China’s AI Has Gone Astray” (把”火药”做成”烟花”？警惕”中国AI跑偏论”). The same day, Nanfang Commentary (南方评论), the op-ed column of Southern Daily , asked : “Is Half of China’s Compute Being Spent on Feel-Good Dramas (爽剧)? Don’t Judge AI by Its Tokens.” Three days earlier, Beijing News (新京报) had questioned whether “‘American AI Boosts Productivity While Chinese AI Makes Tittytainment’ Is a False Proposition.” All three state-affiliated outlets were responding to the same viral claim: that 55% of China’s AI tokens were being used for short dramas and video generation. The claim struck a nerve because AI-generated entertainment in China is both everywhere and widely despised. According to China’s broadcasting regulator, more than 90% of the roughly 430,000 micro-dramas released in the first eight months of 2026 were AI productions. Yet Chinese media and public figures routinely call them wunao (无脑, braindead), cuzhi lanzao (粗制滥造, shoddily made), and zhili dixia (智力低下, intellectually deficient). 1 Why, in a society that places such a premium on productivity, has an industry devoted to manufacturing seemingly useless pleasure become so enormously popular? And what should we make of the shame attached to it? I argue that both answers lie in Chinese society’s obsession with usefulness, where technologies and people alike are judged by their productive value, so much so that resting can feel like a personal failing. AI slop offers an escape from that pressure, and the shame is the same pressure catching up with people even in their escape. Its shoddy production deserves criticism, but the pleasure people take in it should not have to justify itself by being useful. “America uses gunpowder to make weapons; China uses it to make fireworks.” In June 2026, in a meeting with investors, Chinese digital-content company COL Group 中文在线 claimed that AI short dramas and video generation accounted for 55% of China’s AI token consumption, compared with 15% for software development. The figures were picked up in October and circulated widely under the headline “China uses half of its compute to create AI slop.” Alongside this widely circulated claim is an alleged comparison showing the United States using 58% of its tokens for coding, research and enterprise automation. 2 The comparison prompted some Chinese commentators to revive a familiar analogy: the West uses gunpowder to make weapons, while China uses it to make fireworks. The underlying message was that the US had harnessed technology for gains in productivity and national power, while China had squandered its technological potential on entertainment. ChinaTalk is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber. The analogy comes from the Chinese writer Lu Xun, who wrote in his essay “The Benefits and Harms of Electricity” (电的利弊)” in 1933: “Foreigners use gunpowder to make bullets to defend against enemies, while China uses it to make firecrackers to worship the gods.” Writing at a time when China was struggling with foreign aggression and the legacy of its technological decline, Lu Xun invoked a familiar anxiety about China’s failure to turn its own inventions into sources of national strength. Nearly a century later, the same metaphor has resurfaced in discussions of AI, suggesting that the fear of falling behind technologically remains deeply resonant. The controversy became so intense that ByteDance, maker of the top video-generation model Seedance 2.5, publicly challenged the figures, saying that video generation accounted for only around 10% of its own computing capacity. 3 The People’s Daily commentary also rejected the simplistic comparison and enumerated China’s many applications of AI in scientific research and industrial production. How accurate is the 55% claim? According to Chinese business media Caixin, the claim comes from a roadshow and conference call held on June 3 by COL Group (中文在线), an A-share listed digital-content company. The company’s investor relations record states that “among all token consumption in China, AI short dramas and video generation rank first, accounting for as much as 55%; followed by e-commerce and marketing (such as livestream sales and advertising) at 24%; while software development, once highly anticipated due to the popularity of vibe coding, accounts for only 15%, despite the continued advances in coding capabilities of models like Claude.” 4 Caixin also claimed that the number is consistent with the Token Economy Development Research Report (2026) , published on September 30 by the China Academy of Information and Communications Technology (CAICT), China’s state-backed think tank, which states that “AI short dramas account for 55% of token consumption across the entire industry.” However, the report can no longer be found on CAICT’s website. Video generation is extremely token-heavy. On ByteDance’s own Seedance model, a five-second 720p clip costs around 100,000 tokens, roughly the length of 200 chatbot answers. Huang Ying, vice president of the token business group at Runjian (润建股份), which operates “token factories,” told World Journal on October 19 that, from her perspective, more than 50% of tokens are indeed used for video generation. On the one hand, people using generative AI for content are more active; on the other, image and video generation burn far more compute than text. Although some tokens go to R&D, R&D is essentially text, so it falls far behind video and images. A rough calculation suggests released AI micro-dramas alone are unlikely to explain the 55% figure. According to the broadcasting regulator, 430,000 micro-dramas were released between January and August 2026, more than 90% involving AI production. Assuming, generously, that all 430,000 were entirely AI-generated and each contained 200 minutes of finished footage (20 episodes with 10 mins each), producing them at 1080p and 24 frames per second would require roughly 251 trillion video tokens. Even allowing for ten times as much generated footage as the final output to account for discarded takes and repeated attempts, the total is only 2,500 trillion tokens. China’s reported March rate of 140 trillion tokens daily would imply about 34,000 trillion over eight months, so AI micro-short dramas would account for 7.4%. There are some uncertainties in actual production practices. Chinese creators describe repeatedly generating the same shot, sometimes dozens of times, before obtaining satisfactory results (a practice called “card-drawing” 抽卡). Yet reaching 55% from released micro-dramas alone would require roughly 75 times as much footage generated as retained—and considerably more if typical productions are shorter than assumed. The 55% figure likely combines video generation specifically for short micro-drama with other usages. Although the original statistics lacked reliable evidence, the debate revealed an underlying anxiety about the proper use of AI. In this discourse, fireworks are inferior to weapons, entertainment to productivity, and using AI to gen