China’s artificial intelligence video generation sector is experiencing an unprecedented capital frenzy. Within just half a month, ByteDance’s Seedance 2.0 solidified its position as an industry leader with over 80% market share, Kuaishou’s Kling AI set a world record for single-round financing by a multimodal large-scale model company with a post-money valuation of $18 billion, and Aishi Technology closed a Series C round of nearly 3 billion yuan. The influx of billions and even tens of billions of yuan marks AI video’s transformation from the technology toy it was two years ago to the most obvious commercialization path after AI coding.
However, behind the capital frenzy, high computing costs, user aversion to the “face of AI” and the looming IPO valuation adjustment mechanism are creating great uncertainty in this final round of battle.
Capital enters final round: Kling’s $18 billion valuation approaches parent company
On July 14, Aishi Technology announced that it had completed its Series C financing, bringing the total amount raised to 2.98 billion yuan (approximately $441.1 million). The Series C+ round was led by Alibaba, with participation from over 10 institutions including Lollapalooza Capital, Ivy Capital, Huiyuan Capital, Eastern Bell Capital, Mirae Asset, Lion X Fund under the OCBC ecosystem, and BlueFocus.
This is just the tip of the iceberg in the capital extravaganza in the AI video space. Less than half a month later, on July 2nd, Kuaishou announced on the Hong Kong Stock Exchange that its video-generated large model subsidiary, Kling AI, had closed its first round of independent funding of nearly $3 billion, reaching a post-money valuation of $18 billion. This valuation amounted to nearly 76% of Kuaishou’s market capitalization and set a world record for the largest single round of funding by a large, complex model company.
In particular, Kling AI’s investor lineup is very prestigious, jointly led by CPE, Guofang Venture Capital, BlueFive, Tencent, CITIC Securities, and Zhongguancun Science City Fund, with participation from industrial capital such as Alibaba Cloud, Baidu, Hua Ace Film & Television, and Mango Industrial Investment. NetEase noted that the rare appearance of internet giants Tencent, Alibaba and Baidu on the “shared stage” indicates that Kling AI has become a rare asset with a high capital consensus.
Beyond the support of external capital, Kling’s own revenue-generating ability underpins its strong reputation. Kuaishou’s announcement revealed that Kling AI achieved revenue of approximately 1.1 billion yuan (approximately $162.8 million) in 2025. Revenue for the first quarter of 2026 exceeded 650 million yuan (approximately $96.2 million), registering more than 300% year-on-year growth. As of March 2026, the annual revenue run rate was approaching $500 million. Kuaishou founder and CEO Cheng Yixiao revealed that the rapid revenue growth is mainly driven by the dual-engine model of enterprise customer API calls on the B-end and paid user subscriptions on the P-end.
From “cash burning pit” to “productivity tool”: Seedance 2.0 rewrites industry rules
Fast forward to early 2024, OpenAI released its video generation model Sora, and its realistic effects shocked the world. However, high R&D and computing costs quickly weeded out most small and medium-sized players, and Sora itself was shut down by OpenAI in March this year, citing low user retention and obstacles to commercialization.
The reason AI video could turn into a viable business in 2026 lies in crossing the qualitative threshold from “toy” to “productivity”. Earlier this year, ByteDance released its Doubao video generation model Seedance 2.0, setting the market on fire with its ability to “generate Hollywood-level blockbusters in a minute.”
Tan Dai, President of Volcano Engine, observed a significant change. Before the launch of Seedance 2.0, weekday usage for this model was much lower than weekends, indicating that users were mainly engaged in leisure and entertainment. After the release of the new version, weekday loads and usage far exceeded weekends, proving that it has truly penetrated into production. According to NetEase, some AI video producers report that while traditional AI live-action short dramas used to cost hundreds of thousands of yuan to produce, AI can now produce hits for just tens of thousands of yuan.
ByteDance’s Seedance leverages its superior model capabilities to gain a significant lead over its competitors, capturing over 80% market share and becoming ByteDance’s first truly profitable model. Currently, China’s video generation sector has formed a “one superpower, multiple powerful players” composition, with Kuaishou Kling, Alibaba Happy Horse, Aishi Technology, and Shengshu Technology being the key players that make up the final round lineup.
Concerns behind the boom: cost pressures, visceral aversion, and countdown to IPO
Despite the frenzied influx of capital, there are concerns behind the rise of AI video.
First, there are enormous cost pressures. Compared to text large-scale models, the training and inference costs of video generation models are an order of magnitude higher, especially for 4K-level video inference, where the computing costs are very high. This means that it will be nearly impossible to achieve profitability through ‘price wars’ alone, unless pricing power can be seized through technological advantages such as Seedance 2.0.
The second is end user resistance. In short drama tracks, where AI videos are most widely applied, “AI facial uncanny valley effect” recently trended on social media. Industry insiders have pointed out that when artificial intelligence (AI) creates human dramas, the characters often use the same AI template faces across the board. If viewers are frequently shown faces that lack the characteristics of a “living human being,” they may be more likely to have an instinctive reaction of rejection, which could directly impact the business model of AI short dramas.
More importantly, high ratings are often accompanied by strict rating adjustment mechanisms. Kuaishou’s announcement revealed that Kling AI’s funding includes a five-year IPO clause. If the IPO is not completed by October 30, 2031, or if all restructuring matters cannot be completed within the agreed deadline, investors will have the right to demand that the company repurchase its shares at a simple interest rate of 8% per annum.
Furthermore, Kling’s own financial pressures cannot be ignored. According to NetEase, Kling posted a net loss of 500 million yuan (about $74 million) in 2024, which widened to 1.9 billion yuan (about $281.2 million) in 2025. Even with $3 billion in funding, the financial pressures will still be enormous unless rapid self-sustaining profitability is achieved. Market reaction to this financing was also noticeably lukewarm. Kuaisho opened higher the day after the announcement, but closed lower, with a cumulative decline of more than 10% in the following trading days.
As Wang Changhu, founder of Aishi Technology, said, more than two years ago, big technology companies and startups rushed into this space, but now only three to five remain. Securing a ticket to the final round does not guarantee victory. It simply means the beginning of an even tougher contest. For Kling and all the players, raising money is just the first step in a long march. The real test will be whether they can still have a seat at the table the day AI can go beyond producing images to understanding and reasoning about the causal laws of the physical world.
