The industry remains in a capital-intensive phase. On May 12, Kuaishou-W (1024.HK) announced that, to better leverage external financial resources, its board is evaluating a proposed restructuring plan for its Kling AI-related assets and business. This plan may involve introducing external financing. The proposal is still in its preliminary stages, and the company has not entered into any definitive agreements, with no guarantee the plan will proceed.
By midday, Kuaishou's stock rose 3.49% to HK$53.4, giving it a total market capitalization of approximately HK$232.1 billion, or about $29.6 billion.
Previous reports indicated Kuaishou plans to spin off its video model business, Kling AI, with an estimated IPO targeted for next year. The company has reportedly engaged in talks with potential investors for a Pre-IPO financing round, with an overall valuation nearing $20 billion, representing about 67.6% of Kuaishou's total market cap.
The video large model sector has developed over several years but, constrained by high costs, has not been the most capital-attractive direction in AI. However, this year, driven by factors such as the demand for world model implementation and the commercial explosion of AI short dramas and comic adaptations, the sector has accelerated notably.
According to DataEye Research Institute estimates, China's comic drama market is projected to reach 16.8 billion yuan in 2025, with further growth to 24.4 billion yuan expected in 2026. The primary growth driver is the AI-driven explosion in supply-side production capacity, followed by players from the live-action short drama industry chain replicating their business models in the comic drama sector.
Simultaneously, demand for AI video from creative professionals, filmmakers, and marketers is surging. Over the past year, companies like Google, Meta Platforms Inc., and Elon Musk's xAI have all integrated text-to-video and image-to-video capabilities into their respective chatbots.
From a perspective of differences between China and overseas, CITIC Securities research analysis suggests that overseas providers tend to target professional B-end markets like film/TV and advertising, offering capabilities through independent platforms, SaaS, and APIs, primarily via subscriptions and enterprise services. Domestic providers, however, have first embedded themselves into high-frequency, mature scenarios such as short videos, content communities, e-commerce marketing, and short/comic dramas. They drive model iteration and commercial monetization through platform distribution, data feedback, and closed-loop content consumption, giving them stronger implementation advantages.
Furthermore, video large models are seen as the core data source and capability cornerstone for building world models. They enable AI to progress from understanding static images to comprehending space-time, physics, motion, and causality, directly supporting world models in simulating, predicting, and inferring the real world.
Xie Siyuan, Managing Director of Yijing Capital, stated that world models, to some extent, connect the relatively abundant data resources on the internet side with industry demands that are extremely data-scarce, such as embodied intelligence, including areas like autonomous driving. Therefore, from the perspective of matching supply and demand, the logic for this direction is relatively clear.
Driven by these factors, competition in the video model industry has intensified, pushing up computing power costs for various platforms. Xie Siyuan noted that top video models are gradually converging in basic generation capabilities. At this stage, the core of competition shifts to capital and resource investment, including computing power, data, and continuous iteration capabilities. Platform companies with advantages in massive video data and computing power still hold a relative edge. In the short term, breaking through the current situation remains strongly correlated with resource investment, and the industry is still in a capital-intensive driven phase.
This is reflected in the market. Earlier reports suggested ByteDance plans to increase its 2026 capital expenditure by 25% to over $30 billion, citing investments in artificial intelligence and rising memory costs. ByteDance has not officially confirmed this figure. Additionally, other video model companies like Shengshu Technology and Aishu Technology have successively disclosed financing activities.
Against this backdrop, Kuaishou's choice to spin off Kling AI for independent financing represents a more direct and efficient path for capital breakthrough. It can alleviate the computing power and financial pressure on the parent company, Kuaishou, while independently accessing external resources to secure more flexible ammunition supply in the capital-intensive competition.
Comments