Kling AI Valued at RMB 120 Billion, Plans Hong Kong IPO: How to Do the Math on the First Listed Video Foundation Model Stock

In the news flow in early October, Kuaishou's Kling AI appeared with unusual frequency. On October 6, multiple media outlets reported that Kling AI plans to go public in Hong Kong as early as next year, aiming to raise at least US$1 billion, with CICC, Goldman Sachs, and UBS already brought in to prepare for underwriting. On the day the news broke, Kuaishou's Hong Kong shares rose more than 3% in early trading.

For a business that only started in 2023 and broke through with a single video generation model, this pace is not slow. If you read the news only as “a video foundation model is going public,” you will miss the more interesting part: what justifies a valuation in the hundreds of billions, and whether that math can hold up in the market.

How the Valuation Was Stacked Up

First, the numbers. In July this year, Kuaishou disclosed that Beijing Kling completed a restructuring, taking over all of Kling AI's business assets and starting independent fundraising. This capital increase had a ceiling of about RMB 20.447 billion, equivalent to about US$3 billion, with a post-money valuation of about US$18 billion. Participants included industrial capital such as Alibaba Cloud and Baidu, as well as entertainment institutions including Huace Film & TV and Mango-related entities. By public reports, this is already the largest financing round in the global video foundation model track.

On August 31, a Kuaishou announcement showed that the National AI Industry Investment Fund contributed RMB 1.4 billion to Beijing Kling, corresponding to about 1.14% of the enlarged registered capital; an overseas entity, Zhengda Robotics, contributed about RMB 131 million, corresponding to a 0.11% stake. Back-calculating from these two sets of numbers, Kling's overall valuation is about RMB 122.8 billion. For reference, Kuaishou's market capitalization that day was about HK$130.76 billion. A single business unit's valuation is already approaching its parent company's market value itself.

What Holds Up the Valuation Is Revenue, Not Demo Videos

For the valuation to stand, revenue matters. Kling AI's disclosed figures are: revenue in the second quarter of this year exceeded RMB 850 million, up more than 200% year over year; first-quarter revenue exceeded RMB 650 million. As of March this year, the annualized revenue run rate (ARR) was approaching US$500 million, while in December 2025 the figure was about US$240 million. It more than doubled in a year.

This growth curve differs from most AI companies. Many model companies are still talking about API call volumes and token consumption, but Kling sells something more direct: saving individual creators and film/TV teams shooting and visual effects budgets. On the product side, it has launched native 4K direct video output and is penetrating from mass creation into industrialized film and TV workflows. The AI short film “Paper Phone” surpassed 100 million total views and was shortlisted for the AIGC section of the Beijing International Film Festival; generated content has also participated in a Hollywood-related project, completing coordination between AI-generated footage and live-action shooting.

Why This Timing

The timing is deliberate. On September 28, Kling officially announced that Kling 4.0 would launch in October, with the lightweight Kling 4.0 Flash first opening small-scale trials, and the full version entering beta that evening. The new version supports native generation of up to 30-second video in a single pass, 10-bit HDR output at 4K and 1080p, up to 10 keyframe inputs, and 15 multimodal references.

In other words, when the IPO rumors surfaced, the company had just assembled a combination: the new model was in place, commercialization data was rising, and the valuation anchor was clear. For underwriters, the storytelling window matters.

There is also a date hidden in the investor terms. If an IPO is not completed by October 30, 2031, investors have a redemption right. That deadline is not particularly tight, but it shows that when the capital came in, the exit path was written into the agreement. According to reports, after this round of dilution, Kuaishou's stake may fall to about 68%, with control still in its hands.

Three Unavoidable Questions

The first is competition in the track. ByteDance's Seedance is pressing hard in the same interval, and domestic video models have entered an intensive release period since the second half of last year, with parameters, duration, and resolution being repeatedly refreshed. Kling has consistently bet its differentiation on “controllability” and “professional delivery.” This route is friendly to customers' willingness to pay, at the cost of continuously bearing inference costs.

The second is inference economics. Video generation is widely recognized as a compute black hole, and the cost of a 30-second 4K output is far higher than generating an image. Whether revenue growth can outrun inference spending is the question the secondary market will ask most. At present, only revenue is public, not gross margin.

The third is governance structure. The business was spun off from its parent company and raised funds independently, following a typical Chinese internet company spin-off path. The advantage is that the valuation can be priced separately; the drawback is that after listing it will have to face both the parent company's performance fluctuations and its own independent narrative.

A Few Things Worth Watching

The excitement is in the valuation number, but what is worth tracking next are actually a few more mundane things: the paid conversion rate after the official launch of Kling 4.0, repeat purchases by enterprise customers in film and TV projects, and how quickly inference costs fall as models iterate. These determine whether Kling will ultimately be priced as a content tool company or merely as a model team.

An IPO is the endpoint of fundraising, and also the starting point of continuous scrutiny. By then, every quarter's numbers will have to stand on their own.