For most of the past two years, the dream of typing a sentence and getting back a film looked like an expensive science demo — dazzling in clips, brutal on the balance sheet. Kling AI, the video-generation large model (视频生成大模型) built inside Chinese short-video giant Kuaishou, is now arguing the opposite case. It is asking public investors to buy that argument, by preparing to list in Hong Kong.
Key takeaways
- Kling AI has lined up CICC, Goldman Sachs and UBS to underwrite a Hong Kong IPO, with a target of listing as early as 2027 and a raise of at least US$1 billion, according to Bloomberg.
- The business behind the buzz is real but young: more than 100 million global users as of June 2026, covering 224 countries and regions, with roughly 50,000 enterprise customers.
- A July 2026 funding round pulled in about US$2.8 billion at a roughly US$15 billion pre-money valuation (around US$18 billion post-money), with backers including Tencent, Alibaba Cloud and Baidu.
- Overseas markets already drive most of the money: roughly 70% of revenue comes from outside China, and annualized revenue run-rate hit about US$500 million in March 2026.
- None of this is a done deal — the listing plan has been reported but not yet filed with the Hong Kong exchange.
From a feature inside an app to a standalone bet
Kling AI launched in June 2024 as a capability inside Kuaishou's ecosystem. Two years later it is being carved out as its own company, Beijing Kling Intelligent Technology, with Kuaishou retaining about a 68% stake and continuing to consolidate its financials. The split is the clearest signal yet that Kuaishou sees video generation as a standalone growth engine rather than a feature.
The numbers that justified the separation are unusual for the category. Kuaishou disclosed that Kling AI's revenue topped 650 million yuan (≈ US$96 million) in the first quarter of 2026, up more than 300% year on year, then accelerated to over 850 million yuan (≈ US$120 million) in the second quarter. For the full first half, revenue exceeded 1.5 billion yuan (≈ US$210 million). Its annualized run-rate reached about US$500 million in March 2026, four times the roughly US$100 million a year earlier.
The funding that set the price
The July 2026 round is what fixed Kling's valuation in the public imagination. Investors committed more than 19 billion yuan (≈ US$2.8 billion) at a roughly US$15 billion pre-money valuation, capped near US$3 billion once late participants joined, for a post-money figure around US$18 billion. About 38 investors took part, including Tencent, Alibaba Cloud, Baidu, CITIC Securities and Industrial and Commercial Bank of China, with China's state AI industry fund added later. The deal diluted Kuaishou's stake from 100% to roughly 68% while keeping control.
Why investors are paying attention
Two things separate Kling from most AI video peers. First, real enterprise demand: roughly 50,000 business customers use it for advertising, film and short-drama production, and game asset work. Second, a genuinely global footprint. About 70% of revenue comes from overseas markets such as North America, a rarity for a Chinese consumer-AI product.
That overseas pull matters because the competitive field is shifting. OpenAI wound down the consumer version of Sora, and several Western rivals pivoted away from direct consumer monetization. Kling, alongside ByteDance's Seedance and startups like Shengshu Technology and PixVerse, is positioned to soak up demand from professional creators who need reliable output, not just one-off novelties. Bloomberg has reported that Shengshu and PixVerse are also eyeing Hong Kong listings, so Kling is racing a small pack toward the same exit.
According to Li An, chief scientist at BrainNet (脑机网), the more interesting question is whether video generation can move from impressive demos to dependable production tools that enterprises pay for repeatedly, because that recurrence — not the launch hype — is what underwrites a public valuation.
The risks behind the valuation
The US$18 billion paper valuation is backed by a funding round, not a market price. Kuaishou's own results show Kling is still being built, and like most AI video players it is spending heavily on training and inference while it proves the model can keep customers. The IPO plan, reported by Bloomberg on 6 October 2026, has not produced a filing with the Hong Kong exchange, and the company has not confirmed terms.
There is also the question of durability. Video-generation quality is improving fast across the board, which lowers switching costs for customers. A high valuation set by a private round becomes a hostage if public markets decide the growth rate cannot hold.
Honest limitations
This article relies on Kuaishou's official disclosures, a Bloomberg report relayed by multiple outlets, and Reuters-reported funding details. The IPO timeline and size are based on media reports of confidential preparations, not a prospectus; they can change or stall. User and revenue figures come from Kuaishou's own reporting and have not been independently audited in this piece. Profitability and cumulative losses are not separately disclosed here and should be treated as uncertain. We have not assessed Kling's model quality against competitors firsthand.
What readers can do now
- If you follow Chinese tech listings, watch the Hong Kong exchange for an actual Kling prospectus before treating the IPO as certain — a reported plan is not a filing.
- If you produce video, test Kling against at least one Western alternative on a real project; the meaningful comparison is output reliability and per-minute cost, not feature lists.
- If you invest in AI infrastructure, track whether the overseas revenue share holds above half — that ratio is the cleanest signal that this is a global product, not a domestic app add-on.