NUPIAO Reporter |
NUPIAO Editor | Wen Shuqi
“Pulling off a financing round of around $3 billion in such a short time might just be a new record in China’s entire fundraising history,” Lighthouse Capital founder and CEO Zheng Xuanle told NUPIAO.
After every project, the team writes a thank-you list in their group chat to recognize everyone’s hard work. “The thank-you list for this project is definitely the longest we’ve ever had,” Zheng said.
About a week ago, on the evening of July 2, Kuaishou announced on the Hong Kong Stock Exchange that it had completed independent financing for its video generation large model, Kling AI.
This deal set a record in the global video generation model space: Kling AI raised up to $3 billion at a $15 billion pre-money valuation, bringing in Tencent, Alibaba, Baidu, and other internet companies, plus several national-level funds, local state-owned capital, USD funds, and industrial investors. The post-money valuation could reach $18 billion, and Kuaishou retained a 68.33% controlling stake.
From the first market rumors to the official announcement, roughly two months passed. Lighthouse Capital, the sole financial advisor and also a financial investor (its L2F Lighthouse Founders Fund participated), was involved throughout the process. This is the same firm that backed Kuaishou in its early startup days 12 years ago.
“When the decision was truly made to spin off Kling AI for independent financing, for a listed company like Kuaishou, timing was absolutely the most critical factor,” Zheng explained to NUPIAO. The independent financing involved complex asset restructuring, and as a listed company, Kuaishou needed to avoid any unnecessary stock price swings from leaked news, so shrinking the information exposure window was the top priority—time was the number one principle.
The team had to literally break down every milestone to a specific date: when to complete the first round of investor discussions, when to finalize internal judgments, when to hold the investment committee, when to finish negotiating the deal documents. “The vast majority of investors went from first contact to decision in basically two weeks to a month—the timeline was incredibly tight,” he said.
Under such a high-pressure schedule, institutions that couldn’t complete their internal approval processes in time were screened out. Investors with insufficient cash on hand who needed to raise funds from the market were also largely eliminated at this stage.
In AI fundraisings for top-tier projects, it’s almost a given that people keep popping up claiming they have allocations to resell or want to raise money. Kling AI was no exception. Zheng admitted that one of the team’s main jobs was controlling information flow and strictly managing the cap table to prevent the financing quota from circulating layer by layer in the market.
The almost ruthless process management kept the entire fundraising on track without any major derailment. But ultimately, what made investors willing to bet was Kling AI’s commercial value as an investment target—that was the core message the team had to deliver during the roadshow.
While the top tier of large language models is dominated by the overseas “big three”—OpenAI, Anthropic, and Google—video generation models are one of the few tracks where China still holds a global lead. Zheng believes this is the crucial backdrop that attracted industrial capital, national funds, and overseas money to join this round.
Moreover, video models aren’t just a vertical application. If you view them as one of the key technical routes toward world models, their long-term value can rival that of language models. And looking at Kling AI’s ARR (annual recurring revenue) performance, it already has a certain level of commercialization capability and strong growth momentum.In a highly concentrated market where the top models are closer to commercial scenarios, the leading model can capture a larger share of the market and profit margins.
According to Kling AI’s disclosed data, by March 2026, its ARR had reached about $500 million. With a post-money valuation of $18 billion, that’s a price-to-sales ratio of 36x, close to the ratios of OpenAI and Anthropic at the same time.
The diverse shareholder structure in this deal—covering internet platforms like Tencent, Alibaba, and Baidu, national-level funds and local state-owned capital, top RMB funds, USD funds, and industrial investors—was also a major focus of the market.
Zheng said this was the structure the team had designed from the very beginning. “We felt that a beautiful investor lineup must balance both diversity and brand density.Only by combining the two can you help the company secure as much resource support as possible.”
Some outsiders had interpreted the lineup as “BAT joining forces to block ByteDance.” Zheng didn’t directly respond to that, only saying that video generation has become a “must-win battleground” in the AI era, and all investors want to back the most leading company.
After this round, Kling AI will have independent fundraising capabilities, its own employee equity incentive plan, and a standalone path to the capital markets. Kuaishou, meanwhile, retains control, consolidates Kling AI in its financial statements, and strips the business out of the traditional internet platform valuation framework.
In Zheng’s view, this is the deal’s greatest strategic significance. “The ability to attract talent and the decisiveness in computing power investment are the core of sustained competitive advantage for large model companies.” The announcement showed that Kling AI had net losses of RMB 500 million in 2024 and RMB 1.9 billion in 2025; ample funding is needed to sustain the competitive battle ahead.
But high valuation and heavy investment also mean high constraints. According to the announcement, investors have a redemption right if the company fails to IPO by 2031, along with multiple compliance requirements such as algorithm filing, telecom licenses, and overseas entity acquisition. If key conditions aren’t met, investors can demand a buyback at principal plus 8% simple interest.
From a capital market perspective, this is a typical AI company growth pact: exchanging current resources for future growth, and strict governance arrangements for investors’ long-term trust.
It won’t be a quick-win race. The technological path for video models hasn’t converged as much as for large language models. To keep a performance edge, Kling AI’s accurate judgment on technology choices and its pacing of resource investment will remain critical in the near term.
