AI video generation boosts China's Kuaishou stock 84%

Few leading companies are transforming to artificial intelligence (AI) as quickly as Kuaishou Technology.

Kuaishou, long known for lagging behind ByteDance's TikTok in short video streaming, has reinvented itself in 18 months and expanded into AI-generated video. The company's stock has soared 84% over the past year, and excitement about the global potential of the Kling video generator sparked another double-digit gain on Monday (January 5).

According to Sensor Tower, Kuaishou's Kling AI app is the highest-grossing graphics and design app on iPhone in South Korea and Russia as of January 2, and is also in the top 10 in markets including the US, UK, Japan, Australia, and Turkey. Although it spends only a fraction of the money that its better-known rivals are pouring into AI, the company predicts annual revenue from AI video of US$140 million in 2025. It has 60 million users and sales are accelerating as brand awareness grows beyond China.

All of this came about because the Beijing-based company refused to let OpenAI go to market. When Kuaishou, a San Francisco-based startup, unveiled its Sora AI video tool in February 2024, it set an audacious goal to have a comparable solution in place by that summer, and it achieved that goal with the debut of Kling in July.

“This is a resource-constrained game, so timing is very important. If you're too early, you're wasting cash; if you're too late, you lose your advantage,” said Gai Kun, chief engineer at Kuaishou.

Gai is the mastermind behind Kuaisho's rapid rise, overseeing a team of hundreds of people developing Kring. To date, Kling AI has enabled the creation of over 600 million videos. To capitalize on this strong start, he next wants to build an AI-first, video-only content platform. If Kuaishou wants to build it first, he believes there is only a narrow window of opportunity.

Guy describes the short-video specialist as a mid-sized company without deep pockets and believes its biggest opportunity is perfecting the timing of its products and reducing mistakes. Kuaishou, currently valued at around US$41 billion, has regained investor support largely thanks to Kling, a rare consumer AI app with stable revenues. It's also growing among business customers, with more than 30,000 companies and developers integrating its API.

On global benchmarking site Artificial Analysis, Kling is the only Chinese-made model to rank in the top three for both text-to-video and image-to-video performance. To stay ahead of local rivals like ByteDance and startups like MiniMax, Gai's team is keeping the pace of upgrades high.

In December, Kling announced a new O1 model that handles text, images, and video prompts simultaneously. Justin Moore, a partner at Andreessen Horowitz who invests in AI tools, likened the release to Google's image-generating nanobananas, which have been praised for their quality. She shared several examples of Kling O1 transforming objects that move on command, including a clip that transforms a cat into a Chihuahua. Another user showed how to combine Nano Banana and Kling to create high-fidelity movie scenes.

Kling users can now combine video and scripts to drive camera movement, or generate footage from manga-style storyboards combined with character reference images. This level of creative freedom is central to the O1 upgrade, which leverages the relative maturity of the company's technology.

While industry giants such as OpenAI and Alibaba Group Holding view video generation as part of a broader AI ecosystem, Kuaishou has avoided money-wasting competition to build multipurpose foundation models. Kling operates on a unique profit and loss structure, effectively functioning as an internal startup, Guy said.

In the video-focused space, Kling competes with U.S. rivals Runway and Luma AI, both of which raised hundreds of millions of US dollars last year at valuations of more than US$3 billion each. Kling's paid subscription plans range from US$7 to more than US$100 per month, about the same as its peers, but the Chinese platform has shown faster progress in monetization. When it comes to OpenAI, the San Francisco company employs a two-tier system. The company's top-of-the-line Sora 2 video tool is part of the $200 per month ChatGPT Pro subscription, or users can access a free mobile app of the same name targeted at viral TikTok-style sharing.

U.S. services such as Alphabet's Sora and Google Veo are not available in China, which helps Kling domestically, but the majority of paying users are based overseas, Guy said. Some of the prominent groups include movie studios, marketers and social media influencers, he added. As of the first quarter of 2025, about 70% of Kling's revenue will come directly from user subscriptions, with the rest coming from enterprise customers who integrate the software, company executives said.

Looking to the future, Kuaishou's head of technology predicts a paradigm shift in the way AI content is consumed. Rather than adding generated AI video on top of existing services, AI-native entertainment hubs will be required. For example, ordinary users could be portrayed as protagonists in short drama series or interactive video games, Guy said. This change could come within one to three years, he added.

“Every content platform relies on a production ecosystem and has a finite tolerance for disruption from AI,” Guy said. “If you want to be a true leader in your particular field of AI, you need to imagine the future.”

His appetite for high-stakes gambles reflects a career spent in China's cutthroat internet field. A Tsinghua University-trained machine learning scientist, Gai joined Alibaba in 2011 through the Elite Researcher Program and works on the e-commerce giant's advertising platform.

Gai joined Kuaishou in 2020 and was helping assemble a small team of AI researchers by early 2023 after he and CEO Cheng Yixiao concluded the company needed to adopt AI technology. A year later, Gai gave Kuaishou a mandate to beat OpenAI by being the first to release a fully trained, publishable video generation model. It was an uncomfortable time, but essential for timing-conscious executives.

“Everyone there was surprised when I proposed that goal,” Guy said. “But if we want it to go from one person to another, we have to ship it first.”Bloomberg