Kuaishou may be quietly nurturing an asset nearly as significant as itself.

This time, let’s revisit a familiar question: Can Kling truly propel Kuaishou into its next phase of growth?

Similar concerns surfaced last year when Kuaishou pivoted toward e-commerce. While e-commerce GMV once soared by over 70% year-on-year five years ago, growth had dwindled to 15% by 2025, prompting the market to search for the next catalyst to sustain Kuaishou’s valuation.

Following this narrative, Kling was thrust into the spotlight. Launched in June 2024, it has undergone over 30 iterations in 19 months, amassing over 60 million global users, generating over 600 million videos, and achieving an ARR of $500 million—a feat rarely seen among domestic AI video projects.

Recently, Kuaishou signaled an even bolder move: Kling may be spun off independently, with Pre-IPO financing initiated at a staggering $20 billion valuation. What does this figure signify? Based on Kuaishou’s closing market cap on the same day, Kling’s post-spin-off valuation would nearly equal 70% of the parent company’s total market value. In essence, Kuaishou may be harboring an asset almost as substantial as itself.

But beyond the headlines, answering that age-old question requires tangible results. For Kling to “support” Kuaishou’s growth, it must demonstrate independent, quantifiable revenue growth while competing in a fiercely contested arena where Kuaishou struggles to maintain absolute dominance against rivals like ByteDance and Alibaba. Mere technical milestones and rankings won’t suffice.

With these criteria in mind, let’s examine Kling’s recent trajectory.

01 The Landscape Shifts Rapidly

Let’s start with a story.

In early 2024, when Kling was still in its infancy, Sora had just released its demo, and the entire industry was gripped by a mix of excitement and anxiety. AI video was theoretically revolutionary but practically grueling—unstable visuals, inconsistent characters, jerky transitions, and a murky path to commercialization.

The Kling team took a counterintuitive approach: they prioritized character and scene consistency. While this may not sound as flashy as “generating 4K video,” it was a necessity for professionals like short drama producers, advertisers, and film editors. The last thing they wanted was for the same character to appear as two different people in consecutive shots.

Subsequent technical advancements validated this strategy. By late 2025, Kling released the O1 unified multimodal model, the Image O1 model, and version 2.6, achieving breakthroughs in synchronizing audio-visual elements and multimodal fusion in a single generation. Then came the 3.0 series in 2026, upgraded under the All-in-One philosophy, with native 4K output added in April—the first among domestic AI video models to achieve this.

The pace was relentless, and iterations were rapid. But in today’s competitive landscape, product iteration speed is no longer the rarest commodity.

In the first half of 2025, Kling was the only domestic video model that could rival Sora. That claim wasn’t an exaggeration at the time. But by early 2026, the landscape had shifted dramatically.

In February, ByteDance’s Seedance 2.0 launched on Jimeng and Doubao, dubbed “director-level AI” by the industry. BOC Securities calculated: its pure video generation cost dropped to ¥1/second, content usability surged from 20% to over 90%, and the production cost of a 90-minute animated drama fell from over ¥10,000 to around ¥2,000. Jimeng AI’s MAU reached 13.52 million in Q1, with Seedance 2.0’s MAU briefly exceeding 45 million during the Spring Festival.

Then Alibaba entered the fray. HappyHorse-1.0 topped the Artificial Analysis blind test platform under an anonymous identity, with generation costs as low as ¥0.44/second and visual quality rivaling industry leaders.

In less than three months, the AI video competitive landscape was reshuffled. The top three on Artificial Analysis’s text-to-video rankings were now dominated by domestic models: Alibaba’s HappyHorse-1.0, ByteDance’s Seedance 2.0, and Kuaishou’s Kling 3.0.

On the surface, Kling remained in the top tier. But a closer look reveals a subtle truth: no clear technological gap separates these competitors.

When products converge, competition shifts to more pragmatic dimensions: ecosystem, cost, operational depth, and user base. ByteDance leverages the ecosystem of Douyin and Jianying, giving it a naturally larger user base. Alibaba’s HappyHorse slashed costs to rock-bottom levels.

Kling’s position became delicate: it was the earliest mover, but now it had competitors closing in fast—from both ahead and behind.

02 To Stay or To Go

However, if we focus solely on revenue growth, Kling has delivered impressive results.

In 2025, Kling generated approximately ¥1.04 billion ($150 million) in revenue, with ¥340 million in Q4 alone. Monthly revenue exceeded $20 million in December, corresponding to an ARR of $240 million. By January 2026, ARR jumped to over $300 million; by late April, it doubled to $500 million.

The growth curve steepened dramatically, with ARR surging from $240 million to $500 million in six months.

User composition also improved. By the end of 2025, Kling served over 30,000 enterprise clients across marketing, e-commerce, film and TV, short dramas, animation, and gaming. The proportion of professional consumers and commercial users rose, with B-end weight increasing steadily. P-end paid subscriptions contributed nearly 70% of Kling’s revenue.

But the financial picture wasn’t entirely rosy.

Kling is a computational resource hog. Kuaishou’s R&D spending surged from ¥12.2 billion to ¥14.5 billion in 2025, an 18.8% increase, with AI being the primary driver. CFO Jin Bing stated at an earnings call that Kuaishou’s 2026 capital expenditures were expected to reach ¥26 billion, up ¥11 billion from 2025, primarily allocated to training and inference compute for Kling and other large models. Kuaishou’s 2025 adjusted net profit was approximately ¥20.6 billion—nearly all of which was reinvested into the AI arena.

More troubling was pricing. While technology iterated and model capabilities improved, B-end pricing had little room to rise—and even faced downward pressure due to intensifying competition. A senior executive at a domestic video model company put it bluntly: “Terminal prices haven’t changed, but technology keeps iterating. The objective reality is that as competition intensifies, there’s less and less room for premiums.”

This was evident in Kling’s pricing changes: the 2.6 version raised image generation fees compared to earlier versions, with 5-second videos increasing from 20 “inspiration points” in the 1.6 model’s standard mode to 50 points. However, within the broader industry context, such increases reflected quality premiums rather than overall market price hikes.

Compute costs soared while terminal prices faced pressure. Operating in this squeeze felt like running between two walls—faster and faster, but with narrowing space.

If the story ended here, it would be just another AI narrative of “great product, tough commercialization.” But what makes Kling truly interesting lies in the capital markets.

Recently, rumors surfaced that Kuaishou plans to spin off Kling AI, initiating Pre-IPO financing at a $20 billion valuation, with Tencent and other investors already in talks. Kuaishou confirmed in a HKEX announcement that its board was evaluating restructuring options for Kling AI’s assets and business, “which may involve introducing external financing,” but emphasized that the plan was in its preliminary stages.

The $20 billion valuation is key to understanding this story.

Morgan Stanley analyzed the logic clearly: before the spin-off, its valuation of Kling was only about $6 billion; after the spin-off, market pricing could reach $20 billion. Same asset, same revenue, same team—but three times the valuation on a different reporting structure.

What accounts for the difference? The pricing framework.

Kuaishou’s core business is valued based on DAUs, ad ARPU, and e-commerce GMV, with markets using PS or EV/EBITDA multiples typically ranging from 2x to 4x. Kling, as a generative AI model, is benchmarked against frontier model companies like Runway and Pika.

When two valuation logics coexist on a consolidated balance sheet, the market defaults to the more conservative framework—discounting Kling for “parent business drag” and Kuaishou’s core business for “AI investment profit erosion.”

Thus, even as Kling’s revenue surged from $150 million in 2025 to over $500 million in 2026, Kuaishou’s stock price didn’t reflect this growth. Not because Kling was underperforming, but because it was misvalued under the wrong framework.

JPMorgan took an even more aggressive stance. Based on Kling’s expected ARR of $1.3 billion in Q1 2027 and a 40x ARR valuation multiple, it estimated Kling’s total valuation at $52 billion, with Kuaishou’s stake worth approximately $26 billion. Using a sum-of-the-parts (SOTP) valuation, JPMorgan set Kuaishou’s target price at HK$73 per share, implying ~37% upside from current levels.

The spin-off logic becomes simple: it’s not that Kuaishou doesn’t want Kling, but that capital markets only assign premium valuations to independently operated AI assets.

03 Can It Hold Up?

Returning to the original question—can Kling truly drive Kuaishou’s second wave of growth?

First, we must define what “hold up” means.

If the standard is “becoming a significant component of Kuaishou’s valuation,” then the spin-off itself has largely accomplished this. The market’s willingness to value Kling independently at $20 billion indicates it already supports a substantial portion of Kuaishou’s imagination space.

For a company with a market cap below $30 billion to harbor an asset the market values at $20 billion—by any measure, that’s a positive development.

But if the standard is “independently driving Kuaishou’s revenue growth,” there’s still considerable distance to go.

At $500 million in ARR, monthly revenue is approximately $42 million (~¥300 million), while Kuaishou’s quarterly revenue nears ¥40 billion. Kling’s share of Kuaishou’s total revenue remains below 5%. Even if revenue hits $1 billion in 2026, its proportion will still be limited. In terms of scale, it cannot yet carry half of Kuaishou’s revenue load.

More critical is the competitive landscape. In 2025, Kling had a first-mover advantage, but by 2026, ByteDance’s Seedance 2.0 and Alibaba’s HappyHorse had caught up. With no clear technological generation gap, operational efficiency, customer loyalty, and ecosystem depth become decisive. Ultimately, this battle hinges on who can execute more reliably over the long term.

Then there’s the dynamic balance of costs and profits. With ¥26 billion in capital expenditures, profit margins remain uncertain. The AI video arena is trapped in a price war, with soaring compute costs and stagnant terminal pricing ensuring a long ROI cycle.

At this point, I want to share a perspective beyond technology.

After observing countless AI company stories over the past two years, one pattern is clear: in this industry, the rarest commodity isn’t technological breakthroughs or fundraising prowess, but the patience to evolve a product from “impressive” to “reliable.”

Kling’s 19-month iteration cycle of over 30 updates proves the team’s execution and product sense. From being benchmarked against Sora to securing a spot in China’s top tier, and now being independently valued by capital markets, each step has left solid footprints.

But “reliability” hinges on deeper factors: cost structure, ecosystem density, user habits. These won’t be resolved in a quarter or two—they require prolonged refinement and validation.

Kuaishou’s decision to spin off Kling at this juncture makes logical sense. Matching a high-quality asset with an independent pricing system and financing channel is rational in today’s market. Kling’s growth trajectory justifies this expectation.

But to “drive Kuaishou’s second wave of growth,” Kling must continue leading amid fierce competition and prove whether markets will truly buy into independently priced AI assets.

The road is long. For an internal project to reach $500 million in ARR in under two years demands respect. But it hasn’t yet grown to a scale that can single-handedly reshape the competitive landscape. The true test lies in maintaining momentum in the next phase, turning that independent post-spin-off balance sheet into valuation gains.

Before it fully proves itself, spinning it off to let it run freely is a strategic judgment in itself. Sometimes, a company’s foresight lies not in what products it builds, but in whether it dares to let its best creations face the market alone.

Kuaishou has chosen to let Kling face the market independently. The rest is up to Kling.

This article is original to Xinmou. For reprint authorization or business cooperation, please contact WeChat: ycj841642330 (add with company and title noted).

— END —