Kling AI's Rise: How Kuaishou Built China's First Commercially Viable Video Generation Model

What Is Kling, and Why Does It Matter?

Kling is a video generation model developed by Kuaishou, one of China's two dominant short-video platforms. Launched in June 2024, it allows users to generate high-definition video clips—up to two minutes at 1080p—from text prompts or static images.

What distinguishes Kling from most AI products of its era is not just its technical capability, but its commercial performance. Within ten months of launch, Kling's annualized recurring revenue (ARR) crossed $1 billion. For context: Cursor, the AI coding tool widely regarded as one of the fastest-growing software products in recent memory, took twelve months to reach the same milestone. OpenAI's Sora, by comparison, accumulated only $2.1 million in in-app revenue across its entire lifespan before being shut down, against an estimated daily operating cost of $15 million.

In July 2025, Kuaishou announced it would spin Kling out as an independent entity. The fundraising round—$3 billion, backed by Tencent, Baidu, and Alibaba simultaneously—valued the standalone business at over $18 billion. At the time, Kuaishou's total market capitalization had been hovering around $25 billion for months. In other words, a single AI product had become worth roughly two-thirds of the entire parent company.

Why Did Kuaishou, Not a Pure-Play AI Lab, Build This?

The answer lies in structural position, not luck.

By 2023, Kuaishou faced a strategic dilemma that was common across China's internet sector but particularly acute for mid-tier platforms. Daily active users had essentially plateaued. Time-spent-per-user had stabilized around 130 minutes per day. Advertising revenue, while steady, was increasingly difficult to grow through additional sales spending. The company's legacy live-streaming business had stalled.

The broader threat was existential. In the same period, the US online education platform Chegg—functionally similar to China's Zuoyebang or Yuanfudao—watched its stock fall 50% in a single day after disclosing that ChatGPT was suppressing new user growth. The market had delivered a clear verdict: internet companies that could not credibly participate in AI would be repriced as structurally declining businesses, regardless of their current cash flows.

For Kuaishou's CEO Cheng Yixiao, the question was not whether to pursue AI, but how to do so without destroying the balance sheet.

What Made Kuaishou's Position Structurally Unusual?

China's internet industry can be roughly segmented into three tiers. The first tier—ByteDance, Tencent, Alibaba, Meituan—operates diversified business empires with substantial free cash flow and the capacity to absorb multi-year losses on infrastructure bets. The second tier includes platforms like Bilibili, Xiaohongshu, Weibo, and Zhihu, each with concentrated business models, limited cash reserves, and limited ability to fund large-scale AI R&D without external capital.

Kuaishou occupied an unusual middle position. Its e-commerce business—built on an internal traffic loop that sold Kuaishou's own audience to Kuaishou's own merchants, generating both transaction commissions and advertising revenue—had created a durable and growing cash flow engine. The company was generating roughly 20 billion RMB in annual profit, without being drawn into the kind of winner-take-all platform wars that had consumed Meituan's resources.

This is what made Kuaishou's AI bet structurally different from its peers. It had the financial capacity of a first-tier company in a specific domain, without the competitive exposure that typically accompanies that scale. It could invest heavily in a focused area—video generation—without being simultaneously forced to defend market share in food delivery, cloud computing, or enterprise software.

How Did the Technology Come Together?

The technical foundation for Kling was laid earlier than its 2024 launch date suggests.

In late 2023, Kuaishou revived an internal project called "Puji"—a tool for converting static images into two-second animated GIF stickers. The project's lead, Wan Pengfei, became one of Kling's core architects. Research papers published by his team through 2023 show sustained work in image-to-video generation, video instance segmentation, and human motion recovery.

The decisive architectural shift came in February 2024, when OpenAI released Sora and demonstrated the commercial viability of the Diffusion Transformer (DiT) architecture. Unlike earlier video models—which generated video by producing individual frames and attempting to stitch them together—DiT treats video as a unified spatiotemporal structure, breaking it into small patches and training the model to understand abstract relationships between subjects, objects, camera movement, and action. The result is more coherent motion, better long-form consistency, and fewer of the visual artifacts that made earlier AI video obviously artificial.

Kuaishou moved immediately. Kling adopted DiT architecture and launched in June 2024 with support for up to two minutes of 1080p video. ByteDance, by contrast, had two competing internal video teams still debating architectural direction at the same point; its DiT-based product, Jimeng, arrived two months later.

Speed mattered. Kling launched a subscription service within its first month—with a top tier priced at 666 RMB (approximately $92) per month—and iterated through more than twenty versions in six months, adding features including first-and-last-frame control, motion brushes, and expanded generation parameters.

Who Are the Key Players, and Why Is the Field Narrowing?

The competitive dynamics in AI video generation have consolidated rapidly.

OpenAI closed Sora. The economics were unsustainable: high infrastructure costs, limited commercial traction, and no clear path to the kind of professional workflow integration that drives B2B recurring revenue. What remains is a market dominated by two serious players: Kuaishou's Kling and ByteDance's Jimeng.

This consolidation reflects a broader pattern in AI infrastructure. The capital requirements for training and serving frontier video generation models—GPU clusters, data centers, inference optimization—create barriers that most companies cannot clear. ByteDance has the resources to compete. Most others do not.

For Kuaishou, the spin-out structure serves a secondary purpose beyond fundraising. Hong Kong's equity markets have demonstrated a strong preference for pure-play AI companies over conglomerates that mix legacy internet businesses with AI units. Roadshow materials circulating in mid-2025 noted explicitly that mixed entities face meaningful valuation compression relative to standalone AI vehicles. Spinning Kling out allows investors to price the AI business on its own growth trajectory, rather than having it discounted by association with a plateauing short-video platform.

The same logic is driving parallel moves across China's tech sector: Baidu is separating its Kunlun chip business; Alibaba is carving out its semiconductor unit Pingtouge. Kuaishou was the first to execute a transaction at this scale.

What Are the Key Variables Going Forward?

Several structural factors will determine whether Kling's early momentum translates into durable market position.

Market size and ceiling. AI video generation is a real market with genuine commercial use cases—advertising production, short-form drama, creative tools for individual creators. But it is not yet clear whether the total addressable market approaches the scale of AI coding assistance, which benefits from near-universal applicability across software development workflows. Kling's current customer base—advertisers, short-drama producers, professional content creators—is meaningful but bounded.

Competitive intensity from ByteDance. Jimeng has the backing of China's most profitable internet company and access to the largest short-video distribution platform in the world. The competitive gap between Kling and Jimeng is not fixed; it reflects a lead measured in months, not years.

Infrastructure cost curves. AI video generation remains compute-intensive. Kling's commercial viability depends partly on inference costs continuing to fall as hardware and optimization improve. If cost reduction stalls, margin pressure increases.

Governance post-spin-out. Kuaishou has indicated it will retain control of the Kling entity despite the external capital raise. How that control is structured—and whether it creates friction with minority investors over product direction or capital allocation—remains an open question.

What Does This Tell Us About the Broader Transition?

Kuaishou's trajectory illustrates a structural dynamic that will likely repeat across China's internet sector over the next several years.

The mobile internet era created a set of businesses optimized for user acquisition, advertising monetization, and platform lock-in. Those businesses are now mature. The question facing every company in the sector is whether it can generate enough cash from legacy operations to fund a credible AI transition before the window closes.

The window is closing. AI infrastructure costs have risen to levels that most mid-tier internet companies cannot sustain. ByteDance's net profit fell approximately 70% in a recent reporting period due to AI capital expenditure. Alibaba's free cash flow has turned negative. Even Tencent, historically conservative with capital allocation, has significantly increased infrastructure spending.

For companies that cannot make this transition—that remain purely advertising-dependent content platforms without a distinct AI capability—the market has already begun applying a structural discount. The Chegg precedent is instructive: a business that was growing and profitable became a value trap the moment investors concluded it had no AI-era relevance.

Kuaishou's achievement with Kling is not simply that it built a good product. It is that it identified a specific domain where its existing technical capabilities, content ecosystem, and financial position created a defensible entry point—and moved before the door closed.

Whether it can hold that position is a separate question. That it got through the door at all, ahead of better-resourced competitors, is the more important observation.

Related Coverage:

Kuaishou's Kling AI Secures $3B Mega-Round Ahead of Hong Kong IPO