By Shang Ye

Edited by Shen Xiao

Even with the announcement of new organizational changes just before the holiday, Kuaishou's latest developments still captured market attention after the break.

On October 13, Kuaishou's stock price fell to HK$28.24 during intraday trading, marking a historic low. In contrast, Kling's valuation soared to a new high. Reports indicate that after completing a new round of financing worth approximately $3 billion, Kling's post-money valuation neared $18 billion, with underwriters already selected for a potential Hong Kong IPO as early as 2027.

While the parent company hit a new low, the subsidiary's valuation continued to rise, reflecting differing judgments from the capital markets: after all, if Kling were an AI-listed company, how far would Kuaishou, post-spin-off, be from becoming a relic of a bygone era?

A classic joke in investment circles goes: 'Kling steals the soul, Hongguo demands its toll—no one knows where Kuaishou's bottom lies.' The former refers to Kuaishou's most imaginative AI assets being priced separately, while the latter highlights how ByteDance's Hongguo short dramas are cutting into Kuaishou's user time.

While not rigorous, this quip captures Kuaishou's most pressing dilemma.

Latest data from Nomura shows that in August, among China's top mobile apps by user time share, Douyin accounted for 19.3%, WeChat 18.3%, Douyin Lite 6.3%, Hongguo Free Short Dramas 5.9%, and Kuaishou 3.7%.

Kuaishou's main app ranked behind Douyin, WeChat, Douyin Lite, and Hongguo.

More noteworthy are the changes in these shares.

Hongguo's time share increased by 3.7 percentage points year-over-year and 0.9 percentage points month-over-month; Kuaishou's main app saw a 1.1 percentage point year-over-year decline and a 0.2 percentage point month-over-month drop. Kuaishou Lite's time share stood at 2.2%, down 0.4 percentage points year-over-year and 0.1 percentage points month-over-month.

By this metric, Hongguo's time share alone equals the combined 5.9% of Kuaishou's main app and Kuaishou Lite.

Some investors quipped: 'Hongguo is aggressive. While Tencent seems most at risk, Kuaishou has already fallen off ByteDance's radar.'

Yet Kuaishou currently has little time to fret over Hongguo's encroachment on lower-tier user time. The challenge of smoothly spinning off Kling to secure a ticket to the next era already tests managerial wisdom.

As an internal startup within a large corporation, Kling faces an 'internal surgery' before going public independently.

On September 30, Kuaishou announced two personnel changes in one day. Yu Yue, formerly Kuaishou's senior vice president, became Kling's full-time CEO; Kuaishou's community science division was taken over by founder Cheng Yixiao.

Earlier, in July, Kuaishou disclosed Kling's financing and restructuring plan to the Hong Kong Stock Exchange.

The plan revealed Kling's intention to raise up to $3 billion. It also introduced employee incentives, stock ownership, and option plans totaling no more than 15%. If fully utilized, Kuaishou's stake would dilute from 100% to approximately 68.33%.

Kuaishou explained the rationale in its announcement: to reflect Kling's value independently, attract professional investors, align management responsibilities with operational and financial performance, and improve governance and focus.

Prior to this announcement, Kling had experienced a brain drain of core technical talent. Zhang Di, who led Kling's foundational technology and model development, left Kuaishou in August 2025.

After joining Alibaba in November of the same year, his team developed HappyHorse, which anonymously entered Artificial Analysis's blind video generation rankings in April, topping the text-to-video and image-to-video categories (without audio), surpassing models like Kling.

However, staff turnover did not halt after the new incentive plan. In August, Kling's key personnel Wang Xintao and Wang Meng were reported to have left.

Wang Xintao, lead author of open-source projects like Real-ESRGAN and GFPGAN, joined Kuaishou in 2024 to lead video generation efforts. Wang Meng, with research experience at Google and Baidu Apollo, was also reported to have participated in Kling's early development.

Pre-IPO turbulence is inevitable for AI firms. Before Zhipu's listing, holding platforms Huihui and Zhideng owned approximately 9.80% and 6.75% of shares, respectively. Employees transitioned from 'early participants' to explicit shareholders, with their stakes disclosed in filings.

MiniMax established a pre-IPO equity incentive plan, granting options equivalent to about 20.89 million shares to 392 participants. Founder Yan Junjie retained approximately 25.22% economic interest but controlled about 71.49% of voting rights.

Zhipu addressed institutionalizing employee wealth, while MiniMax ensured founder control post-dilution. Kling, meanwhile, must rebalance power among the parent company, business leaders, and new investors.

Before an AI company goes public, relationships once sustained by trust must transform into auditable, disclosable, tradable, and accountable systems.

Kimi demonstrates the consequences of failing to act in time. Before advancing its IPO, Moonshot AI still addressed equity and governance disputes with its predecessor, Recurrent Intelligence, and some investors. Unresolved issues may affect governance, ownership, and disclosure during the listing process.

The primary market tolerates ambiguity for growth stories; the IPO market does not.

Moreover, transforming a large company's internal startup into an independent listed entity requires solving transparency challenges.

Unlike native AI model founders, Kling's professional manager Yu Yue must navigate personnel, power, and interests (translated as 'interests') dynamics within a corporate hierarchy, testing interpersonal wisdom beyond formal announcements.

Kling's uniqueness lies in its pre-IPO CEO—a technical leader rather than a founder or investor.

Yu Yue, formerly Kuaishou's senior vice president, was thrust into leading Kling, a model company incubated from Kuaishou's operations, capital, and teams.

This sets Kling apart from native large model companies like Zhipu and MiniMax.

In those firms, technical founders typically drive establishment, financing, and operations. Research direction, corporate control, and personal interests align from the start. Technical judgment precedes capital expenditure in their DNA.

Kling followed the opposite path. It began as an internal business and technical project before becoming independent through capital increases and restructuring. Yu Yue stepped into operations, but Cheng Yixiao remains chairman, and Kuaishou the controlling shareholder. Under the capital increase plan, Kuaishou's stake dilutes to ~68.33%, retaining control and consolidating Kling in its financial statements.

This means Yu Yue is accountable for model capabilities and business results but may lack independent resource allocation authority. To offset this, Kuaishou granted Yu Yue an incentive package resembling a founder's.

Announcements show Yu Yue received 3% of Kling's equity in rewards and 1% in options. Uniquely, the 3% equity carries ten votes per share in special voting rights. Future exercise of the 1% options may also include special voting rights.

Under current arrangements disclosed on August 31, without superposition (translated as ' superposition ' or ' superposition options'), Yu Yue holds ~23.62% of Kling's voting rights. If the 1% options are fully exercised and other shares remain at one vote each, his 4% equity would yield 40 votes out of 136 total shareholder votes.

Thus, Yu Yue's voting rights would be ~29.41%; Kuaishou's ~50.25%.

In essence, Yu Yue can significantly influence shareholder meetings with minimal equity but cannot unilaterally challenge Kuaishou's control.

This reflects a classic corporate spin-off strategy: offer technical leaders substantial operational incentives and bargaining power while preserving the parent company's ultimate control.

However, under HKEX's current weighted voting rights (WVR) rules, special voting rights beneficiaries must collectively hold at least 10% economic interest at the time of initial listing. The exchange may accept a lower threshold (≥5%) at its discretion, provided the beneficiary's stake corresponds to a market value of ≥HK$4 billion.

Even if Yu Yue exercises all 1% options, his economic interest would be 4%, below the 5% exception threshold. If Kling issues new shares during its IPO, this ratio would dilute further.

Thus, the current special voting rights resemble a private placement-stage temporary agreement rather than a confirmed go public structure.

Pre-IPO, Kling has several adjustment paths: add new WVR beneficiaries, allocate additional shares to Yu Yue, abandon WVR for 'one share, one vote,' or revise the listing entity, venue, or overall plan.

Realistically, the first two options are more plausible. Whichever path is chosen, Kling must realign equity, incentives, and control before filing for listing.

Kling's future power structure is clear: Kuaishou holds controlling equity, Yu Yue wields partial special voting rights, and external investors gain contractual constraints over major decisions.

While manageable in ordinary internet businesses, this governance structure may impose competitive costs in generative AI.

Model firms require continuous investment in training compute, data, and R&D personnel. Each training round may enhance capabilities or merely increase costs; betting on a new technical path may create generational advantages or prove misguided months later.

Technical founder-led companies can directly translate such judgments into corporate decisions. Yu Yue must instead convert technical insights into capital proposals and persuade Kuaishou, the board, and external investors to accept them.

Kling's client structure amplifies this pressure.

In Q1, overseas revenue accounted for nearly 80% of Kling's total, with API revenue comprising ~60%. Within the API segment, super-large clients contributed over 80% of income.

A significant portion of Kling's revenue comes from backend calls by AI aggregation platforms, which typically integrate multiple model providers. The provider generating usable footage at lower costs gains more calls.

As of October 15, in Artificial Analysis's audio-included text-to-video rankings, Tongyi Wanxiang Wan 3.0, Utopai X, and Seedance 2.5 ranked top three, while Kling 3.0 Omni and 3.0 Pro placed 15th and 16th.

Pricing-wise, Kling lacks a clear edge. On the call platform Fal, Kling 3.0 Pro charges $0.168 per second for 1080p audio-included text-to-video, compared to Wan 3.0's $0.20 and H3 Max's $0.16.

Kling opened 4.0 Flash beta access to select annual subscribers in late September, with a full rollout planned this month.

Whether the new version retains major clients amid price wars and technical competition, how to allocate model investments, and how to structure equity and compensation for competitive teams—all these issues land on Yu Yue's desk, where adjustments to one end will ripple across the entire system.

At a post-money valuation of $18 billion, Kling is now worth ~1.1 times Kuaishou. For comparison, Kuaishou's market cap was ~HK$127 billion as of October 14.

Assuming full utilization of this financing round and 15% incentives, Kuaishou would retain ~68.33% of Kling, valued at ~HK$96 billion—three-fourths of its own market cap.

If Kling's reported $30 billion IPO expectation materializes (without subsequent dilution), this stake would be worth ~HK$160 billion, exceeding Kuaishou's current market cap.

Compared to its peak of ~HK$1.74 trillion at the 2021 IPO, Kuaishou's market cap has evaporated by ~HK$1.61 trillion. On October 13, its stock price fell to HK$28.24 intraday, a record low.

As Kling grows more valuable, Kuaishou becomes cheaper. How much imagination does the market leave for Kuaishou's legacy business?

Kuaishou's growth increasingly relies on Kling. In Q2, group revenue grew just 1.4% YoY, adding RMB 489 million. Kling contributed over RMB 850 million in revenue, up >200% YoY, accounting for >RMB 560 million of the group's total revenue increase—surpassing the overall group increment.

Within legacy businesses, advertising growth slowed, and live streaming continued to contract. Q2 advertising revenue grew 4.4%, while live streaming revenue fell 13.5%—the former's increment barely offset the latter's decline. Other services (including Kling) grew 18.5%, barely sustaining the group's marginal revenue increase.

Daily active users (DAUs) and profits also faced pressure. Q2 DAUs grew just 0.8% YoY, with core active user growth nearly stagnant. Adjusted net profit fell 30.3% YoY to RMB 3.913 billion.

According to Nomura's August data on China's top 50 apps by usage time, gaps in advertisers' budget allocation among platforms widened further.

ByteDance-owned apps collectively accounted for 41.8% of usage time, up from 34.2% YoY; Hongguo short dramas alone captured 5.9%, surpassing Kuaishou's main app (3.7%). User attention continues to shift toward ByteDance's ecosystem.

Kuaishou is also doubling down on short dramas to save itself. In Q2, short drama clients' ad spend on Kuaishou grew >100% YoY. Platform supply surged >5x from January to June. Yet this failed to lift overall ad growth into double digits.

For comparison, Tencent's marketing services revenue grew 22% in Q2, outpacing Kuaishou by 17.6 percentage points.

The e-commerce business also finds it difficult to make up for the growth gap alone. In the second quarter, the scale of active paying buyers on Kuaishou's e-commerce platform remained basically unchanged on a quarter-on-quarter basis, while the number of newly settle in merchants increased by approximately 10% quarter-on-quarter. The transaction volume and marketing investment growth rate of top brand merchants were higher than the platform's overall average.

With a lack of growth in existing businesses and a contraction in group profits, it is difficult for KwaiDissi's (Kling) financing valuation to solely support Kuaishou's stock price.

After the spin-off, Kuaishou can still share in KwaiDissi's growth. However, if investors can directly invest in KwaiDissi, they will no longer need to bet on video models through Kuaishou.

By then, the more expensive KwaiDissi becomes, the more pointed the market's questioning of Kuaishou will be: Apart from holding this stake in KwaiDissi, what else can investors gain from holding Kuaishou itself?