A collapse in market confidence is often harder to reverse than operational difficulties.
A better-than-expected earnings report ironically became the catalyst for a sharp decline in Kuaishou-W's stock price. On March 25, Kuaishou-W released its full-year and fourth-quarter 2025 results: full-year revenue increased 12.5% year-over-year to 142.8 billion yuan, while adjusted net profit grew 16.5% to 20.6 billion yuan. For Q4 2025, revenue was 39.57 billion yuan and adjusted net profit was 5.46 billion yuan, representing year-over-year growth of 11.8% and 16.2% respectively, with core performance metrics exceeding market expectations across the board.
However, the capital market responded with an extreme reaction: the day after the earnings release, Kuaishou-W's Hong Kong stock opened sharply lower, plunging over 14% in a single day. This marked its largest drop in nearly 11 months, wiping out hundreds of billions in market capitalization. Its total market cap fell below 200 billion Hong Kong dollars, a nearly 90% decline from its peak market value of 1.738 trillion Hong Kong dollars at the time of its listing. This raises the question: why would a fundamentally sound report showing growth in both revenue and profit amplify market division and trigger a panic sell-off?
The core issue is the failure of the high-growth narrative that Kuaishou-W had relied upon. Against the backdrop of peaking short-video traffic, the twilight of its core "Laotie" community economy, and slowing growth in its main businesses, Kuaishou-W announced it would increase its 2026 capital expenditure to 26 billion yuan, up from 15 billion yuan in 2025. This all-in bet on AI is a high-stakes gamble: success could potentially create a new Kuaishou-W, but failure could drag the company into an abyss due to massive costs, causing it to miss the AI era entirely.
**The Twilight of the 'Laotie' Economy**
The capital market's negative vote was not impulsive but reflected the weakening of the three core pillars that previously supported Kuaishou-W's high valuation: user growth, e-commerce expansion, and its live-streaming foundation. All three showed poor performance in 2025. As a traffic-driven platform, Kuaishou-W's primary revenue streams—advertising, e-commerce, and live-streaming—depend on user scale and engagement time. In 2025, Kuaishou-W's traffic红利 (dividend) hit a ceiling, and its growth engine nearly stalled.
The financial report shows that Kuaishou-W's average daily active users (DAU) in 2025 were 410 million, a mere 2.76% increase year-over-year. Average monthly active users (MAU) reached 725 million, up just 2.11%. Looking at a longer trend from 2021 to 2025, Kuaishou-W's DAU growth rate slid from 15.58% down to 2.76%, while MAU growth slowed from 12.68% to 2.11%, approaching stagnation. More concerning was the quarter-on-quarter decline: DAU figures for the four quarters of 2025 were 408 million, 409 million, 416 million, and 408 million respectively. Impacted by the fade of summer holiday traffic红利 among other factors, Q4 saw a loss of 8 million users compared to Q3. The year-over-year MAU growth rate plummeted to 0.7%, the lowest since its listing in 2021. As user competition shifted from acquiring new users to battling for existing ones, the growth logic Kuaishou-W depended on fundamentally changed.
First, the live-streaming business, once the backbone, has now become a segment experiencing negative growth. In Q4 2025, Kuaishou-W's live-streaming revenue was 9.655 billion yuan, a 1.9% decrease from 9.846 billion yuan in the same period last year. For the full year, live-streaming revenue was 39.087 billion yuan, growing only 5.5%, a rate significantly lower than other business lines. This indicates that traditional entertainment live-streaming is nearing its peak, user growth has stalled, and the "Laotie" gifting model that Kuaishou-W was built on has lost its former glory.
Second, the e-commerce business, which had been the primary growth engine in recent years, experienced a cliff-like slowdown in growth. For the full year 2025, Kuaishou-W's Gross Merchandise Volume (GMV) grew 15% year-over-year to 1.6 trillion yuan. This growth rate represents a significant annual decline compared to 78% in 2021, 33% in 2022, 31% in 2023, and 17% in 2024. In Q4 2025, e-commerce GMV grew 12.9% to 521.8 billion yuan, slowing further from the previous quarters and marking a clear departure from the previous era of frequent 30%+ high growth.
More pessimistically for the market, Kuaishou-W announced it would stop disclosing GMV figures separately starting in 2026. This is widely interpreted as an acknowledgment by management that double-digit high growth is unsustainable and the narrative of explosive e-commerce expansion has ended. In the存量 (stock) competition of the e-commerce market, facing pressure from rivals like Douyin, Taobao, and Pinduoduo, the growth ceiling for Kuaishou-W's e-commerce business is clearly visible.
The only segment still showing growth, online marketing services (advertising), also harbors hidden worries. In 2025, advertising revenue reached 81.5 billion yuan, a 12.5% increase, outperforming the broader domestic internet advertising market. However, the core driver of this growth was not market share gains but improved efficiency from AI technology—generative recommendation models and intelligent bidding models contributed approximately 5 percentage points of the revenue growth. The market views this type of growth, driven by internal efficiency optimization, as having a clear ceiling, making it difficult to support a long-term high valuation. More critically, the advertising business is highly correlated with the macroeconomy. In an environment where merchants are becoming more cautious with spending, the sustainability of its growth is highly uncertain.
**Kling Carrying Kuaishou-W Forward?**
With core businesses under pressure and growth narratives weakening, Kling became the central theme of Kuaishou-W's earnings report. Management's focus on Kling has escalated from a technical investment to a full corporate strategy. In late April 2025, Kuaishou-W formally established the Kling AI Division, positioning it as a first-tier department alongside core businesses like e-commerce and commercialization, reporting directly to founder and CEO Cheng Yixiao. Kling was thus elevated from an internal tech project to a core growth engine for the company. In August 2025, after former technology head Zhang Di departed, Senior Vice President Gai Kun concurrently assumed the role of head of the Kling AI Technology Department, still reporting directly to Cheng Yixiao, further cementing Kling's status as a top-priority "Number One Project."
In Cheng Yixiao's earnings call comments, Kling's importance was repeatedly emphasized: from the vision of "striving for Kling AI to become the video generation AI application with the largest global revenue scale as soon as possible" to setting the ambitious goal of "achieving over 100% year-on-year revenue growth for Kling in 2026." Cheng Yixiao has positioned Kling as the decisive factor in this AI revolution. After user, e-commerce, and live-streaming growth faltered, Kuaishou-W desperately needs a new story to support its valuation in the capital markets.
To this end, Kuaishou-W has significantly accelerated its pace: it launched the Kling O1 and Kling 2.6 series models in Q4 2025, followed by the Kling 3.0 series in February 2026, completing a transition from single-point generation to an integrated, professional creation system.
Correspondingly, Kling established a commercial闭环 (closed loop) within a year. Revenue for Kling AI across the four quarters of 2025 was 150 million yuan, 250 million yuan, 300 million yuan, and 340 million yuan respectively, totaling approximately 1.04 billion yuan for the full year, far exceeding the initial 2025 target of $60 million USD. Although quarter-on-quarter growth slowed, Kling's commercialization efficiency is still accelerating: in December 2025, Kling AI's monthly revenue surpassed $20 million USD, corresponding to an Annual Recurring Revenue (ARR) of $240 million USD. By January 2026, its ARR had further climbed to over $300 million USD.
In terms of user base, Kling also achieved counter-trend expansion. By the end of 2025, Kling AI's global user base exceeded 60 million, having generated over 600 million videos. It provides API services to more than 30,000 enterprise clients and developers, covering industries like advertising/marketing, film/animation, and game production. It is reported that in January 2026, Kling AI's Monthly Active Users (MAU) surpassed 12 million, with professional creators (P-end) and enterprise clients (B-end) contributing nearly 70% of the revenue, forming a stable commercial foundation. Kling is one of the few AI video products that has established advantages in model capability, revenue scale, and user volume simultaneously.
Of course, Kling's core value also lies in its渗透 (infiltration) into Kuaishou-W's entire commercial ecosystem like a utility, forming a dual闭环 of "external commercialization + internal full-scenario enablement" that pure-play AI companies cannot replicate. It has become a core engine for reducing costs, increasing efficiency, and building competitive moats for the main business.
In advertising, AI technology enables the full chain: the fully automated UAX ad delivery product achieved nearly 80% penetration in consumption for non-ecommerce marketing services, with 90% automation across the delivery process. In Q4 2025, consumption driven by AIGC marketing creatives reached 4 billion yuan, with overall conversion rates improving 20%-30% compared to manually created素材 (materials).
In e-commerce, continuous iteration of the end-to-end generative search architecture OneSearch drove search-based order volume on the marketplace up by approximately 3%. The generative recommendation model OneRec-V2 was extended from general shelf scenarios to live-streaming rooms and short videos, boosting GMV across all channels.
Regarding content ecosystem and organizational efficiency, the self-developed multimodal large model Keye-671B enhanced the platform's video understanding capability. The TagNex tag system based on this model directly contributed to increases in both user time spent and retention rates. In live-streaming scenarios, the "AI Wanxiang" gift feature surpassed 1 million cumulative generations. The self-developed AI programming tool CoderFlicker accounted for over 40% of newly generated code at Kuaishou-W, significantly improving R&D efficiency.
This complete layout of "tool + platform + ecosystem" is precisely the core barrier that distinguishes Kuaishou-W from pure AI companies like MiniMax and Zhipu AI. The technology落地 (implementation) for AI-native companies still requires a difficult search for consumption scenarios like e-commerce, advertising, and content, whereas Kuaishou-W inherently possesses top-tier domestic traffic and commercial scenarios, which is its greatest advantage in the AI era.
**All-in on AI: A High-Stakes Gamble It Cannot Afford to Lose**
Despite Kling's impressive report card, the capital market remained unmoved, even expressing pessimism through the stock plunge. The core reason is that while Kuaishou-W's gamble appears promising, the path is long and fraught with challenges, and the underlying logic of the capital market has subtly shifted.
When the AI wave swept the globe in 2023, the capital market was willing to buy into any AI-related narrative, even mere concepts, driving stock prices soaring. But by 2026, investors have changed their focus to short-term profit quality and cash flow, adopting an extremely cautious stance towards long-term, high-investment AI stories. This is the core reason behind the collective stock decline of several Hong Kong-listed internet companies after they disclosed increased AI capital expenditures recently—even heavyweight constituents of the Hang Seng Tech Index like Tencent and Alibaba received negative market feedback. For Kuaishou-W, with its already weak fundamentals, this shift in market sentiment is particularly damaging.
Under this market logic, the risks of Kuaishou-W's AI gamble are being scrutinized and amplified.
First, the mismatch between investment and output has caused excessive concern. Kuaishou-W's CFO, Jin Bing, stated on the earnings call that the group's 2026 capital expenditure would reach approximately 26 billion yuan, the vast majority allocated to computing power support for the Kling large model and other foundational models. This includes inference computing demands from Kling's expanding user base and training computing reserves needed for model upgrades. This means Kuaishou-W plans to add 11 billion yuan in investment on top of the already high 2025 base of 15 billion yuan. The full-year capital expenditure of 26 billion yuan far exceeds market expectations of around 18 billion yuan. Notably, Kuaishou-W's full-year 2025 adjusted net profit was only 20.6 billion yuan, meaning the 2026 capital expenditure will exceed the total net profit for the entire previous year.
In contrast, Kling's revenue base is still small: its full-year 2025 revenue was only about 1.04 billion yuan, accounting for less than 1% of Kuaishou-W's total revenue. Even if it meets Cheng Yixiao's expectation of doubling revenue in 2026, Kling's annual revenue would only reach around 2 billion yuan. Essentially, Kuaishou-W is betting 26 billion yuan of certain investment on a potential revenue increase of roughly 2 billion yuan. Setting aside the time lag between computing power deployment and commercial monetization, where unit computing power cannot immediately generate revenue, even if Kling maintains rapid growth, the incremental revenue it brings would be far from sufficient to cover the massive depreciation and R&D costs of the computing power.
Second, amid intense industry competition, Kling's breathing room is being squeezed, with ByteDance posing a threat domestically and strong rivals encircling it overseas. In February of this year, ByteDance's Volcano Engine launched Seedance 2.0, which achieved viral spread凭借 (relying on) powerful model capabilities, being called the "DeepSeek moment" for AI video generation, with even Elon Musk remarking on its "astonishing development speed." According to third-party data, Seedance 2.0's MAU reached around 45 million during the Spring Festival, nearly four times that of Kling AI during the same period. More critically, Seedance 2.0 is deeply integrated into ByteDance's product matrix, launching on Jiemeng AI and Doubao, and officially landing on CapCut (the overseas version of Jianying) in March, initially for users in countries like Brazil, Indonesia, and Thailand, with plans for gradual global rollout. Given that 70% of Kling's revenue comes from overseas markets, a direct confrontation with Seedance seems inevitable.
Although Cheng Yixiao stated on the call that Seedance 2.0's multimodal input technical direction aligns with that of Kling O1, validating Kuaishou-W's foresight, it's undeniable that ByteDance possesses world-class traffic ecosystems, product operation capabilities, and R&D strength—negotiating with such a competitor puts Kling at a disadvantage from the start. Furthermore, Kling must also face pressure from giants like Google and xAI in overseas markets, both in terms of pricing and technology. If Kling chooses to cut prices to gain market share, it will further dilute the gross profit per unit of computing power, falling into a "revenue growth without profit growth" dilemma. If it maintains prices, it risks losing market share in the competition, a precarious position.
The most critical point is that the profit logic Kuaishou-W has barely maintained is poised for a complete reversal, with profit erosion seeming inevitable. Over the past two years, Kuaishou-W's sustained profitability was aided not only by core business growth but also heavily by stringent cost control. The financial report shows that in 2025, the ratio of cost of sales to total revenue decreased to 45% from 45.4% in 2024; the ratio of selling and marketing expenses to total revenue decreased to 29.6% from 32.4% in 2024; overall employee costs grew only 4%, and promotion/marketing expenses remained flat year-over-year. It was this极致 (extreme) expense control that allowed Kuaishou-W to marginally improve its profit margin while increasing AI R&D investment.
Simultaneously, Kuaishou-W's R&D expenses reached 14.5 billion yuan in 2025, an 18.8% increase, and equipment depreciation costs grew 10%. If the massive 26 billion yuan capital expenditure materializes in 2026, coupled with continuously growing AI R&D investment, depreciation costs and R&D expenses will expand further, undoubtedly pulling Kuaishou-W back into the quagmire of losses.
Nomura Securities pointed out in a research report that Kuaishou-W's 2026 performance outlook is weak, expecting revenue growth of only about 4%, with adjusted net profit projected to decline 15%-18%. Huatai Securities also believes Kuaishou-W's 2026 profit will decline year-over-year due to AI investments, with soft advertising momentum and increased AI capital expenditure putting continued pressure on profit margins.
In the eyes of the capital market, Kuaishou-W's all-in AI bet is essentially "gambling an uncertain future with certain profit erosion." OpenAI's shutdown of Sora has already created significant恐慌 (panic) in the capital market regarding the prospects of the AI video generation sector. How can Kling guarantee it won't meet a similar fate? This pessimistic expectation of profit decline,质疑 (doubts) about the return on investment, and collective pessimism about the sector's前景 (prospects) ultimately converged into the panic selling of Kuaishou-W stock the day after its earnings report. A collapse in market confidence is often harder to reverse than the operational difficulties themselves.
Comments