26 billion gamble on AI and capital pricing anxiety. Kuaishou profits hit a new high but stock price drops 14%

Ask AI · Kuaishou profit hits a new high, but the stock price still plunges—Is the AI gamble the main reason?

China Jing News reporter Li Jing, Beijing report

When a company’s annual financial report showing both revenue and profit hitting historical highs is released, the market responds with a 14.04% plunge in the share price. This is what happened to Kuaishou (01024.HK) after it released its 2025 annual results.

On the evening of March 25, 2026, Kuaishou released its 2025 full-year and fourth-quarter financial results. The data showed that Kuaishou’s 2025 full-year revenue was RMB 142.776 billion, up 12.5% year over year, and its adjusted net profit was RMB 20.647 billion, up 16.5% year over year. Both revenue and profit hit new highs.

However, on the next day (March 26), Kuaishou’s share price fell by more than 9% straight after the open, and then kept sliding. By the close on March 26, the stock price was HKD 45.6 per share, down 14.04%, with its market value falling below HKD 200 billion.

In the earnings call, Kuaishou management said it would pursue a long-term strategy to empower all businesses with AI technology, and planned to invest RMB 26 billion in 2026 in capital expenditures for AI computing power infrastructure.

Although Kuaishou is actively embracing AI, capital markets have raised questions about the sustainability of its profits and the efficiency of rolling out new growth engines. Yuan Shuai, an expert from the China Jing Media think tank and deputy secretary-general of the Zhongguancun Internet of Things Industry Alliance, said directly to reporters from China Business Journal: “Behind this lies anxiety in the market’s dual pricing logic for Kuaishou, reflecting the general anxiety of content platforms in the face of inventory competition across the internet industry—that is, as the industry moves into a phase of stock competition.”

Basic business under pressure

Kuaishou’s 2025 annual financial report, to a certain extent, exposes the reality that the driving force behind its traditional growth engine has weakened.

The financial report shows that in 2025, Kuaishou achieved full-year revenue of RMB 142.776 billion, up 12.5% year over year. Of this, online marketing service revenue was RMB 81.462 billion, live streaming revenue was RMB 39.087 billion, and other services revenue was RMB 22.227 billion.

Among these, the highly watched e-commerce business—despite still reaching GMV of RMB 521.8 billion in the fourth quarter of 2025—its year-over-year growth rate of 12.9% is no longer comparable to the high growth seen during the industry’s boom period.

What has further shaken market confidence is that Kuaishou announced it would follow industry trends and stop disclosing quarterly and annual GMV data separately.

“Kuaishou’s decision to stop disclosing GMV is actually a snapshot of the industry as a whole downplaying GMV behavior,” Yuan Shuai said. “Previously, GMV—an essential metric for measuring the scale of an e-commerce platform—was an important anchor for investors to evaluate platforms. But as the industry shifts from incremental competition to stock competition, pure GMV growth can no longer reflect a platform’s true operating quality. In the future, indicators used to assess e-commerce health will focus more on ‘quality indicators’ such as user repurchase rates, average order value, and platform monetization rates.”

Kuaishou CEO Cheng Yixiao also acknowledged in the earnings call that the e-commerce business faces pressure in 2026. The e-commerce strategy for 2026 is to return to the essence of content e-commerce, focusing on supply-side reform and buyer penetration.

Cheng Yixiao said it would invest more resources to support brands and business merchants in the industrial sector, and strengthen its exploration and understanding of users’ interest in e-commerce content. This suggests that Kuaishou’s e-commerce strategy is shifting from a rough, growth-at-any-cost approach focused on transaction scale to a more in-depth development model focused on cultivating an ecosystem and improving conversion efficiency.

Liang Zhenpeng, a senior industry observer, believes this is a sign that the industry is moving toward maturity, and investors need to shift their perspective from focusing on “scale” to focusing on “profitability health.”

Meanwhile, another Kuaishou cash-cow business—live streaming—has all but hit a standstill in terms of growth. In 2025, full-year live streaming revenue was RMB 39.09 billion, up 5.4% year over year, which is almost the same as the RMB 39.05 billion live streaming revenue in 2023.

Yuan Shuai analyzed: “This seems to signal that the ceiling for the business model of the live-streaming business has already emerged. Behind it lies the widespread predicament of saturated paid paid-for show-style live-streaming and fierce competition in e-commerce live streaming.”

The AI gamble

As the growth curves of existing businesses level off, Kuaishou is pinning its hopes for the future on artificial intelligence (AI)—especially its in-house video generation large model, “Keling AI.”

Kuaishou’s total capital expenditure in 2026 is expected to reach as much as RMB 26 billion, up roughly RMB 11 billion from 2025. Kuaishou’s chief financial officer, Jin Bing, clearly stated that these investments will be mainly used for computing power on the Keling large model and other foundational large models, as well as for building data centers.

Behind the massive investment is management’s firm belief that its AI strategy can produce a “dual-wheel drive” effect.

On the one hand, it hopes that AI capabilities will feed back into Kuaishou’s traditional businesses. In the call, Cheng Yixiao explained in detail that in addition to video generation, Kuaishou will continue investing in generative recommendation large models and multimodal understanding large models to optimize content distribution, ad placement, and e-commerce conversion efficiency on its main platform. Its in-house multimodal foundational large model, “Keyi,” has been used for video parsing and user-behavior reasoning. In addition, Kuaishou plans to roll out AI agents in e-commerce and marketing scenarios to achieve functions such as automated ad placement and intelligent customer service, directly targeting “driving greater commercial value.”

On the other hand, it expects Keling to open up new incremental volume. According to disclosures from Kuaishou management, the commercialization of Keling AI is accelerating: its annualized run rate (ARR) for January 2026 has already exceeded US$300 million. “We are very confident that this year Keling’s revenue will achieve more than 100% year-over-year growth.”

Cheng Yixiao also shared application examples of Keling AI in top-tier film and television production (such as special effects for 《Taiping Nian》), and emphasized that its model capabilities rank ahead in authoritative ranking lists, establishing differentiated advantages among professional creators and enterprise clients.

Jin Bing further revealed that Keling AI has already surpassed 60 million global users, providing API services for more than 30,000 enterprises, and that its app ranks number one by downloads in more than 40 countries and regions.

However, what is worth noting is that OpenAI recently adjusted its video generation model strategy, shutting down the widely popular video generation app Sora and related APIs. Kuaishou clearly cannot afford to give up Keling.

“Kuaishou apparently has no possibility of choosing to follow,” Yuan Shuai said. “This is not stubbornness about a technical path, but a necessary choice driven by its own business needs and strategic positioning. For Kuaishou, a video large model is the underlying support for its content ecosystem; giving up the video large model is equivalent to giving up core control over the content ecosystem.”

At the beginning of 2026, ByteDance released its video generation model Seedance 2.0, drawing market attention. People in the industry told reporters: “OpenAI’s abandonment of Sora may also be related to this, because ByteDance has video library data advantages for training video generation models that other companies can hardly match.”

Yet the huge mismatch between AI investment and output is the key reason the market is not buying in. Yuan Shuai said: “The situation that Keling AI’s annual revenue has only just broken RMB 1 billion, yet it needs to invest in computing power on the scale of billions of yuan, makes its ROI (return on investment) cycle the focus of market attention.”

Liang Zhenpeng also believes that Kuaishou’s AI currently is more of an efficiency tool serving Kuaishou’s core business—by optimizing areas such as content recommendations, ad placement, and e-commerce product selection to feed back into the main business—but it has not yet become an independent second growth curve that can carry Kuaishou’s future.

For Kuaishou’s AI gamble, institutions have also shown anxiety. After Kuaishou released its financial report, multiple investment banks such as Furi, Citi, and Morgan Stanley mostly maintained “buy” ratings, but they collectively cut their target prices by large margins, ranging from 20% to 30%. The reasons for their cuts generally point to a slowdown in growth expectations and AI spending eroding short-term profits. For example, a research report from Huatai Securities noted that AI spending will increase depreciation and compensation expenses, or may cause Kuaishou’s profits in 2026 to decline year over year.

(Editor: Zhang Jingchao, Reviewer: Li Zhenghao, Proofread by: Yan Jingning)