"AI Entry Battle" - ByteDance has become a "forcing trend," Alibaba and Tencent "cannot afford to lose"

The AI competition among China's internet giants is rapidly evolving from a simple contest of model parameters into a life-and-death struggle for "default entry." With ByteDance making simultaneous efforts in AI cloud infrastructure and consumer applications, its aggressive pricing strategy and vast traffic ecosystem have created substantial pressure on Alibaba and Tencent, forcing them to increase investments to defend their core territories by the strategic turning point of 2026.

According to a report by the Financial Times on Wednesday, ByteDance's Volcano Engine has quickly grown to become China's second-largest AI infrastructure provider, with a market share of 13% in the "AI cloud" sector in the first half of 2025, second only to Alibaba's 23%. This data indicates that ByteDance is leveraging its first-mover advantage in the AI field to disrupt the long-standing cloud market dominated by Alibaba, Tencent, and Huawei.

On the consumer application front, ByteDance itself controls users' time. The public account "Yanwai Zhiyi" points out that the most frightening aspect of the AI entry point is its position being ahead of search; once users get used to asking AI first, traditional apps will face the risk of being marginalized, which is the fundamental reason why Tencent and Alibaba "cannot afford to lose."

Goldman Sachs has defined 2026 as a "critical year" for China's internet giants in its latest research report. In the face of ByteDance's Doubao App surpassing 100 million daily active users and having the highest daily Token consumption in the country, Alibaba and Tencent must significantly increase their capital and operational expenditures on AI for the consumer end. Goldman Sachs noted that ByteDance's profit scale reached $50 billion in 2025, surpassing Tencent's $36 billion and Alibaba's $15 billion, providing solid support for its aggressive expansion in the AI field.

The change in this competitive landscape not only concerns technological routes but also directly affects the pricing logic in the capital market. Goldman Sachs emphasizes that the market will no longer simply reward "visions" in 2026 but will focus on profit growth and the realization of new narratives. As the "super entry" battle rewrites traffic distribution and the internet profit pool, a defensive and offensive battle around "entry win rates" and business closed-loop capabilities has become the core window for investors to observe the revaluation of Chinese tech stocks.

ByteDance's "AI Cloud" Breakthrough: 13% Share Disrupts the Tripod

According to IDC data cited by the Financial Times, ByteDance is actively promoting diversification in the cloud market by leveraging its advancements in the AI field. In the first half of 2025, ByteDance accounted for nearly 13% of China's AI cloud service revenue, valued at approximately $390 million, second only to Alibaba's 23%, surpassing Tencent and Huawei. Although its overall market share in the cloud sector is only about 3%, ByteDance has established a significant advantage in the rapidly growing AI service segment.

According to company employees, clients, and competitors, ByteDance's Volcano Engine has rapidly expanded by increasing its sales team and "undermining" competitors with pricing advantages in recent months. The core of its strategy lies in marketing products based on its vast database and computing infrastructure to enterprise clients, such as building customized AI agents using its proprietary models Charlie Dai, Vice President and Chief Analyst at Forrester, pointed out that ByteDance's growth trajectory and AI-driven strategy indicate that it could become one of the dominant players as AI demand accelerates.

This strategy is supported by substantial hardware investments. Edison Lee, Head of China Technology Analysis at Jefferies, believes that ByteDance has strong software capabilities and ample hardware resources to gain market share, and is currently in a phase of "catching up and taking share from Tencent and Huawei." In contrast, Tencent has stated that it will prioritize GPU resources for internal use, while Huawei has scaled back its ambitions for AI cloud over the past year, focusing instead on directly selling Ascend chips to customers.

Entry Equals Pricing Power: The "Forced" Logic Under Internet Iron Laws

The rise of ByteDance is not just a numbers game of market share, but a fundamental reconstruction of user behavior habits.

A recent article from the public account "Yan Wai Zhi Yi" deeply analyzes the essence of this "entry shift." The article points out that there is an iron law in the Chinese internet industry: the shift of entry points is never announced in advance. Just as portals were replaced by search engines and PCs were replaced by mobile phones, the most frightening aspect of the AI entry point is that it stands even further ahead than search.

The entry war equals the redistribution of screen time. When AI can complete tasks for you: the frequency of "searching" will decrease, while the time spent on "content/social/transactions" will increase, and the winner will capture both advertising and transaction cash flows.

"Yan Wai Zhi Yi" emphasizes that once users become accustomed to asking AI first, the original apps, although they will not disappear, will be forced to step back. In the history of the internet, taking this step back often means "never returning." This is the root of Tencent and Alibaba's anxiety—they are not looking to attack, but are being pushed by this trend.

Currently, ByteDance has virtually no competitors in the short term. Its advantage lies not in the model itself, but in its control over users' time. When users are scrolling through Douyin in a state of "not sure what to do," AI (like Doubao) can pre-arrange the next steps.

Goldman's judgment aligns with this: the watershed for Chinese AI is not in the model, but in whether "the entry can be closed-loop." Whoever can stand at the moment the user makes a decision will hold the new pricing power.

Goldman Sachs Outlook for 2026: The Game of Super Entry and Capital Expenditure

In its latest lengthy research report, Goldman Sachs detailed the "AI super entry" battle. Goldman believes that 2026 is not a "model competition," but a life-and-death battle for the "default entry." Whoever secures the default entry will rewrite traffic distribution, advertising budgets, transaction loops, and even the entire internet profit pool.

The report points out that investors should focus on three "super entry paths": system-level mobile assistants/OS-level agents, social-level IM agents, and transaction-level application agents. What ByteDance is doing is a system-level "background agent," with Goldman specifically naming capabilities like "Phone Assistant," which handle trivial tasks in the background and rewrite human-computer interaction Data shows that Doubao's DAU has surpassed 100 million, and its average daily token consumption ranks first in the country. This is no longer an experiment, but an efficient distribution machine.

In the face of ByteDance's offensive, Goldman Sachs predicts that Alibaba and Tencent will make significant strategic shifts in 2026. Tencent's advantage lies in WeChat's potential to serve as an agentic assistant, relying on the mini-program ecosystem to create task loops; while Alibaba emphasizes the integration of the Tongyi Qianwen App's "agent capabilities for goods and services" with "local life/3D world model."

This means that the giants must increase their investment in AI To-C (capital expenditure + operating expenditure), while also focusing more on defending their core market positions—Alibaba's leading e-commerce GMV and Meituan's top scale in in-store and travel services.

Investment Logic Reconstruction: From Model Competition to Business Loop

For investors, the stock selection logic in 2026 will undergo fundamental changes. Goldman Sachs emphasizes that the market will not reward "visions" first; 2026 will be a year more focused on Alpha (excess returns), driven by profit growth and new narratives, rather than mere valuation expansion.

Goldman Sachs advises investors to use "entry win rate" rather than "model sentiment" to determine positions, focusing on hard indicators such as system-level cooperation, embedded depth in IM, progress in transaction loops, and migration of advertising products.

A true moat equals "default entry + closed-loop tasks + low-cost reasoning." Without a closed loop, AI is just a more expensive search box; without cost advantages, the larger the scale, the faster the losses.

This AI super entry battle does not require "everyone to win"; it only needs one default entry, while others may become "called functions." This is the harsh truth that 2026 will be a "critical year" for China's internet