"AI Entry Battle" -- ByteDance has become a "driving force under pressure", Alibaba and Tencent "can't afford to lose"

The AI competition among China's internet giants is rapidly evolving from a simple contest of model parameters to a survival battle for the "default entry point." With ByteDance making simultaneous efforts in both AI cloud infrastructure and consumer applications, and leveraging its aggressive pricing strategies and massive traffic ecosystem, it has created substantial "reverse pressure" on Alibaba and Tencent, forcing them to increase investment by the strategic turning point in 2026 to defend their core territories.

According to the Financial Times on Wednesday, ByteDance's Volcano Engine has quickly grown into China's second largest AI infrastructure provider, with its "AI cloud" market share reaching 13% in the first half of 2025, second only to Alibaba's 23%. This data shows ByteDance is leveraging its first-mover advantage in AI to break the cloud market structure long dominated by Alibaba, Tencent, and Huawei.

On the other hand, in the consumer application sector, ByteDance itself controls users' time. The WeChat public account "Yanwai Zhiyi" points out that the most daunting aspect of the AI entry point is that it is positioned ahead of search; once users get used to asking AI first, traditional apps could face the risk of being marginalized, which is the fundamental reason why Tencent and Alibaba "cannot afford to lose."

Goldman Sachs' latest research report defines 2026 as the "critical year" for China's internet giants. Facing ByteDance's Doubao app, whose daily active users have surpassed 100 million and has the highest daily token consumption domestically, Alibaba and Tencent must significantly increase capital and operational expenditure on AI for consumers. Goldman Sachs notes that ByteDance's profit scale is expected to reach $50 billion in 2025, exceeding Tencent's $36 billion and Alibaba's $15 billion, providing robust financial support for its aggressive expansion in the AI field.

The shift in this competitive landscape concerns not just technical approaches but directly impacts valuation logic in the capital markets. Goldman Sachs emphasizes that, in 2026, the market will no longer simply reward "vision" but will focus on profit growth and the realization of new narratives. As the battle for the “super entry point” rewrites traffic allocation and the internet profit pool, an offensive and defensive war centering on entry point win rate and commercial closed loop capability has become the key window for investors to reassess the value of Chinese tech stocks.

ByteDance’s “AI Cloud” Breaks Through: 13% Share Disrupts the Triumvirate

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

According to company employees, customers, and competitors, ByteDance's Volcano Engine has expanded rapidly by enlarging its sales team and weakening competitors through price advantage over the past few months. The core of its strategy is to market products to enterprise clients based on its vast databases and computing infrastructure, such as custom AI agents built using its proprietary models.

Charlie Dai, VP and Principal Analyst at Forrester, noted that ByteDance's growth trajectory and AI-driven strategy suggest that, as AI demand accelerates, it may become one of the dominant players.

This strategy is backed by massive hardware investment. Edison Lee, Head of China Tech Analysis at Jefferies, believes ByteDance has strong software capabilities and ample hardware resources to gain market share, currently in the stage of “catching up and taking share from Tencent and Huawei.” In contrast, Tencent said it would prioritize GPU resources for internal use, while Huawei in the past year has scaled back its AI cloud ambitions, focusing instead on direct sales of Ascend chips to customers.

Entry Point is Pricing Power: The “Reverse Pressure” Logic of Internet Iron Rules

ByteDance’s rise is not just a numbers game in market share but a fundamental reconstruction of user behavioral habits.

The WeChat account "Yanwai Zhiyi" in a recent article deeply analyzes the essence of this “entry point shift.” The article highlights that there is an iron rule in the Chinese internet industry: entry point transitions never come with advance notice. Just as portals were replaced by search, and PCs by mobile phones, the scariest aspect of the AI entry point is that it stands ahead of search.

Entry point battle = screen time redistribution. When AI can accomplish tasks for you: the number of times you open “search” decreases, while time spent on “content/social/transactions” increases—the winner will devour both the advertising and transaction cash flows.

“Yanwai Zhiyi” emphasizes that once users get used to asking AI first, the original apps may not disappear but will be forced to take a step back. In internet history, stepping back often means never returning. This is the anxiety at the core of Tencent and Alibaba—they are not attacking, but being forced forward by the prevailing trend.

Currently, ByteDance has almost no competitors in the short term. Its advantage isn't the model itself, but its control over users' time. When users are browsing Douyin in an "what shall I do next?" state, AI (such as Doubao) can prearrange their next move.

Goldman Sachs' view aligns: the watershed for Chinese AI is not the model, but whether the entry point can achieve a closed loop. Whoever stands at the moment the user makes a decision controls the new pricing power.

Goldman Sachs Outlook for 2026: Super Entry Point vs. Capital Expenditure

In its latest lengthy research report, Goldman Sachs breaks down the “AI super entry point” battle. Goldman Sachs believes that 2026 is not a “model race” but a life-or-death contest for the “default entry point”. Whoever wins the default entry rewrites traffic allocation, ad budgets, closed transaction loops, and the entire internet profit pool.

The report points out that investors should pay attention to three “super entry point paths”: system-level phone assistant/OS-level agents, social-level IM agents, and transaction-level application agents. What ByteDance is doing is a system-level “background agent,” with Goldman Sachs highlighting “Phone Assistant” capabilities, handling affairs in the background and reimagining human-computer interaction. Data shows Doubao’s DAU has surpassed 100 million, and its daily token consumption ranks first nationally; this is no longer an experiment, but an efficient distribution mechanism.

In the face of ByteDance's offensive, Goldman Sachs predicts Alibaba and Tencent will initiate major strategic shifts in 2026. Tencent’s advantage lies in WeChat’s potential as an agentic assistant, closing task loops via mini-program ecosystems; while Alibaba emphasizes its Tongyi Qianwen app, combining “goods and services agent capability” and “local life/3D world model.”

This means the giants must ramp up investment in AI for consumers (capital + operating expenses), while focusing more on defending their core market leadership—Alibaba with first place in e-commerce GMV, and Meituan with first in in-store/services/hospitality scale.

Reconstructing Investment Logic: From Model Races to Business Closed Loops

For investors, the stock selection logic in 2026 will fundamentally change. Goldman Sachs stresses the market will not reward “vision” first; 2026 will emphasize alpha (excess return), driven by profit growth and new narratives rather than pure valuation expansion.

Goldman Sachs recommends investors set positions by “entry point win rate” rather than “model sentiment,” closely tracking system-layer partnerships, IM embedding depth, transaction closed-loop progress, and ad product migration—these are the hard metrics.

The real moat is “default entry point + closed-loop tasks + low-cost inference.” Without closed loops, AI is just a more expensive search box; without cost advantage, the larger the scale, the faster the loss.

This AI super entry point battle does not require “everyone to win”—it only needs one default entry point. Others could end up as “functions to be called,” and this is the harsh truth of why 2026 will be a “critical year” for China’s internet.

Risk Disclosure and Disclaimer ClauseThe market has risks; investment must be cautious. This article does not constitute personal investment advice, nor does it consider the unique investment objectives, financial status, or needs of particular users. Users should consider whether any opinions, viewpoints, or conclusions in this article suit their specific circumstances. Investment based on the above is at your own risk.