Is Doubao AI's mobile phone competitor Xiaomi? Goldman Sachs: AI "system-level integration" faces challenges, which further validates Xiaomi's long-term competitiveness

Author: Long Yue

Source: Hard AI

Recently, ByteDance prominently released the preview version of the "Doubao AI Phone Assistant," aiming to reshape the mobile interaction experience through a large language model (LLM), attracting significant market attention.

Goldman Sachs released its latest research report on December 4, analyzing that the "Doubao AI Phone Assistant" marks the entry of AI competition into the operating system layer. However, third-party AI agents face numerous challenges in obtaining system-level permissions, user data, and cross-application connectivity, which further validates the long-term competitive strength of existing mobile giants like Xiaomi.

Doubao AI is Coming, but System-Level Integration Faces Three Major Challenges

According to the report, the "Doubao AI Phone Assistant," released by ByteDance on December 1, is a system-level graphical user interface (GUI) agent that deeply integrates the Doubao large model into the mobile operating system, aiming to achieve visual interpretation of screen content and complex task operations across applications, such as cross-platform price comparisons and automatic coupon collection demonstrated in the presentation.

The report shows that Doubao has become a popular large language model deployed by several smartphone original equipment manufacturers (OEMs) such as OPPO, vivo, and Honor. In May, the hardware AI assistant generated an average of 1.3 trillion tokens daily on Doubao (accounting for 8% of Doubao's total token consumption).

Despite demonstrating potential technically, Goldman Sachs points out that third-party AI agents face three core challenges to achieve large-scale application:

1. System-Level Operating Permissions: AI agents need to obtain permissions to read screen content, simulate user behavior, and access system services. Currently, mainstream smartphone OEMs still enjoy strategic advantages in the AI ecosystem and may not easily open up the complete system kernel and permission system to third parties.

2. System-Level Memory Capability: Mainstream OEMs can access and store comprehensive user behavior and preference data, which is crucial for training and optimizing AI agents. Third-party AIs are at a natural disadvantage in this regard.

3. Cross-Application Interface Connectivity: The cross-application operational capability of AI agents depends on the openness of interfaces from third-party internet applications. The report suggests that some large companies aiming to build closed-loop ecosystems may restrict the use of external agents, while frequent changes in the application's own UI (such as pop-ups) may also affect the stability of AI agents.

Goldman Sachs expects that in the foreseeable mid-term, the competition in the value chain among consumer AI terminals (mobile manufacturers), internet platforms, and third-party LLM/AI agents will continue.

Mobile Market Highly Concentrated, Software Disruption Difficult

Goldman Sachs emphasizes in the report that a key premise for understanding the impact of AI on the mobile industry is the market structure. The report cites data showing that the Chinese smartphone market has become highly consolidated, with the top six manufacturers (vivo, OPPO, Honor, Apple, Xiaomi, Huawei) accounting for over 90% of the shipment volume share.

This contrasts sharply with the Chinese new energy vehicle (NEV) market, which is more fragmented, leaving space for new entrants with software differentiation (such as models empowered by Huawei HIMA). In the smartphone sector, less than 10% of the market is occupied by niche brands and counterfeit models, leaving very limited space for new players to disrupt the market This means that the competitive barriers in the mobile phone industry are extremely high, and the market position of mobile giants is relatively stable.

Xiaomi's Moat: A Vast Ecosystem and Deeply Integrated AI Agent

Goldman Sachs believes that the challenges faced by Doubao AI highlight the long-term advantages of companies like Xiaomi. Xiaomi is actively promoting its "People x Vehicles x Home" full ecosystem strategy, with AI being the core that runs through it.

The report points out that Xiaomi's investment in AI is substantial, with R&D spending on AI expected to exceed 7 billion RMB by 2025, accounting for 22% of its total annual R&D expenses. The company has released professional LLMs covering various fields such as vision, audio, and speech, and is committed to applying AI capabilities to its vast ecosystem—by the third quarter of 2025, the number of globally connected AIoT devices by Xiaomi is expected to approach 1 billion.

Xiaomi's AI agent "Super XiaoAI" has made significant progress. According to QuestMobile data, "Super XiaoAI" is one of the top three native AI assistants in China by MAU, with a penetration rate of 71% among Xiaomi phone users (as of September 2025).

Goldman Sachs found through channel research that "Super XiaoAI" has achieved deep functional integration in several core scenarios:

• Social Media: Able to post notes, send chat messages, like, and share content on platforms such as Douyin, Weibo, and QQ (excluding WeChat).

• E-commerce: Able to help users find products or services on almost all mainstream platforms like Taobao, Meituan, JD.com, Pinduoduo, and Ctrip, all the way to the selection/payment page, and can confirm receipt and leave comments.

• Productivity and Content Services: Supports editing images, real-time translation, text writing, and summarization based on users' natural language input.

• Memory Capability: Able to locally remember information, knowledge bases, and user habits (such as AIoT usage, medication, and viewing time), including sensitive data, for future applications.

Goldman Sachs concludes that Xiaomi has built a strong competitive barrier through its comprehensive strengths in operating systems, hardware, a vast AIoT ecosystem, and deeply integrated AI agents. This systemic advantage makes it difficult for third-party AI agents to pose a substantial threat in the short term, further validating the long-term value of its "People x Vehicles x Home" ecosystem strategy.

This article is from WeChat Official Account "Hard AI". For more cutting-edge AI news, please click here.