Summary of Key Points
AI-powered smartphones have finally moved from the stage of hype to practical implementation: For the first time, the state has registered system-level AI services provided by seven smartphone manufacturers, marking the official opening of the door to legal commercial use. The DouBao smartphone served as a turning point in the industry, promoting the widespread adoption of intelligent features but also highlighting potential security risks related to user permissions. Model manufacturers (such as ByteDance and OpenAI) and traditional smartphone manufacturers (like Huawei and Xiaomi) are each displaying their capabilities in the race for control over future traffic flows. The business models are undergoing a reshuffle, with smartphone manufacturers and internet platforms competing for control over service distribution rights. While the technical approaches are becoming clearer, the ultimate outcome remains to be seen.
1. Regulatory Approvals: The First Step towards Legalization of AI Smartphones
The recent registration of seven AI services for smartphones by the National Internet Information Office represents the first time that regulation has extended to the system level. Previous registrations focused on standalone apps or cloud-based services, but this time it involves smartphone operating systems, which have direct access to users' private information such as photos, calendars, and device status, and can perform cross-app actions (for example, automatically opening WeChat to send messages). The significance of these approvals is that only with the necessary “licenses” can manufacturers officially incorporate AI features into their products for commercial use. Industry insiders expect that smartphone companies will begin updating their systems in the second half of this year, allowing AI functions to be rolled out smoothly. In simple terms, it’s like obtaining a driver’s license before hitting the road—system-level AI has finally gained the official permission to operate.
2. The DouBao Smartphone: A Game-Changer
Last year, the DouBao M153, a collaboration between ByteDance and ZTE, marked a turning point for AI smartphones. This device could understand user commands and automatically open apps to perform actions. For instance, if you said “book a flight to Shanghai for tomorrow,” it would directly open the booking app to fill in the details; or if you asked to compare prices for a robotic vacuum cleaner, it would search multiple platforms. Its ability to take proactive actions, unlike earlier AI phones that were limited to tasks like photo editing and translation, led to rapid sales of the prototype, with its second-hand price soaring to 7,999 yuan. However, this model encountered issues when apps like WeChat and Alipay issued security warnings due to its use of a graphical user interface (GUI) that mimicked human interaction without waiting for app-specific APIs. DouBao later suspended WeChat functionality in high-risk scenarios. This year’s second-generation DouBao (Nubia NaviX Ultra) has improved performance and reliability, with all operations requiring user authorization to ensure complete control. ByteDance’s goal is not just to sell phones but to integrate DouBao as a system plugin into smartphones, thereby gaining a stable source of traffic for potential subscription-based services.
3. The Competition for the AI Smartphone Market
The AI smartphone market is currently divided into two camps:
- Model Manufacturers: Companies like ByteDance (partnering with Nubia), OpenAI (acquiring hardware companies and collaborating with chip manufacturers like MediaTek and Qualcomm to develop processors), and JieYueXingChen (launching an intelligent operating system for smartphones). Their aim is not to sell hardware but to bind users to their services, encouraging them to use their models over time and ultimately control which platform services are used (for example, deciding whether to order coffee from Luckin or Starbucks).
- Traditional Smartphone Manufacturers: Brands like Huawei (focusing on self-developed chips to enhance computing power), Xiaomi (leveraging their IoT ecosystems to control costs with established external solutions), OPPO, and Honor (adopting a more lightweight approach where the system identifies user needs and calls upon external model services). These manufacturers have hardware and user bases but lack AI capabilities; some have invested over tens of billions in AI research with little return, leading to internal adjustments in their strategies.
In essence, model manufacturers aim to gain control over the hardware through software, while traditional manufacturers seek to combine hardware and software strengths. The winner in this competition will dominate traffic flows for decades to come.
4. The Hidden Battle for Business Models: Who Will Profit from AI Services?
The biggest change brought by AI smartphones is in the way services are distributed. Previously, users had to open apps like Taobao or Meituan on their own; now, AI can directly assist them in making purchases (for example, selecting a robotic vacuum cleaner within a certain budget). This threatens the revenue models of internet platforms, which lose opportunities for advertising and recommending products (e.g., no more membership pop-ups when ordering food). As a result, there is a struggle for control over user permissions between smartphone manufacturers and platforms. For instance, vivo wants DouBao to act solely as an executor, while ByteDance seeks full operational control; Alibaba failed to negotiate higher permissions with Alipay; Tencent once opposed a manufacturer’s use of accessible APIs for WeChat operations. However, platforms are becoming more flexible—if they don’t open up these capabilities, users might directly use AI assistants to place orders, turning apps into mere background tools.
Potential future revenue models include:
- B2B (hotels and airlines paying commissions to smartphone manufacturers),
- C2C (users subscribing to advanced AI services),
- Operators selling token packages (similar to how data plans are sold today, with users purchasing AI processing power in the future).
5. Technological Paths: Which Approach Will Be Successful?
Several technical approaches are being explored in the industry:
- GUI: AI operates apps through a graphical interface without waiting for APIs, but this approach is risky as it leaves manufacturers vulnerable to platform restrictions.
- APIs: Platforms offer APIs to AI services, but they are reluctant to open them unilaterally.
- A2A (Agent to Agent): Two AI systems communicate directly, exchanging only necessary information (e.g., specifying a delivery time for food), while maintaining data privacy. OPPO believes this approach balances the interests of all parties and offers efficient intelligent operations.
In conclusion, the A2A model seems more balanced and may become the mainstream in the future.
The ultimate outcome of the AI smartphone revolution is still uncertain, but just as the iPhone enabled humans to walk upright by freeing their hands, AI-powered smartphones could enable us to “look up at the stars” with intelligent assistants at our side. This competition has just begun.