The world is gradually transitioning into the era of hybrid artificial intelligence (AI), fostering the emergence of edge computing and the development of end devices capable of directly running generative AI applications. Since the fourth quarter of 2023, global smartphone brands have introduced their AI phones, offering users diverse experiences with generative AI applications. The development of large language models (LLMs) has become a focal point for smartphone brands seeking AI dominance, alongside the enhancement of smartphone chip computing capabilities, expected to be a key trend in AI phones this year.
Table of Contents
1. Background: AI Phones Enable On-Device Computing1.1 Generative AI Transforms User Experiences via LLMs
1.1.1 Mainstream Generative AI Applications of: Image and Text Generation, Productivity Tools
1.1.2 LLMs: Utilizing Model Parameters, Tokens, and Penalty Mechanisms to Enhance Content Accuracy and User Relevance
1.2 Rise of Edge Computing Drives Mobile Chip Performance for Local AI
1.2.1 Emergence of Hybrid AI: Convergence of Cloud AI and Edge AI Models
1.2.2 Mobile Chips Redesigned for On-Device AI: Specification Upgrades and Configuration Changes
1.2.3 10 Billion Parameters: Mobile AI's Current Limit Due to Chip and Memory Constraints
2. Brands Focus on Developing LLM and Diverse AI Applications
2.1 Smartphone Brands Shifting from Hardware Specifications to Proprietary LLM Development
2.1.1 Android Camp: Pioneering AI Phone Launches
2.1.2 iOS Camp: Building Competitiveness with Proprietary LLMs as Market Followers
2.2 Software Services: Focusing on End-user Applications and Upgrading Proprietary OS
3. Opportunities and Challenges: Memory specs as a Bareeir to AI Development
3.1 Opportunities: AI Trends Boost Sales, Chip Supply Chain Benefits from Hardware Upgrades
3.1.1 Smartphone Brands: Generative AI Optimizes User Experience, Aligning with Social Interaction Needs of Users
3.1.2 Mobile Supply Chain: Suppliers Benefit from Specs Upgrades in Early AI Phone Development
3.2 Challenges: Memory Limits as Key Barrier for AI Phone Development
4. The Analyst's Perspective
4.1 Mobile Brands Enhance Generative AI with LLM
4.2 Generative AI Spurs Phone Hardware Upgrades, but Memory Limits Future Development
Appendix
List of Companies
List of Tables
- Table 1: Comparison of Flagship Chip Specifications between Qualcomm and MediaTek (2022 vs. 2023)
- Table 2: Comparison of LLMs in Android Smartphone Camp
- Figure 1: Qualcomm's Hybrid AI Concept
- Figure 2: Distribution of Generative AI Applications and LLM Parameters
Companies Mentioned (Partial List)
A selection of companies mentioned in this report includes, but is not limited to:
- Amazon
- Apple
- Arm
- Baidu
- Ferret
- Gamma
- MediaTek
- Meta
- Microsoft
- Microsoft
- OpenAI
- Oppo
- Qualcomm
- Samsung
- TikTok
- TSMC
- Vivo
- X (Twitter)
- Xiaomi
Methodology
Primary research with a holistic, cross-domain approach
The exhaustive primary research methods are central to the value that the analyst delivers. A combination of questionnaires and on-site visits to the major manufacturers provides a first view of the latest data and trends. Information is subsequently validated by interviews with the manufacturers' suppliers and customers, covering a holistic industry value chain. This process is backed up by a cross-domain team-based approach, creating an interlaced network across numerous interrelated components and system-level devices to ensure statistical integrity and provide in-depth insight.
Complementing primary research is a running database and secondary research of industry and market information. Dedicated research into the macro-environmental trends shaping the ICT industry also allows the analyst to forecast future development trends and generate foresight perspectives. With more than 20 years of experience and endeavors in research, the methods and methodologies include:
Method
- Component supplier interviews
- System supplier interviews
- User interviews
- Channel interviews
- IPO interviews
- Focus groups
- Consumer surveys
- Production databases
- Financial data
- Custom databases
Methodology
- Technology forecasting and assessment
- Product assessment and selection
- Product life cycles
- Added value analysis
- Market trends
- Scenario analysis
- Competitor analysis
LOADING...
