Generative AI Phone Industry Whitepaper

Jointly released by Counterpoint Research and MediaTek along with other partners, including 01.AI, Alibaba Cloud Tongyi Qianwen, Baichuan-AI, Huya, Kugou, OPPO, Soul, Tencent AI Lab, Tencent Hunyuan and vivo (in alphabetical order).

PDF | 33 pages
Published date: May 7, 2024

Overview

The integration of generative AI technology is undoubtedly the main theme of the current and future development of the smartphone industry. It will open a new chapter in the evolution of smartphone intelligence and inject new momentum into the continued prosperity of the mobile internet ecosystem. In this context, Counterpoint and MediaTek, together with industry partners, have released the Generative AI Phone Industry Whitepaper which proposes the concept of generative AI phones and defines the necessary features of such devices.

• On-device or hybrid processing of foundational models: This enables the smartphone to perform complex generative AI tasks without relying solely on cloud-based servers.

• Multimodal capabilities: This allows the device to process text, images, voice and other forms of input to generate a variety of outputs, with use cases such as translations, image generation and video generation.

• A user experience that is fluid and seamless: This enables the device to provide natural interactions and intuitive responses to the user's inputs and requests.

• Hardware specifications that are capable of supporting the above capabilities: These include but are not limited to a processor built on the latest process node designed to undertake AI workload leveraging integrated or standalone neural network processing units (such as APU/NPU/TPU), large memory capacity with high bandwidth, stable and high-speed connectivity options, and hardware-level and software-level security solutions.

Counterpoint’s Definition of Generative AI Phone

The whitepaper also delves into the ecosystem of generative AI phones, discussing the AI strategies of various stakeholders, including chip vendors, smartphone OEMs, large AI model providers, and application developers, as well as the technology landscape of both hardware and software for generative AI phones. It also presents Counterpoint's predictions on the development of generative AI phones. Counterpoint expects the installed base of generative AI phones to exceed 1 billion by 2027 as generative AI technology achieves widespread adoption. This rapid increase is the result of powerful mobile computing platforms, a rich array of AI models of varying sizes, a complete set of supporting tools, and the active exploration and ever-increasing applications of AI by various industry stakeholders.

Generative AI Phone Installed Base Forecast (In Million Units)

Table of Contents

• Preface
• Chapter 1: The Onset of New Generative AI Phone Era
  • The History of Handset Smartization
  • Counterpoint’s Definition of Generative AI Phones
  • Device+Cloud Hybrid To be Mainstream in Deploying Generative AI
  • Multi-modal Capacity Key to Realize Vision of AI Agents
  • The Advantages of On-device Large AI Models
• Chapter 2: Generative AI Phone Ecosystem
  • Current Status of LLM and Forecast
    • AI Yi Model Series
    • Tongyi Qianwen by Alibaba Cloud
    • Baichuan AI
    • Tecent Hunyuan
  • Large AI Model Size Projected to Grow in Next Two Years
  • App-based UI and AI Agent UI Coexist for Years
  • Generative AI Strategies of Chip Vendors
  • Generative AI Strategies of Smartphone OEMs
    • OPPO’s Generative AI Strategy
    • vivo’s Generative AI Strategy
  • Generative AI Strategies of Developers
    • Huya
    • Kugou
    • Soul
    • Tencent AI Lab
• Chapter 3: Hardware and Software Technology Landscape of Generative AI Phones
  • Hardware Requirements for Deploying Large AI Models
  • Demand for Software Ecosystem
  • Current Smartphone SoC Platforms Support AI on Device
• Chapter 4: Generative AI Phone Forecast