Research Reports

Overview

Summary
Values of AI servers and corresponding computing cores GPGPU (General-Purpose Computing on Graphics Processing Unit) in parallel computing as LLM (Large Language Model), which excels in NLP (Natural Language Processing), and diffusion model, which excels in image processing, march towards mass commercialization, have become essential drivers for the human society in improving productivity through technical advancement.

Among which, AI servers possess a higher degree of customization compared to that of cloud servers. Taiwanese ODMs, including Foxconn Industrial Internet, Inventec, MiTAC, Wistron, Wiwynn, and Quanta, are once again portraying essential roles in the supply chain for AI servers with their capabilities in customization, while server OEMs such as Dell, Gigabyte, HPE, Lenovo, and Supermicro are relying on AI server solutions as important growth dynamics in the global server market amidst the post-pandemic era.

Table of Contents
1. AI Server Supply Chain
   (1) ODM Direct for AI Servers
   (2) OEM Mode for AI Servers

2. Three Major Suppliers of GPGPU for AI Servers: NVIDIA, AMD, Intel
   (1) NVIDIA’s Data Center GPU
   (2) AMD’s Instinct MI Accelerator
   (3) Intel’s Data Center GPU Flex/Max

3. TRI’s View
   (1) Production and sales of AI servers can be largely divided into ODM direct and OEM mode, with the
   major difference being the fewer segments of production and sales for the former
   (2) Three Major CPU Suppliers NVIDIA, AMD, Intel to Expedite Core Architecture Designs, Hardware
   Configurations, Packaging Technology, and High-Speed Interconnection Technology for Respective
   GPGPU Products

＜Total Pages：11＞