Special Topic: 23rd Annual Management Conference The "23rd Annual Management Conference—Connectors of the Intelligent Era" was held in Beijing on November 27, 2025. Pan Xiaoming, Senior Vice President of Advanced Micro Devices (AMD) and President of Greater China, attended and delivered a speech.

Below is the transcript of his speech:

Today, AMD is no longer just a chip company. Instead, we have evolved into an industry leader providing comprehensive solutions of all scales and forms, centered around our chip products for high-performance computing (HPC) and artificial intelligence (AI) computing.

We firmly believe that high-performance and AI computing will help humanity tackle the world’s most critical challenges, driving social progress and human development!

Founded in 1969, AMD has now been in operation for 56 years. Today, we employ 28,000 people globally, including over 4,800 in China—more than half of whom are R&D personnel—demonstrating our strong commitment to innovation.

Over the past few years, AMD has achieved rapid growth in market performance and financial results, with our market capitalization exceeding $300 billion. We are also one of only nine companies globally to deliver an annualized shareholder return of over 30% in the last decade.

In terms of branding, we have made remarkable achievements: According to the global authoritative brand research firm BRANDZ, AMD ranked 47th among the world’s most valuable brands in 2025.

Our CEO, Lisa Su, was not only named *Time* magazine’s CEO of the Year in 2024 but was also recently elected Chair of the Board for the Semiconductor Industry Association (SIA), marking a new height in AMD’s brand value and industry influence.

Recently, you may have noticed the news: We have entered strategic partnerships with OpenAI and Oracle based on AI accelerators, signifying strong recognition from the world’s leading AI solution providers and large-scale data center operators for AMD’s AI data center solutions.

Today, AMD’s computing power touches the daily lives of billions worldwide. Whether in edge scenarios and personal devices, cloud and large-scale data centers, or high-performance computing, AMD’s computational capabilities underpin the most critical services across these key domains.

In this era of AI, AMD offers complete AI-based solutions across all three computing scenarios. Through ubiquitous AI, we are delivering our computing power to industries ranging from smart manufacturing and life sciences to aviation, telecommunications, and beyond.

Since entering the Chinese market in 1993, AMD has adhered to the principle of "In China, For China," continuously deepening our local presence. This year marks our fifth consecutive participation in the China International Import Expo (CIIE).

At this year’s CIIE, we actively responded to the national "AI+" initiative, showcasing AMD’s empowerment across industries with full-stack products and solutions spanning cloud, edge, and terminal devices. We also highlighted our latest achievements in technological innovation and ecosystem collaboration.

Additionally, we have expanded open-source cooperation by opening ROCm (Radeon Open Compute). In July, AMD inaugurated its first ROCm lab in Nanjing, China, to accelerate the deep application of ROCm in AI, biopharmaceuticals, smart manufacturing, and other critical fields.

Earlier this year, the rise of DeepSeek made AI a national-level topic, followed by trends such as advanced Mixture-of-Experts (MoE) architecture models, MCP, and Agent-driven long-context demands. Against this backdrop, the launch of AMD’s Ryzen AI Max+395 processor was perfectly timed. The Mini AI workstation based on this processor has become a preferred platform for edge-side multi-agent or AI inference and was also exhibited at CIIE.

Thanks to its unified memory architecture, this workstation can allocate up to 96GB of its 128GB memory as dedicated VRAM, effectively alleviating memory constraints when loading large models. Equipped with an integrated GPU (8060s) rivaling the performance of a discrete RTX 4070, it achieves an impressive inference speed of over 40 tokens/s when deploying **GPT-OSS-120B**—a 120-billion-parameter MoE model.

For real-world enterprise scenarios, especially AI inference applications requiring cost efficiency, privacy, space optimization, and real-time performance, the Mini AI workstation with robust edge-side deployment capabilities is the ideal choice. Small and medium-sized enterprises, departments, teams, and even individual startups can efficiently build customized AI solutions by locally deploying high-quality foundational models combined with local knowledge bases and MCP.

AMD is not just a chip company but an industry leader in high-performance and AI computing. We will continue to collaborate with partners to drive technological innovation and embrace the limitless possibilities of the AI era!

Disclaimer: All conference transcripts are real-time records and have not been reviewed by speakers.