Lisa Su in Shanghai: AI Is Redefining Every Layer of Computing
At AMD's first AI Developer Conference in Shanghai, CEO Lisa Su declared that AI is redefining every layer of computing. The competitive focus is shifting from model capabilities to systems engineering and full-stack optimization. With China leading in open ecosystems, AMD is doubling down on its developer ecosystem, providing end-to-end computing from cloud to edge, powered by the ROCm open-source platform, to help developers tackle the new challenges of the agent era.
“AI is redefining every layer of computing.” That was the latest assessment of the AI industry from Dr. Lisa Su, Chair and CEO of AMD, as the company’s AI Developer Conference landed in Shanghai for the first time today.
QbitAI was invited by AMD to attend and observe the conference. After a full day of activities, it was clear that the industry is accelerating its transformation—from a focus on model capabilities to systems engineering and full-stack optimization. The challenges developers face in inference, training, and fine-tuning are becoming more concrete and engineering-driven.
What developers truly need is a practical, optimizable, and continuously evolving engineering system. This is especially true in China. Over the past two years, projects like DeepSeek and Qwen have shown that Chinese developers are not just consumers of AI applications but builders of AI infrastructure.
AMD’s response is a systematic approach to this trend.
AI Developers Need a New Engineering System
The cost of AI deployment has become a central issue. Turing Award winner David Patterson warned in early 2026 that large-scale AI deployment faces a cost crisis. Although token prices keep dropping, enterprise AI budgets are rising. The reason is a fundamental shift in AI work patterns: agents and multi-step workflows consume far more compute than single-turn Q&A.
AMD’s strategy is to provide full-stack computing from cloud to edge, centered on the ROCm open-source platform, so developers have the right tools for every deployment scenario.
Lisa Su: China Is Leading the Open Ecosystem
At the conference, Su emphasized that in the agent era, every person could have 5, 10, or even 100 agents. The structure of compute consumption has changed fundamentally; simply stacking GPUs is insufficient. A complete end-to-end combination of GPU and CPU is needed.
AMD’s vision is to be a platform company, not just a chip vendor. Software openness prevents lock-in, and hardware iteration supports the ecosystem. In China, AMD has invested for over 30 years, with Shanghai being one of its largest R&D centers. Su stated that China is not only an important market but a key part of AMD’s global roadmap. She noted that China is leading the open ecosystem, which aligns perfectly with AMD’s strategy.
Concretely, AMD continues to build the local developer community, collaborate with local open-source projects, and lower the barriers to AI development and deployment.
AI Enters Systematic Engineering Practice
The conference featured hands-on workshops and technical talks covering inference, training, edge computing, and infrastructure. Topics included optimizing inference for agent workloads, efficient RLHF training on single GPUs, MoE architecture stability, edge AI for privacy and low latency, and low-level kernel development. AMD also launched the AMD AI Developer Program – China, a membership ecosystem offering technical resources, courses, community interaction, and events.
Doubling Down on China’s Developer Ecosystem
Building a developer ecosystem requires long-term commitment. AMD is earning trust by supporting Chinese open-source models like DeepSeek and Qwen, and by actively engaging with local developers. The company’s logic is clear: let Chinese developers truly use and benefit from AMD’s platforms in their daily engineering practice. The deepest moat in the AI era is developers choosing to build on your platform—and AMD is making that happen in China.