Following the departure of a prominent industry figure, AMD's CEO Lisa Su arrived in Shanghai to lead the company's first AI Developer Conference in China. Eschewing a leather jacket, she opened by energizing the crowd with a question: "Are you excited?" The venue was packed, with attendees filling every aisle.
While AMD has sold chips in China for over three decades, this visit carried distinct significance. Su presented ROCm, eight hands-on workshops, and a comprehensive open-source ecosystem. She also highlighted the scale of AMD's Shanghai R&D center, noting it is one of the company's largest globally.
She shared a projection: the global number of active AI users is expected to surpass five billion in the coming years.
A significant portion of Su's address focused on a paradigm shift. She declared that the era of large language models simply providing a single answer to a single question is ending. "Each of us could have 5 agents, 10 agents, or 100 agents. Think about how much more you could get done," she stated.
Supporting these agents requires more than just large models; it necessitates robust inference, learning, and dataflow capabilities. This emphasis signals AMD's strategic bet on the future of Agentic AI.
Su also praised the Chinese developers in attendance and positioned the Chinese market as a core driver of AMD's technology roadmap.
The following is a summary of Lisa Su's keynote speech:
"Hello everyone! Let me start by asking: Are you excited? It's wonderful to be here and to see so many of you. The energy in this room is incredible—completely full, even overflowing. I am incredibly proud to be in Shanghai today.
This event is for you—our developers, our builders, and our partners. It's a chance for us to innovate together, and that is what makes this moment so special.
AI is innovating at a truly unprecedented pace. Because of this, it's more important than ever for bright minds to gather frequently to discuss the latest advancements.
At AMD, you know our mission is to push the boundaries of high-performance computing and AI. We believe this technology is a force for good, enabling us to tackle the world's most fascinating problems and its most critical challenges.
Together, we are building the foundation for the next generation of AI computing.
We are proud that AMD technology touches billions of people every day. Whether in data centers, PCs, or edge devices, AI is fundamentally redefining every layer of computing.
Our philosophy is that AI should be everywhere. It should serve everyone, every workload, and every form factor. This means we need to drive innovation at every single layer of the entire technology stack.
AMD's commitment to the Chinese market is extraordinary. China hosts one of the world's most vibrant AI ecosystems, and AMD has been deeply invested here for over thirty years.
We view China as a core part of driving our roadmap—encompassing chips, AI, software, and platform engineering. In fact, our Shanghai and China R&D centers are among our largest globally.
We do this because we firmly believe we need to be here, engaging with some of the world's brightest AI practitioners.
From AMD's perspective, our investment extends far beyond R&D; it focuses on the entire stack—from silicon and software to systems. This is why we have established AI Centers of Excellence across multiple fields in China; this is why we partner with China's largest and most influential cloud and enterprise companies; this is why we are incredibly committed to the ecosystem here.
What's truly exciting about the Chinese ecosystem is that you are leading the development of Agentic AI and enterprise AI. This is a unique strength and forms the foundation for what we at AMD believe will propel AI systems forward.
The progress we've seen in AI is incredible. Today, there are over a billion active users globally. But more exciting is our forecast: that number will exceed five billion in the next few years.
The key story supporting this user growth is having the right applications. There is no single app or workflow. You need a diverse range of models and workflows.
We have the building blocks to revolutionize everything in our lives, businesses, and interactions with AI. That's why it's such an exciting time to be in technology. With over thirty years in the industry, I've never seen a more thrilling era.
You all know the acceleration in AI over recent years. But in just the last few months, we've witnessed numerous AI successes. We've seen not just open-source models, but Agentic AI becoming immensely popular.
Enterprises are actively discussing how to leverage AI. I can tell you, every CEO, from large corporations to media firms and small businesses, is talking about using AI to impact their operations more effectively.
It's also true that AI now runs in every corner of the ecosystem—from the cloud to PCs, to the edge, and into physical systems and robotics. We see immense innovation happening, and openness is critical.
Looking ahead, a key development over the last four months is this: while traditional large language models are impressive—you ask a question, you get an answer—the overwhelming trend now, fueled by open-source and new innovations around agents, is that Agentic AI is revolutionizing how we use AI.
I hope you get a chance to see the demos from our partners outside. They show the real possibilities: what happens when you think beyond asking a model multiple questions, to how we can truly have multiple AI agents.
Each of us could have 5 agents, 10 agents, or 100 agents. Think about how much more you could get done.
Along with this, the underlying technology is changing. You no longer just need a large language model at the core; you truly need inference capabilities, learning capabilities, and dataflow capabilities. This allows for continuous reasoning, processing more data, and returning to learn—exactly what agents require.
What does this mean? Of course, everyone loves GPUs. Do you love GPUs? Yes! Do you want more GPUs? Yes! There will be more GPUs.
But GPUs will be everywhere, not just in the cloud, but across the entire ecosystem. Beyond GPUs, you also need significant CPU processing power to enable the full capability of running a complete agent.
So, you need a complete, end-to-end spectrum of computing. And this is AMD's focus: delivering complete, end-to-end compute capability.
Our commitment is to build the computing foundation for the AI era. This means having the best, most leadership roadmaps for the cloud and the edge. I must specifically note here: in China, there is so much innovation happening in Agentic AI and local AI—you are doing some of the world's best work here.
We truly value this partnership.
Our goal is also to collaborate with the ecosystem, with all the open-source projects here in China, with models, and to achieve deep integration with AMD hardware. This gives customers a real opportunity to choose the right compute, the right model, and the right capability for each of their applications.
We will discuss many things today, but a key point to understand is this: AMD's focus is purely on delivering the best compute for AI, extending our compute technology leadership, and truly engaging with the ecosystem. Crucially, we want to partner with you—with all developers, software developers, and our ROCm and developer ecosystem.
Then, we aim to bring AI to every corner of the ecosystem. This is the philosophy of delivering the right compute for the right application."
Comments