AMD’s Lisa Su Predicts 5 Billion AI Users by 2030
Shanghai, China — AMD Chair and CEO Dr. Lisa Su (Su Zifeng) delivered a resounding vote of confidence in China’s artificial intelligence ecosystem on Tuesday, holding the company’s first-ever AI Developer Day in Shanghai and predicting that global daily active AI users will surge from 1 billion today to 5 billion by 2030.
The event, held at the Pudong Shangri-La Hotel on May 19, drew over 2,000 registered developers — more than double the original capacity — underscoring the intense interest in AMD’s strategy for the Chinese market. It was AMD’s only AI Developer Day outside of North America, a deliberate signal of the company’s commitment to what Su called “the world’s most vibrant AI ecosystem.”
A Strategic Pivot to China
Su’s visit to China was carefully choreographed. On May 18, she met with CCPIT Chairman Ren Hongbin in Beijing, where she stated that “the Chinese market is extremely important” and expressed her desire to deepen cooperation. The following morning, she flew to Shanghai to deliver a keynote address that laid out AMD’s vision for the AI-powered future.
“China’s ecosystem is the most exciting — this is a place that truly understands open innovation,” Su said, according to Sina Finance. “This is a very special sample.”
AMD has operated in China for over 30 years, maintaining R&D centers in Beijing, Shanghai, Shenzhen, and Taipei that employ more than 4,000 engineers. The Shanghai R&D center is one of AMD’s largest globally. The company’s EPYC processors already power over 700 cloud instances at China’s leading cloud service providers, and AMD has cultivated partnerships with more than 100 software providers, startups, and universities across the country.
The 5 Billion User Prediction
In her keynote, Su projected that global daily active AI users would grow from approximately 100 million in 2020 to 1 billion in 2025, and then to 5 billion by 2030 — meaning over half the world’s population would interact with AI on a daily basis within five years.
“I have been in the tech industry for over 30 years, but never has there been a time more exciting than today,” Su told the audience, as reported by Wall Street CN.
She outlined three key transformations driving the industry: exponential growth in inference and agentic AI demand, structural shifts in computing requirements that are elevating the role of CPUs, and the transition of AI deployment from enterprise pilots to大规模普及 (mass-scale adoption).
The CPU Renaissance in AI
One of the most technically significant revelations from the event was the shifting ratio of CPUs to GPUs in AI computing nodes. Su noted that from 2022 to 2025, the typical CPU-to-GPU ratio in major AI nodes was approximately 1:4. That ratio is now rapidly moving toward 1:1, driven by the rise of multi-agent AI systems that require substantial CPU resources for task orchestration, planning, evaluation, and execution.
This structural shift plays directly to AMD’s strengths. Unlike NVIDIA, which is primarily GPU-focused, or Intel, which is CPU-focused but weaker in AI acceleration, AMD offers a full-stack portfolio spanning EPYC server CPUs, Ryzen client processors, Radeon GPUs, and Instinct AI accelerators. The company has achieved four consecutive quarters of record server CPU revenue, and Bank of America projects that the AI CPU market alone could grow from $9 billion in 2025 to $87.6 billion by 2030 — a compound annual growth rate of 57%.
Open Source as Competitive Strategy
AMD is positioning open-source software as a key differentiator against NVIDIA’s proprietary CUDA ecosystem. The company’s ROCm software platform, which supports over 3 million models through Hugging Face and China’s ModelScope community, now offers Day-0 support for leading Chinese open-source models including DeepSeek, Alibaba’s Qwen, MiniMax, Kimi, StepFun, and Xiaomi’s MiMo.
Nick Ni, AMD’s SVP of AI Business, announced that AMD is launching its first free public AI developer cloud for China, powered by Radeon GPUs. The company also unveiled an enterprise AI agent all-in-one device developed in partnership with 01.AI (Lingyiwanwu), founded by renowned AI scientist Kai-Fu Lee.
The Agentic AI Paradigm
Kai-Fu Lee, who joined Su for a fireside chat on the “New Paradigm of AI Agents,” offered a vivid characterization of the difference between Silicon Valley and China’s AI ecosystems.
“If Silicon Valley’s AI giants are like geniuses hoping to win Nobel Prizes, China’s AI ecosystem is more like a vibrant, decentralized study group,” Lee said, as reported by 21st Century Business Herald. “Everyone faces the exam together, learns from each other, builds together, and even when there is competition in business, they still share at the open-source level.”
Lee framed the evolution of AI capabilities in three stages: 2024 asked whether AI could complete a single task; 2025 asked whether AI could complete an entire workflow; and 2026 asks whether AI can operate an enterprise function. The ultimate horizon, he suggested, is whether AI can run an entire company.
Geopolitical Context
Su’s visit comes shortly after NVIDIA CEO Jensen Huang’s trip to China, highlighting the strategic importance both major GPU makers place on the Chinese market despite ongoing US-China tech tensions. AMD has previously confirmed it received approval to export AI chips to China, subject to a 15% tariff paid to the U.S. government.
By holding its only non-North American AI Developer Day in Shanghai, AMD is signaling that China is central to its global AI strategy — a deliberate counterpoint to NVIDIA’s dominance and a recognition that China’s unique open-source culture and engineering efficiency make it a critical market.
What’s Next
AMD’s deepening engagement with China’s AI ecosystem represents a bet on the country’s distinctive path forward. Constrained hardware resources have driven Chinese developers to focus on algorithmic optimization and architectural innovation rather than brute-force compute scaling — resulting in world-leading open-source models like DeepSeek and Qwen, which rank among the most downloaded globally.
As Jack Huynh, AMD’s SVP of Computing and Graphics, put it: “The winners in the next five years won’t be the teams that rent the most GPU cloud computing time, but the teams that design with efficiency as the core from day one.”
With a free AI developer cloud, new enterprise partnerships, and a growing ROCm ecosystem, AMD is betting that China’s developer community will help write the next chapter of the AI revolution — and that AMD’s open, full-stack approach will be the platform they choose to build on.