AMD CEO Lisa Su on the AI Supercycle and the Future of AI Chips

AMD’s chair and CEO, Dr. Lisa Su, believes her company is only at the start of an AI supercycle, driven by the rapid development of powerful AI chips. Speaking at the Goldman Sachs Communacopia & Technology Conference, Su highlighted AMD’s accelerated AI roadmap, which promises major new product releases over the next few years.

New AI Chips Target Nvidia’s Dominance

AMD is set to roll out its MI325 AI chip later this year, with the MI350 scheduled for 2025 and the MI400 expected in 2026. These chips are designed to compete directly with Nvidia, the current leader in the AI chip market. Su is confident that AMD will secure a significant share of the market for training and inference of large language models (LLMs), a key segment where Nvidia currently dominates.

Building on the Success of the MI300x

The upcoming MI chips follow the successful launch of AMD’s MI300x chip last year, which features an impressive 192GB of memory and 153 billion transistors. These capabilities make it ideal for training LLMs like OpenAI’s ChatGPT. AMD expects to generate $4.5 billion in sales from this chip alone in 2024, a huge leap from the $100 million in AI-related revenue it earned last year.

“The MI300 guidance raise [in the second quarter] is a positive vs. falling expectations (including ours),” said Jefferies analyst Blayne Curtis, especially considering earlier concerns about supply and technical challenges. AMD has reported strong demand from major customers, including Microsoft, Meta, and Oracle.

AI Powering AMD's Growth

AMD’s AI chip sales are fueling rapid growth for the company. In the second quarter, sales and earnings rose 9% and 19% year-over-year, respectively. AMD expects third-quarter revenue to grow by 16% compared to the same period last year, and 15% sequentially.

“AI is a much larger cycle than I would have expected five years ago,” Su said. “We’re making big bets for the next five years.”

The AI Supercycle: A New Era of Innovation

The “AI Supercycle” refers to a period of accelerated growth in AI technology, driven by the increasing power of machine learning algorithms, big data, and AI adoption across industries. This supercycle follows earlier waves of AI innovation, such as the development of foundational algorithms in the 1950s and 60s, and the resurgence of AI with deep learning in the 2000s.

Experts predict that the next decade will bring massive innovation and adoption of AI. Andrew Ng, founder of Deeplearning.ai, expects AI to transform every major industry. Kai-Fu Lee, CEO of Sinovation Ventures, has called AI “the new electricity,” essential to every industry and job.

As AI continues to evolve, it is expected to revolutionize industries like healthcare, finance, and transportation, while also changing how people work and live. The AI supercycle promises a new era of technological advancement that could reshape society.