AMD Accelerates AI Chip Layout, Targeting NVIDIA Market Position

AMD CEO Lisa Su said at a communication and technology conference that the field of artificial intelligence (AI) is entering an unprecedented period of rapid development, and AMD is actively adjusting its strategy to accelerate the launch of high-performance AI chips, challenging the current market leader NVIDIA's position.

Su revealed that AMD plans to release a series of MI series AI chips intensively in the next few years, including MI325 later this year, MI350 next year, and MI400 in 2026. These new chips will have powerful memory capacity, up to 192GB, and integrate an astonishing 15.3 billion transistors, aiming to support training and inference tasks of large language models such as OpenAI's ChatGPT.

Su emphasized that AMD has accelerated its AI product roadmap and plans to launch new products at a rate of one generation per year to seize the super development opportunities in the current AI field. She pointed out that although NVIDIA dominates the AI chip market, AMD is rapidly narrowing the gap through technological innovation and market expansion.

In terms of financial outlook, Su expects AMD's AI chip business to achieve rapid growth. In particular, the sales of the MI300 chip at the end of 2024 are expected to reach $4.5 billion, far exceeding the total revenue of AI-related chips last year (about $100 million) and surpassing the company's earlier expectations. This growth expectation demonstrates AMD's strong momentum and market potential in the AI field.

Wall Street analysts are also optimistic about AMD's AI chip prospects. Jefferies analyst Blayne Curtis said that although the market has generally lowered its expectations for the second quarter, AMD's guidance for the MI300 chip is positive. He expects AMD's AI chip revenue to remain around $5 billion this year and support the potential revenue of $8-9 billion next year. This further validates AMD's competitiveness and growth potential in the AI field.