AMD Challenges Nvidia: "New AI GPU to Launch Within This Year"

Price Competitiveness Advantage Expected
But AMD Stock Falls 3% After Announcement

AMD has thrown down the gauntlet to Nvidia, which is leading the artificial intelligence (AI) semiconductor boom. Amid a recent shortage of graphics processing units (GPUs) essential for building AI programs, AMD announced that it will release a cutting-edge AI GPU with better performance than Nvidia's within this year.

According to Bloomberg and other sources on the 13th (local time), AMD unveiled its advanced AI GPU named 'MI300X' and announced that it will be officially launched by the end of this year. AMD plans to start technology sampling with key customers in the third quarter and begin full-scale production in the fourth quarter.

Lisa Su AMD CEO
Photo by Reuters Yonhap News

To develop large language models (LLMs) for generative AI like 'ChatGPT,' auxiliary chips such as GPUs are needed to assist the central processing unit (CPU). Nvidia GPUs are the semiconductors optimized for this purpose, and based on this, Nvidia holds over 80% of the AI GPU market share.

AMD explained that the new MI300X chip is designed for large language models (LLMs) and other advanced AI models. AMD emphasized that the MI300X chip is equipped with up to 192GB of memory, allowing it to be installed in large AI models. This surpasses the 120GB memory of the competing Nvidia H100.

Lisa Su, AMD's Chief Executive Officer (CEO), stated, "The MI300X chip offers 2.4 times the memory density and more than 1.6 times the bandwidth compared to the Nvidia H100." She added, "As LLMs continue to grow larger, multiple GPUs are required to run them, but with AMD's chip, many GPUs will not be necessary."

CEO Su forecasted rapid growth in the AI semiconductor market. She said, "We are at a very, very early stage in the AI lifecycle," and added, "We expect the data center AI semiconductor market to grow from $30 billion (approximately 38 trillion KRW) this year by more than 50% annually, reaching over $150 billion by 2027."

There are expectations that AMD's announcement of the new AI GPU could deal some blows to Nvidia. This is because the performance surpasses the H100 chip, and although AMD has not disclosed pricing, it is expected to be lower than the H100. Lower GPU prices would reduce the cost of generative AI development, with the Nvidia H100 reportedly costing over $30,000.

Nevertheless, there seemed to be significant skepticism among investors about whether AMD could surpass Nvidia. In fact, AMD's stock price rose nearly 95% this year but closed down 3.61% on the day. Nvidia's stock price has risen more than 180% this year alone, and on the same day, it set a record as the first semiconductor company to surpass a market capitalization of $1 trillion.

Nvidia's AI GPUs are so hard to obtain that people say it is "harder than getting drugs." The U.S. economic media Business Insider reported on the same day that well-funded venture capitalists are directly acquiring Nvidia GPUs to support startups they invest in.

This content was produced with the assistance of AI translation services.