On October 12, 2024, Advanced Micro Devices (AMD) announced plans to begin mass production of its new AI chip, the MI325X, in the fourth quarter of this year. This move aims to strengthen AMD's position in a market currently dominated by Nvidia.
During an event in San Francisco, CEO Lisa Su revealed that the company intends to launch its next series of chips, the MI350, in the second half of 2025. These new chips are expected to feature increased memory and a new architecture, which AMD claims will significantly enhance performance compared to previous models, the MI300X and MI250X.
Despite the anticipated announcements, AMD's stock fell nearly five percent, as investors were disappointed by the absence of new major clients in the cloud computing sector for these chips. Analysts noted that demand for AI processors from tech giants like Microsoft and Meta Platforms has far exceeded supply from Nvidia and AMD, allowing these semiconductor companies to sell everything they produce.
The semiconductor sector has experienced a notable rise over the past two years, with AMD's shares increasing approximately 30 percent since a recent low in early August. However, Kinngai Chan, an analyst at Summit Insights, remarked that no new clients had been announced, which contributed to the stock's decline.
Based in Santa Clara, California, AMD stated that suppliers like Super Micro Computer will begin shipping the MI325X AI chip to customers in the first quarter of 2025. The design is intended to compete with Nvidia's Blackwell architecture.
The MI325X utilizes the same architecture as the MI300X, launched last year, but includes a new type of memory that AMD claims will accelerate AI calculations. The next-generation AI chips from AMD are expected to increase pressure on Intel, which has struggled to implement a coherent strategy in this area. Intel anticipates AI chip sales exceeding $500 million in 2024.
During the event, Su also stated that the company has no plans to use chip manufacturers outside of TSMC in Taiwan for advanced manufacturing processes, which are crucial for producing high-speed AI chips. 'We would love to utilize more capacity outside of Taiwan. We are very focused on using TSMC facilities in Arizona,' Su commented.