At the CES 2023 exhibition held in early 2023, AMD showcased its latest flagship accelerated processor, Instinct MI300, designed for artificial intelligence computations. This product features up to 13 small chips, with some utilizing 3D stacking technology to enhance density, totaling approximately 146 billion transistors. AMD is set to host a press conference on June 13th for the official launch of the MI300 processor.
According to MoneyDJ reports, as the launch event approaches, analysts from American institutions have expressed optimism about AMD. Harsh Kumar, an analyst from research firm Piper Sandler, has revised AMD's target stock price from $110 to $150, citing the expected high demand for AMD processors in the second half of 2023. He predicts that following the release of AMD Instinct MI300, it is expected to contribute over $200 million in revenue to AMD in the fourth quarter. The revenue from AI data center products will strongly drive AMD's performance in the third quarter, thereby dismissing concerns about lower-than-expected Q3 revenue for AMD.
After Kumar's prediction, AMD's stock price surged by 5.34%, resulting in a cumulative increase of 91.8% since the beginning of the year.
A research report published by Morgan Stanley on May 12th stated that AMD's progress in the AI field has exceeded expectations, and the upcoming MI300 processor's performance is highly anticipated based on current rumors. The firm had previously conservatively estimated AMD's AI-related revenue to reach $100 million in 2024 but has recently revised this forecast to $400 million, with a possibility of reaching as high as $1.2 billion if all goes well.
Details about the AMD Instinct MI300 accelerator processor were leaked as early as the beginning of 2023. This processor integrates 24 Zen 4 architecture CPU cores, CDNA 3 architecture GPU cores, and is surrounded by 8 HBM3 high-speed caches with a capacity of up to 128GB. With a total of 146 billion transistors, this processor is the largest-scale chip currently produced by AMD. According to reports from foreign media, Microsoft and HP may have already made use of the MI300, while future customers such as Amazon and Google are expected to purchase this processor.
AMD previously stated that the MI300 would offer a performance improvement of up to 8 times and a 5 times increase in energy efficiency for AI computations compared to the previous generation Instinct MI250. Utilizing this processor for training large-scale AI models like ChatGPT and DALL-E could reduce training time from months to weeks, resulting in savings of millions of dollars in electricity costs.