This photo provided by Samsung Electronics Co. on May 21, 2020, shows the company’s chip plant in Pyeongtaek, south of Seoul.
SEOUL, June 27 (Korea Bizwire) – Samsung Electronics Co. is pushing for the development and production of next-generation DRAM, including high bandwidth memory (HBM), which is essential for artificial intelligence (AI) server development.
According to industry sources, the chip giant is currently shipping samples of 16GB and 12-layer 24GB HBM3, boasting the highest capacity and ultra-low power, along with 6.4Gbps performance.
Mass production has already been prepared.
Furthermore, the company has plans to release next-generation HBM3P with even higher performance to meet market demand.
While HBM represents a small fraction of the overall DRAM market, it is considered a potential breakthrough in the stagnant memory market due to its crucial role as a primary component in graphics processing units (GPUs) used by AIs like ChatGPT for data processing.
Industry sources have revealed that Samsung’s HBM3 is included in the recently unveiled MI300 superchip for AI by American chip maker AMD Inc.
Additionally, the cutting-edge memory products from Samsung are reportedly being used in the Aurora Supercomputer, established by Intel and Argonne National Laboratory.
According to market research firm TrendForce, last year’s top players in the global HBM market were SK hynix (50 percent), Samsung (40 percent) and Micron (10 percent).
“Samsung will begin supplying HBM3 to North American GPU manufacturers starting in the fourth quarter,” said analyst Kim Dong-won from KB Securities.
“Samsung’s HBM3 sales will increase from 6 percent this year to 18 percent of total DRAM sales next year.”
Kevin Lee (kevinlee@koreabizwire.com)
