Samsung Plans to Launch AI Chip Using LPDDR Instead of HBM

During today’s Samsung Electronics shareholders' meeting, Samsung Electronics announced its plans to launch the Mach-1 AI chip, utilizing LPDDR memory, by the end of this year or the beginning of next year.

The Mach-1 chip has completed its technical validation based on FPGA and is currently in the SoC design phase. The manufacturing process of the AI chip is expected to be completed by the end of this year, with the launch of AI systems based on it slated for early next year.

According to Korean media reports, the Mach-1 chip is based on a non-traditional architecture, which can reduce the bottleneck between external memory and the computing chip to 1/8 of that of existing AI chips.

Moreover, the chip is positioned as a lightweight AI chip, opting for LPDDR memory over the more expensive HBM.

Analysts claim that while Samsung has strengths in memory, it currently trails behind its old rival SK Hynix in terms of market share and development progress in advanced HBM, and has also been outpaced by Micron in obtaining permission to supply to industry leader Nvidia. Therefore, it is crucial for Samsung to make a significant push in the AI chip market.