According to a recent report by Sedaily, Samsung Electronics has unveiled its plans to introduce its own AI accelerator chip, dubbed “Mach-1,” in early 2025. This could mark a significant stride for the company as it endeavors to compete with industry giants like Nvidia. However, Samsung has explicitly stated that it harbors no intention to directly challenge Nvidia’s top-tier AI solutions, such as the H100, B100, or B200 series.
The Mach-1 chip from Samsung is internally developed and tailored to excel in demanding tasks, specifically crafted for edge computing applications. Unlike its high-performance counterparts, Mach-1 is geared towards delivering accelerated capabilities optimized for edge scenarios.
Employing an application-specific integrated circuit (ASIC) design and incorporating LPDDR memory, Mach-1 aims to substantially reduce memory bandwidth requirements during the inference process, potentially yielding efficiency gains. However, Samsung has not disclosed the specifics of how this efficiency is achieved.
Kye Hyun Kyung, director of Samsung Electronics‘ device solutions department, has affirmed that the Mach-1 chip’s design has undergone validation using field-programmable gate array (FPGA) technology and is currently in the final stages of physical implementation as part of a system-on-chip (SoC) configuration. He has assured stakeholders that the chip will be fully prepared for deployment by year-end, with plans for an artificial intelligence system powered by Mach-1 slated for an early launch next year.
Samsung’s Mach-1 chip is strategically positioned to address the burgeoning demand for AI solutions in edge computing, prioritizing factors such as low power consumption, compact form factor, and cost-effectiveness. While Mach-1 is not directly positioned against high-performance AI processors like AWS Trainium or Nvidia’s H100 series, it may encounter competition within the inference-oriented market segment, potentially from offerings such as AWS Inferentia.
In addition to Mach-1, Kye Hyun Kyung has disclosed Samsung’s broader involvement in AI semiconductor development, highlighting the establishment of a specialized research laboratory in Silicon Valley solely dedicated to advancing general artificial intelligence (AGI). This initiative underscores Samsung’s commitment to meeting future processing requirements for AGI systems through the development of innovative processor and memory technologies.